1 /* 2 * libjingle 3 * Copyright 2004--2005, Google Inc. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions are met: 7 * 8 * 1. Redistributions of source code must retain the above copyright notice, 9 * this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright notice, 11 * this list of conditions and the following disclaimer in the documentation 12 * and/or other materials provided with the distribution. 13 * 3. The name of the author may not be used to endorse or promote products 14 * derived from this software without specific prior written permission. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR IMPLIED 17 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 18 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO 19 * EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 20 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, 21 * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; 22 * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, 23 * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR 24 * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF 25 * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 26 */ 27 28 #include "talk/base/virtualsocketserver.h" 29 30 #include <errno.h> 31 32 #include <algorithm> 33 #include <cmath> 34 #include <map> 35 #include <vector> 36 37 #include "talk/base/common.h" 38 #include "talk/base/logging.h" 39 #include "talk/base/physicalsocketserver.h" 40 #include "talk/base/socketaddresspair.h" 41 #include "talk/base/thread.h" 42 #include "talk/base/timeutils.h" 43 44 namespace talk_base { 45 #ifdef WIN32 46 const in_addr kInitialNextIPv4 = { {0x01, 0, 0, 0} }; 47 #else 48 // This value is entirely arbitrary, hence the lack of concern about endianness. 49 const in_addr kInitialNextIPv4 = { 0x01000000 }; 50 #endif 51 // Starts at ::2 so as to not cause confusion with ::1. 52 const in6_addr kInitialNextIPv6 = { { { 53 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2 54 } } }; 55 56 const uint16 kFirstEphemeralPort = 49152; 57 const uint16 kLastEphemeralPort = 65535; 58 const uint16 kEphemeralPortCount = kLastEphemeralPort - kFirstEphemeralPort + 1; 59 const uint32 kDefaultNetworkCapacity = 64 * 1024; 60 const uint32 kDefaultTcpBufferSize = 32 * 1024; 61 62 const uint32 UDP_HEADER_SIZE = 28; // IP + UDP headers 63 const uint32 TCP_HEADER_SIZE = 40; // IP + TCP headers 64 const uint32 TCP_MSS = 1400; // Maximum segment size 65 66 // Note: The current algorithm doesn't work for sample sizes smaller than this. 67 const int NUM_SAMPLES = 1000; 68 69 enum { 70 MSG_ID_PACKET, 71 MSG_ID_CONNECT, 72 MSG_ID_DISCONNECT, 73 }; 74 75 // Packets are passed between sockets as messages. We copy the data just like 76 // the kernel does. 77 class Packet : public MessageData { 78 public: 79 Packet(const char* data, size_t size, const SocketAddress& from) 80 : size_(size), consumed_(0), from_(from) { 81 ASSERT(NULL != data); 82 data_ = new char[size_]; 83 std::memcpy(data_, data, size_); 84 } 85 86 virtual ~Packet() { 87 delete[] data_; 88 } 89 90 const char* data() const { return data_ + consumed_; } 91 size_t size() const { return size_ - consumed_; } 92 const SocketAddress& from() const { return from_; } 93 94 // Remove the first size bytes from the data. 95 void Consume(size_t size) { 96 ASSERT(size + consumed_ < size_); 97 consumed_ += size; 98 } 99 100 private: 101 char* data_; 102 size_t size_, consumed_; 103 SocketAddress from_; 104 }; 105 106 struct MessageAddress : public MessageData { 107 explicit MessageAddress(const SocketAddress& a) : addr(a) { } 108 SocketAddress addr; 109 }; 110 111 // Implements the socket interface using the virtual network. Packets are 112 // passed as messages using the message queue of the socket server. 113 class VirtualSocket : public AsyncSocket, public MessageHandler { 114 public: 115 VirtualSocket(VirtualSocketServer* server, int family, int type, bool async) 116 : server_(server), family_(family), type_(type), async_(async), 117 state_(CS_CLOSED), error_(0), listen_queue_(NULL), 118 write_enabled_(false), 119 network_size_(0), recv_buffer_size_(0), bound_(false), was_any_(false) { 120 ASSERT((type_ == SOCK_DGRAM) || (type_ == SOCK_STREAM)); 121 ASSERT(async_ || (type_ != SOCK_STREAM)); // We only support async streams 122 } 123 124 virtual ~VirtualSocket() { 125 Close(); 126 127 for (RecvBuffer::iterator it = recv_buffer_.begin(); 128 it != recv_buffer_.end(); ++it) { 129 delete *it; 130 } 131 } 132 133 virtual SocketAddress GetLocalAddress() const { 134 return local_addr_; 135 } 136 137 virtual SocketAddress GetRemoteAddress() const { 138 return remote_addr_; 139 } 140 141 // Used by server sockets to set the local address without binding. 142 void SetLocalAddress(const SocketAddress& addr) { 143 local_addr_ = addr; 144 } 145 146 virtual int Bind(const SocketAddress& addr) { 147 if (!local_addr_.IsNil()) { 148 error_ = EINVAL; 149 return -1; 150 } 151 local_addr_ = addr; 152 int result = server_->Bind(this, &local_addr_); 153 if (result != 0) { 154 local_addr_.Clear(); 155 error_ = EADDRINUSE; 156 } else { 157 bound_ = true; 158 was_any_ = addr.IsAnyIP(); 159 } 160 return result; 161 } 162 163 virtual int Connect(const SocketAddress& addr) { 164 return InitiateConnect(addr, true); 165 } 166 167 virtual int Close() { 168 if (!local_addr_.IsNil() && bound_) { 169 // Remove from the binding table. 170 server_->Unbind(local_addr_, this); 171 bound_ = false; 172 } 173 174 if (SOCK_STREAM == type_) { 175 // Cancel pending sockets 176 if (listen_queue_) { 177 while (!listen_queue_->empty()) { 178 SocketAddress addr = listen_queue_->front(); 179 180 // Disconnect listening socket. 181 server_->Disconnect(server_->LookupBinding(addr)); 182 listen_queue_->pop_front(); 183 } 184 delete listen_queue_; 185 listen_queue_ = NULL; 186 } 187 // Disconnect stream sockets 188 if (CS_CONNECTED == state_) { 189 // Disconnect remote socket, check if it is a child of a server socket. 190 VirtualSocket* socket = 191 server_->LookupConnection(local_addr_, remote_addr_); 192 if (!socket) { 193 // Not a server socket child, then see if it is bound. 194 // TODO: If this is indeed a server socket that has no 195 // children this will cause the server socket to be 196 // closed. This might lead to unexpected results, how to fix this? 197 socket = server_->LookupBinding(remote_addr_); 198 } 199 server_->Disconnect(socket); 200 201 // Remove mapping for both directions. 202 server_->RemoveConnection(remote_addr_, local_addr_); 203 server_->RemoveConnection(local_addr_, remote_addr_); 204 } 205 // Cancel potential connects 206 MessageList msgs; 207 if (server_->msg_queue_) { 208 server_->msg_queue_->Clear(this, MSG_ID_CONNECT, &msgs); 209 } 210 for (MessageList::iterator it = msgs.begin(); it != msgs.end(); ++it) { 211 ASSERT(NULL != it->pdata); 212 MessageAddress* data = static_cast<MessageAddress*>(it->pdata); 213 214 // Lookup remote side. 215 VirtualSocket* socket = server_->LookupConnection(local_addr_, 216 data->addr); 217 if (socket) { 218 // Server socket, remote side is a socket retreived by 219 // accept. Accepted sockets are not bound so we will not 220 // find it by looking in the bindings table. 221 server_->Disconnect(socket); 222 server_->RemoveConnection(local_addr_, data->addr); 223 } else { 224 server_->Disconnect(server_->LookupBinding(data->addr)); 225 } 226 delete data; 227 } 228 // Clear incoming packets and disconnect messages 229 if (server_->msg_queue_) { 230 server_->msg_queue_->Clear(this); 231 } 232 } 233 234 state_ = CS_CLOSED; 235 local_addr_.Clear(); 236 remote_addr_.Clear(); 237 return 0; 238 } 239 240 virtual int Send(const void *pv, size_t cb) { 241 if (CS_CONNECTED != state_) { 242 error_ = ENOTCONN; 243 return -1; 244 } 245 if (SOCK_DGRAM == type_) { 246 return SendUdp(pv, cb, remote_addr_); 247 } else { 248 return SendTcp(pv, cb); 249 } 250 } 251 252 virtual int SendTo(const void *pv, size_t cb, const SocketAddress& addr) { 253 if (SOCK_DGRAM == type_) { 254 return SendUdp(pv, cb, addr); 255 } else { 256 if (CS_CONNECTED != state_) { 257 error_ = ENOTCONN; 258 return -1; 259 } 260 return SendTcp(pv, cb); 261 } 262 } 263 264 virtual int Recv(void *pv, size_t cb) { 265 SocketAddress addr; 266 return RecvFrom(pv, cb, &addr); 267 } 268 269 virtual int RecvFrom(void *pv, size_t cb, SocketAddress *paddr) { 270 // If we don't have a packet, then either error or wait for one to arrive. 271 if (recv_buffer_.empty()) { 272 if (async_) { 273 error_ = EAGAIN; 274 return -1; 275 } 276 while (recv_buffer_.empty()) { 277 Message msg; 278 server_->msg_queue_->Get(&msg); 279 server_->msg_queue_->Dispatch(&msg); 280 } 281 } 282 283 // Return the packet at the front of the queue. 284 Packet* packet = recv_buffer_.front(); 285 size_t data_read = _min(cb, packet->size()); 286 std::memcpy(pv, packet->data(), data_read); 287 *paddr = packet->from(); 288 289 if (data_read < packet->size()) { 290 packet->Consume(data_read); 291 } else { 292 recv_buffer_.pop_front(); 293 delete packet; 294 } 295 296 if (SOCK_STREAM == type_) { 297 bool was_full = (recv_buffer_size_ == server_->recv_buffer_capacity_); 298 recv_buffer_size_ -= data_read; 299 if (was_full) { 300 VirtualSocket* sender = server_->LookupBinding(remote_addr_); 301 ASSERT(NULL != sender); 302 server_->SendTcp(sender); 303 } 304 } 305 306 return static_cast<int>(data_read); 307 } 308 309 virtual int Listen(int backlog) { 310 ASSERT(SOCK_STREAM == type_); 311 ASSERT(CS_CLOSED == state_); 312 if (local_addr_.IsNil()) { 313 error_ = EINVAL; 314 return -1; 315 } 316 ASSERT(NULL == listen_queue_); 317 listen_queue_ = new ListenQueue; 318 state_ = CS_CONNECTING; 319 return 0; 320 } 321 322 virtual VirtualSocket* Accept(SocketAddress *paddr) { 323 if (NULL == listen_queue_) { 324 error_ = EINVAL; 325 return NULL; 326 } 327 while (!listen_queue_->empty()) { 328 VirtualSocket* socket = new VirtualSocket(server_, AF_INET, type_, 329 async_); 330 331 // Set the new local address to the same as this server socket. 332 socket->SetLocalAddress(local_addr_); 333 // Sockets made from a socket that 'was Any' need to inherit that. 334 socket->set_was_any(was_any_); 335 SocketAddress remote_addr(listen_queue_->front()); 336 int result = socket->InitiateConnect(remote_addr, false); 337 listen_queue_->pop_front(); 338 if (result != 0) { 339 delete socket; 340 continue; 341 } 342 socket->CompleteConnect(remote_addr, false); 343 if (paddr) { 344 *paddr = remote_addr; 345 } 346 return socket; 347 } 348 error_ = EWOULDBLOCK; 349 return NULL; 350 } 351 352 virtual int GetError() const { 353 return error_; 354 } 355 356 virtual void SetError(int error) { 357 error_ = error; 358 } 359 360 virtual ConnState GetState() const { 361 return state_; 362 } 363 364 virtual int GetOption(Option opt, int* value) { 365 OptionsMap::const_iterator it = options_map_.find(opt); 366 if (it == options_map_.end()) { 367 return -1; 368 } 369 *value = it->second; 370 return 0; // 0 is success to emulate getsockopt() 371 } 372 373 virtual int SetOption(Option opt, int value) { 374 options_map_[opt] = value; 375 return 0; // 0 is success to emulate setsockopt() 376 } 377 378 virtual int EstimateMTU(uint16* mtu) { 379 if (CS_CONNECTED != state_) 380 return ENOTCONN; 381 else 382 return 65536; 383 } 384 385 void OnMessage(Message *pmsg) { 386 if (pmsg->message_id == MSG_ID_PACKET) { 387 //ASSERT(!local_addr_.IsAny()); 388 ASSERT(NULL != pmsg->pdata); 389 Packet* packet = static_cast<Packet*>(pmsg->pdata); 390 391 recv_buffer_.push_back(packet); 392 393 if (async_) { 394 SignalReadEvent(this); 395 } 396 } else if (pmsg->message_id == MSG_ID_CONNECT) { 397 ASSERT(NULL != pmsg->pdata); 398 MessageAddress* data = static_cast<MessageAddress*>(pmsg->pdata); 399 if (listen_queue_ != NULL) { 400 listen_queue_->push_back(data->addr); 401 if (async_) { 402 SignalReadEvent(this); 403 } 404 } else if ((SOCK_STREAM == type_) && (CS_CONNECTING == state_)) { 405 CompleteConnect(data->addr, true); 406 } else { 407 LOG(LS_VERBOSE) << "Socket at " << local_addr_ << " is not listening"; 408 server_->Disconnect(server_->LookupBinding(data->addr)); 409 } 410 delete data; 411 } else if (pmsg->message_id == MSG_ID_DISCONNECT) { 412 ASSERT(SOCK_STREAM == type_); 413 if (CS_CLOSED != state_) { 414 int error = (CS_CONNECTING == state_) ? ECONNREFUSED : 0; 415 state_ = CS_CLOSED; 416 remote_addr_.Clear(); 417 if (async_) { 418 SignalCloseEvent(this, error); 419 } 420 } 421 } else { 422 ASSERT(false); 423 } 424 } 425 426 bool was_any() { return was_any_; } 427 void set_was_any(bool was_any) { was_any_ = was_any; } 428 429 private: 430 struct NetworkEntry { 431 size_t size; 432 uint32 done_time; 433 }; 434 435 typedef std::deque<SocketAddress> ListenQueue; 436 typedef std::deque<NetworkEntry> NetworkQueue; 437 typedef std::vector<char> SendBuffer; 438 typedef std::list<Packet*> RecvBuffer; 439 typedef std::map<Option, int> OptionsMap; 440 441 int InitiateConnect(const SocketAddress& addr, bool use_delay) { 442 if (!remote_addr_.IsNil()) { 443 error_ = (CS_CONNECTED == state_) ? EISCONN : EINPROGRESS; 444 return -1; 445 } 446 if (local_addr_.IsNil()) { 447 // If there's no local address set, grab a random one in the correct AF. 448 int result = 0; 449 if (addr.ipaddr().family() == AF_INET) { 450 result = Bind(SocketAddress("0.0.0.0", 0)); 451 } else if (addr.ipaddr().family() == AF_INET6) { 452 result = Bind(SocketAddress("::", 0)); 453 } 454 if (result != 0) { 455 return result; 456 } 457 } 458 if (type_ == SOCK_DGRAM) { 459 remote_addr_ = addr; 460 state_ = CS_CONNECTED; 461 } else { 462 int result = server_->Connect(this, addr, use_delay); 463 if (result != 0) { 464 error_ = EHOSTUNREACH; 465 return -1; 466 } 467 state_ = CS_CONNECTING; 468 } 469 return 0; 470 } 471 472 void CompleteConnect(const SocketAddress& addr, bool notify) { 473 ASSERT(CS_CONNECTING == state_); 474 remote_addr_ = addr; 475 state_ = CS_CONNECTED; 476 server_->AddConnection(remote_addr_, local_addr_, this); 477 if (async_ && notify) { 478 SignalConnectEvent(this); 479 } 480 } 481 482 int SendUdp(const void* pv, size_t cb, const SocketAddress& addr) { 483 // If we have not been assigned a local port, then get one. 484 if (local_addr_.IsNil()) { 485 local_addr_ = EmptySocketAddressWithFamily(addr.ipaddr().family()); 486 int result = server_->Bind(this, &local_addr_); 487 if (result != 0) { 488 local_addr_.Clear(); 489 error_ = EADDRINUSE; 490 return result; 491 } 492 } 493 494 // Send the data in a message to the appropriate socket. 495 return server_->SendUdp(this, static_cast<const char*>(pv), cb, addr); 496 } 497 498 int SendTcp(const void* pv, size_t cb) { 499 size_t capacity = server_->send_buffer_capacity_ - send_buffer_.size(); 500 if (0 == capacity) { 501 write_enabled_ = true; 502 error_ = EWOULDBLOCK; 503 return -1; 504 } 505 size_t consumed = _min(cb, capacity); 506 const char* cpv = static_cast<const char*>(pv); 507 send_buffer_.insert(send_buffer_.end(), cpv, cpv + consumed); 508 server_->SendTcp(this); 509 return static_cast<int>(consumed); 510 } 511 512 VirtualSocketServer* server_; 513 int family_; 514 int type_; 515 bool async_; 516 ConnState state_; 517 int error_; 518 SocketAddress local_addr_; 519 SocketAddress remote_addr_; 520 521 // Pending sockets which can be Accepted 522 ListenQueue* listen_queue_; 523 524 // Data which tcp has buffered for sending 525 SendBuffer send_buffer_; 526 bool write_enabled_; 527 528 // Critical section to protect the recv_buffer and queue_ 529 CriticalSection crit_; 530 531 // Network model that enforces bandwidth and capacity constraints 532 NetworkQueue network_; 533 size_t network_size_; 534 535 // Data which has been received from the network 536 RecvBuffer recv_buffer_; 537 // The amount of data which is in flight or in recv_buffer_ 538 size_t recv_buffer_size_; 539 540 // Is this socket bound? 541 bool bound_; 542 543 // When we bind a socket to Any, VSS's Bind gives it another address. For 544 // dual-stack sockets, we want to distinguish between sockets that were 545 // explicitly given a particular address and sockets that had one picked 546 // for them by VSS. 547 bool was_any_; 548 549 // Store the options that are set 550 OptionsMap options_map_; 551 552 friend class VirtualSocketServer; 553 }; 554 555 VirtualSocketServer::VirtualSocketServer(SocketServer* ss) 556 : server_(ss), server_owned_(false), msg_queue_(NULL), stop_on_idle_(false), 557 network_delay_(Time()), next_ipv4_(kInitialNextIPv4), 558 next_ipv6_(kInitialNextIPv6), next_port_(kFirstEphemeralPort), 559 bindings_(new AddressMap()), connections_(new ConnectionMap()), 560 bandwidth_(0), network_capacity_(kDefaultNetworkCapacity), 561 send_buffer_capacity_(kDefaultTcpBufferSize), 562 recv_buffer_capacity_(kDefaultTcpBufferSize), 563 delay_mean_(0), delay_stddev_(0), delay_samples_(NUM_SAMPLES), 564 delay_dist_(NULL), drop_prob_(0.0) { 565 if (!server_) { 566 server_ = new PhysicalSocketServer(); 567 server_owned_ = true; 568 } 569 UpdateDelayDistribution(); 570 } 571 572 VirtualSocketServer::~VirtualSocketServer() { 573 delete bindings_; 574 delete connections_; 575 delete delay_dist_; 576 if (server_owned_) { 577 delete server_; 578 } 579 } 580 581 IPAddress VirtualSocketServer::GetNextIP(int family) { 582 if (family == AF_INET) { 583 IPAddress next_ip(next_ipv4_); 584 next_ipv4_.s_addr = 585 HostToNetwork32(NetworkToHost32(next_ipv4_.s_addr) + 1); 586 return next_ip; 587 } else if (family == AF_INET6) { 588 IPAddress next_ip(next_ipv6_); 589 uint32* as_ints = reinterpret_cast<uint32*>(&next_ipv6_.s6_addr); 590 as_ints[3] += 1; 591 return next_ip; 592 } 593 return IPAddress(); 594 } 595 596 uint16 VirtualSocketServer::GetNextPort() { 597 uint16 port = next_port_; 598 if (next_port_ < kLastEphemeralPort) { 599 ++next_port_; 600 } else { 601 next_port_ = kFirstEphemeralPort; 602 } 603 return port; 604 } 605 606 Socket* VirtualSocketServer::CreateSocket(int type) { 607 return CreateSocket(AF_INET, type); 608 } 609 610 Socket* VirtualSocketServer::CreateSocket(int family, int type) { 611 return CreateSocketInternal(family, type); 612 } 613 614 AsyncSocket* VirtualSocketServer::CreateAsyncSocket(int type) { 615 return CreateAsyncSocket(AF_INET, type); 616 } 617 618 AsyncSocket* VirtualSocketServer::CreateAsyncSocket(int family, int type) { 619 return CreateSocketInternal(family, type); 620 } 621 622 VirtualSocket* VirtualSocketServer::CreateSocketInternal(int family, int type) { 623 return new VirtualSocket(this, family, type, true); 624 } 625 626 void VirtualSocketServer::SetMessageQueue(MessageQueue* msg_queue) { 627 msg_queue_ = msg_queue; 628 if (msg_queue_) { 629 msg_queue_->SignalQueueDestroyed.connect(this, 630 &VirtualSocketServer::OnMessageQueueDestroyed); 631 } 632 } 633 634 bool VirtualSocketServer::Wait(int cmsWait, bool process_io) { 635 ASSERT(msg_queue_ == Thread::Current()); 636 if (stop_on_idle_ && Thread::Current()->empty()) { 637 return false; 638 } 639 return socketserver()->Wait(cmsWait, process_io); 640 } 641 642 void VirtualSocketServer::WakeUp() { 643 socketserver()->WakeUp(); 644 } 645 646 bool VirtualSocketServer::ProcessMessagesUntilIdle() { 647 ASSERT(msg_queue_ == Thread::Current()); 648 stop_on_idle_ = true; 649 while (!msg_queue_->empty()) { 650 Message msg; 651 if (msg_queue_->Get(&msg, kForever)) { 652 msg_queue_->Dispatch(&msg); 653 } 654 } 655 stop_on_idle_ = false; 656 return !msg_queue_->IsQuitting(); 657 } 658 659 int VirtualSocketServer::Bind(VirtualSocket* socket, 660 const SocketAddress& addr) { 661 ASSERT(NULL != socket); 662 // Address must be completely specified at this point 663 ASSERT(!IPIsUnspec(addr.ipaddr())); 664 ASSERT(addr.port() != 0); 665 666 // Normalize the address (turns v6-mapped addresses into v4-addresses). 667 SocketAddress normalized(addr.ipaddr().Normalized(), addr.port()); 668 669 AddressMap::value_type entry(normalized, socket); 670 return bindings_->insert(entry).second ? 0 : -1; 671 } 672 673 int VirtualSocketServer::Bind(VirtualSocket* socket, SocketAddress* addr) { 674 ASSERT(NULL != socket); 675 676 if (IPIsAny(addr->ipaddr())) { 677 addr->SetIP(GetNextIP(addr->ipaddr().family())); 678 } else if (!IPIsUnspec(addr->ipaddr())) { 679 addr->SetIP(addr->ipaddr().Normalized()); 680 } else { 681 ASSERT(false); 682 } 683 684 if (addr->port() == 0) { 685 for (int i = 0; i < kEphemeralPortCount; ++i) { 686 addr->SetPort(GetNextPort()); 687 if (bindings_->find(*addr) == bindings_->end()) { 688 break; 689 } 690 } 691 } 692 693 return Bind(socket, *addr); 694 } 695 696 VirtualSocket* VirtualSocketServer::LookupBinding(const SocketAddress& addr) { 697 SocketAddress normalized(addr.ipaddr().Normalized(), 698 addr.port()); 699 AddressMap::iterator it = bindings_->find(normalized); 700 return (bindings_->end() != it) ? it->second : NULL; 701 } 702 703 int VirtualSocketServer::Unbind(const SocketAddress& addr, 704 VirtualSocket* socket) { 705 SocketAddress normalized(addr.ipaddr().Normalized(), 706 addr.port()); 707 ASSERT((*bindings_)[normalized] == socket); 708 bindings_->erase(bindings_->find(normalized)); 709 return 0; 710 } 711 712 void VirtualSocketServer::AddConnection(const SocketAddress& local, 713 const SocketAddress& remote, 714 VirtualSocket* remote_socket) { 715 // Add this socket pair to our routing table. This will allow 716 // multiple clients to connect to the same server address. 717 SocketAddress local_normalized(local.ipaddr().Normalized(), 718 local.port()); 719 SocketAddress remote_normalized(remote.ipaddr().Normalized(), 720 remote.port()); 721 SocketAddressPair address_pair(local_normalized, remote_normalized); 722 connections_->insert(std::pair<SocketAddressPair, 723 VirtualSocket*>(address_pair, remote_socket)); 724 } 725 726 VirtualSocket* VirtualSocketServer::LookupConnection( 727 const SocketAddress& local, 728 const SocketAddress& remote) { 729 SocketAddress local_normalized(local.ipaddr().Normalized(), 730 local.port()); 731 SocketAddress remote_normalized(remote.ipaddr().Normalized(), 732 remote.port()); 733 SocketAddressPair address_pair(local_normalized, remote_normalized); 734 ConnectionMap::iterator it = connections_->find(address_pair); 735 return (connections_->end() != it) ? it->second : NULL; 736 } 737 738 void VirtualSocketServer::RemoveConnection(const SocketAddress& local, 739 const SocketAddress& remote) { 740 SocketAddress local_normalized(local.ipaddr().Normalized(), 741 local.port()); 742 SocketAddress remote_normalized(remote.ipaddr().Normalized(), 743 remote.port()); 744 SocketAddressPair address_pair(local_normalized, remote_normalized); 745 connections_->erase(address_pair); 746 } 747 748 static double Random() { 749 return static_cast<double>(rand()) / RAND_MAX; 750 } 751 752 int VirtualSocketServer::Connect(VirtualSocket* socket, 753 const SocketAddress& remote_addr, 754 bool use_delay) { 755 uint32 delay = use_delay ? GetRandomTransitDelay() : 0; 756 VirtualSocket* remote = LookupBinding(remote_addr); 757 if (!CanInteractWith(socket, remote)) { 758 LOG(LS_INFO) << "Address family mismatch between " 759 << socket->GetLocalAddress() << " and " << remote_addr; 760 return -1; 761 } 762 if (remote != NULL) { 763 SocketAddress addr = socket->GetLocalAddress(); 764 msg_queue_->PostDelayed(delay, remote, MSG_ID_CONNECT, 765 new MessageAddress(addr)); 766 } else { 767 LOG(LS_INFO) << "No one listening at " << remote_addr; 768 msg_queue_->PostDelayed(delay, socket, MSG_ID_DISCONNECT); 769 } 770 return 0; 771 } 772 773 bool VirtualSocketServer::Disconnect(VirtualSocket* socket) { 774 if (socket) { 775 // Remove the mapping. 776 msg_queue_->Post(socket, MSG_ID_DISCONNECT); 777 return true; 778 } 779 return false; 780 } 781 782 int VirtualSocketServer::SendUdp(VirtualSocket* socket, 783 const char* data, size_t data_size, 784 const SocketAddress& remote_addr) { 785 // See if we want to drop this packet. 786 if (Random() < drop_prob_) { 787 LOG(LS_VERBOSE) << "Dropping packet: bad luck"; 788 return static_cast<int>(data_size); 789 } 790 791 VirtualSocket* recipient = LookupBinding(remote_addr); 792 if (!recipient) { 793 // Make a fake recipient for address family checking. 794 scoped_ptr<VirtualSocket> dummy_socket( 795 CreateSocketInternal(AF_INET, SOCK_DGRAM)); 796 dummy_socket->SetLocalAddress(remote_addr); 797 if (!CanInteractWith(socket, dummy_socket.get())) { 798 LOG(LS_VERBOSE) << "Incompatible address families: " 799 << socket->GetLocalAddress() << " and " << remote_addr; 800 return -1; 801 } 802 LOG(LS_VERBOSE) << "No one listening at " << remote_addr; 803 return static_cast<int>(data_size); 804 } 805 806 if (!CanInteractWith(socket, recipient)) { 807 LOG(LS_VERBOSE) << "Incompatible address families: " 808 << socket->GetLocalAddress() << " and " << remote_addr; 809 return -1; 810 } 811 812 CritScope cs(&socket->crit_); 813 814 uint32 cur_time = Time(); 815 PurgeNetworkPackets(socket, cur_time); 816 817 // Determine whether we have enough bandwidth to accept this packet. To do 818 // this, we need to update the send queue. Once we know it's current size, 819 // we know whether we can fit this packet. 820 // 821 // NOTE: There are better algorithms for maintaining such a queue (such as 822 // "Derivative Random Drop"); however, this algorithm is a more accurate 823 // simulation of what a normal network would do. 824 825 size_t packet_size = data_size + UDP_HEADER_SIZE; 826 if (socket->network_size_ + packet_size > network_capacity_) { 827 LOG(LS_VERBOSE) << "Dropping packet: network capacity exceeded"; 828 return static_cast<int>(data_size); 829 } 830 831 AddPacketToNetwork(socket, recipient, cur_time, data, data_size, 832 UDP_HEADER_SIZE, false); 833 834 return static_cast<int>(data_size); 835 } 836 837 void VirtualSocketServer::SendTcp(VirtualSocket* socket) { 838 // TCP can't send more data than will fill up the receiver's buffer. 839 // We track the data that is in the buffer plus data in flight using the 840 // recipient's recv_buffer_size_. Anything beyond that must be stored in the 841 // sender's buffer. We will trigger the buffered data to be sent when data 842 // is read from the recv_buffer. 843 844 // Lookup the local/remote pair in the connections table. 845 VirtualSocket* recipient = LookupConnection(socket->local_addr_, 846 socket->remote_addr_); 847 if (!recipient) { 848 LOG(LS_VERBOSE) << "Sending data to no one."; 849 return; 850 } 851 852 CritScope cs(&socket->crit_); 853 854 uint32 cur_time = Time(); 855 PurgeNetworkPackets(socket, cur_time); 856 857 while (true) { 858 size_t available = recv_buffer_capacity_ - recipient->recv_buffer_size_; 859 size_t max_data_size = _min<size_t>(available, TCP_MSS - TCP_HEADER_SIZE); 860 size_t data_size = _min(socket->send_buffer_.size(), max_data_size); 861 if (0 == data_size) 862 break; 863 864 AddPacketToNetwork(socket, recipient, cur_time, &socket->send_buffer_[0], 865 data_size, TCP_HEADER_SIZE, true); 866 recipient->recv_buffer_size_ += data_size; 867 868 size_t new_buffer_size = socket->send_buffer_.size() - data_size; 869 // Avoid undefined access beyond the last element of the vector. 870 // This only happens when new_buffer_size is 0. 871 if (data_size < socket->send_buffer_.size()) { 872 // memmove is required for potentially overlapping source/destination. 873 memmove(&socket->send_buffer_[0], &socket->send_buffer_[data_size], 874 new_buffer_size); 875 } 876 socket->send_buffer_.resize(new_buffer_size); 877 } 878 879 if (socket->write_enabled_ 880 && (socket->send_buffer_.size() < send_buffer_capacity_)) { 881 socket->write_enabled_ = false; 882 socket->SignalWriteEvent(socket); 883 } 884 } 885 886 void VirtualSocketServer::AddPacketToNetwork(VirtualSocket* sender, 887 VirtualSocket* recipient, 888 uint32 cur_time, 889 const char* data, 890 size_t data_size, 891 size_t header_size, 892 bool ordered) { 893 VirtualSocket::NetworkEntry entry; 894 entry.size = data_size + header_size; 895 896 sender->network_size_ += entry.size; 897 uint32 send_delay = SendDelay(static_cast<uint32>(sender->network_size_)); 898 entry.done_time = cur_time + send_delay; 899 sender->network_.push_back(entry); 900 901 // Find the delay for crossing the many virtual hops of the network. 902 uint32 transit_delay = GetRandomTransitDelay(); 903 904 // Post the packet as a message to be delivered (on our own thread) 905 Packet* p = new Packet(data, data_size, sender->local_addr_); 906 uint32 ts = TimeAfter(send_delay + transit_delay); 907 if (ordered) { 908 // Ensure that new packets arrive after previous ones 909 // TODO: consider ordering on a per-socket basis, since this 910 // introduces artifical delay. 911 ts = TimeMax(ts, network_delay_); 912 } 913 msg_queue_->PostAt(ts, recipient, MSG_ID_PACKET, p); 914 network_delay_ = TimeMax(ts, network_delay_); 915 } 916 917 void VirtualSocketServer::PurgeNetworkPackets(VirtualSocket* socket, 918 uint32 cur_time) { 919 while (!socket->network_.empty() && 920 (socket->network_.front().done_time <= cur_time)) { 921 ASSERT(socket->network_size_ >= socket->network_.front().size); 922 socket->network_size_ -= socket->network_.front().size; 923 socket->network_.pop_front(); 924 } 925 } 926 927 uint32 VirtualSocketServer::SendDelay(uint32 size) { 928 if (bandwidth_ == 0) 929 return 0; 930 else 931 return 1000 * size / bandwidth_; 932 } 933 934 #if 0 935 void PrintFunction(std::vector<std::pair<double, double> >* f) { 936 return; 937 double sum = 0; 938 for (uint32 i = 0; i < f->size(); ++i) { 939 std::cout << (*f)[i].first << '\t' << (*f)[i].second << std::endl; 940 sum += (*f)[i].second; 941 } 942 if (!f->empty()) { 943 const double mean = sum / f->size(); 944 double sum_sq_dev = 0; 945 for (uint32 i = 0; i < f->size(); ++i) { 946 double dev = (*f)[i].second - mean; 947 sum_sq_dev += dev * dev; 948 } 949 std::cout << "Mean = " << mean << " StdDev = " 950 << sqrt(sum_sq_dev / f->size()) << std::endl; 951 } 952 } 953 #endif // <unused> 954 955 void VirtualSocketServer::UpdateDelayDistribution() { 956 Function* dist = CreateDistribution(delay_mean_, delay_stddev_, 957 delay_samples_); 958 // We take a lock just to make sure we don't leak memory. 959 { 960 CritScope cs(&delay_crit_); 961 delete delay_dist_; 962 delay_dist_ = dist; 963 } 964 } 965 966 static double PI = 4 * std::atan(1.0); 967 968 static double Normal(double x, double mean, double stddev) { 969 double a = (x - mean) * (x - mean) / (2 * stddev * stddev); 970 return std::exp(-a) / (stddev * sqrt(2 * PI)); 971 } 972 973 #if 0 // static unused gives a warning 974 static double Pareto(double x, double min, double k) { 975 if (x < min) 976 return 0; 977 else 978 return k * std::pow(min, k) / std::pow(x, k+1); 979 } 980 #endif 981 982 VirtualSocketServer::Function* VirtualSocketServer::CreateDistribution( 983 uint32 mean, uint32 stddev, uint32 samples) { 984 Function* f = new Function(); 985 986 if (0 == stddev) { 987 f->push_back(Point(mean, 1.0)); 988 } else { 989 double start = 0; 990 if (mean >= 4 * static_cast<double>(stddev)) 991 start = mean - 4 * static_cast<double>(stddev); 992 double end = mean + 4 * static_cast<double>(stddev); 993 994 for (uint32 i = 0; i < samples; i++) { 995 double x = start + (end - start) * i / (samples - 1); 996 double y = Normal(x, mean, stddev); 997 f->push_back(Point(x, y)); 998 } 999 } 1000 return Resample(Invert(Accumulate(f)), 0, 1, samples); 1001 } 1002 1003 uint32 VirtualSocketServer::GetRandomTransitDelay() { 1004 size_t index = rand() % delay_dist_->size(); 1005 double delay = (*delay_dist_)[index].second; 1006 //LOG_F(LS_INFO) << "random[" << index << "] = " << delay; 1007 return static_cast<uint32>(delay); 1008 } 1009 1010 struct FunctionDomainCmp { 1011 bool operator()(const VirtualSocketServer::Point& p1, 1012 const VirtualSocketServer::Point& p2) { 1013 return p1.first < p2.first; 1014 } 1015 bool operator()(double v1, const VirtualSocketServer::Point& p2) { 1016 return v1 < p2.first; 1017 } 1018 bool operator()(const VirtualSocketServer::Point& p1, double v2) { 1019 return p1.first < v2; 1020 } 1021 }; 1022 1023 VirtualSocketServer::Function* VirtualSocketServer::Accumulate(Function* f) { 1024 ASSERT(f->size() >= 1); 1025 double v = 0; 1026 for (Function::size_type i = 0; i < f->size() - 1; ++i) { 1027 double dx = (*f)[i + 1].first - (*f)[i].first; 1028 double avgy = ((*f)[i + 1].second + (*f)[i].second) / 2; 1029 (*f)[i].second = v; 1030 v = v + dx * avgy; 1031 } 1032 (*f)[f->size()-1].second = v; 1033 return f; 1034 } 1035 1036 VirtualSocketServer::Function* VirtualSocketServer::Invert(Function* f) { 1037 for (Function::size_type i = 0; i < f->size(); ++i) 1038 std::swap((*f)[i].first, (*f)[i].second); 1039 1040 std::sort(f->begin(), f->end(), FunctionDomainCmp()); 1041 return f; 1042 } 1043 1044 VirtualSocketServer::Function* VirtualSocketServer::Resample( 1045 Function* f, double x1, double x2, uint32 samples) { 1046 Function* g = new Function(); 1047 1048 for (size_t i = 0; i < samples; i++) { 1049 double x = x1 + (x2 - x1) * i / (samples - 1); 1050 double y = Evaluate(f, x); 1051 g->push_back(Point(x, y)); 1052 } 1053 1054 delete f; 1055 return g; 1056 } 1057 1058 double VirtualSocketServer::Evaluate(Function* f, double x) { 1059 Function::iterator iter = 1060 std::lower_bound(f->begin(), f->end(), x, FunctionDomainCmp()); 1061 if (iter == f->begin()) { 1062 return (*f)[0].second; 1063 } else if (iter == f->end()) { 1064 ASSERT(f->size() >= 1); 1065 return (*f)[f->size() - 1].second; 1066 } else if (iter->first == x) { 1067 return iter->second; 1068 } else { 1069 double x1 = (iter - 1)->first; 1070 double y1 = (iter - 1)->second; 1071 double x2 = iter->first; 1072 double y2 = iter->second; 1073 return y1 + (y2 - y1) * (x - x1) / (x2 - x1); 1074 } 1075 } 1076 1077 bool VirtualSocketServer::CanInteractWith(VirtualSocket* local, 1078 VirtualSocket* remote) { 1079 if (!local || !remote) { 1080 return false; 1081 } 1082 IPAddress local_ip = local->GetLocalAddress().ipaddr(); 1083 IPAddress remote_ip = remote->GetLocalAddress().ipaddr(); 1084 IPAddress local_normalized = local_ip.Normalized(); 1085 IPAddress remote_normalized = remote_ip.Normalized(); 1086 // Check if the addresses are the same family after Normalization (turns 1087 // mapped IPv6 address into IPv4 addresses). 1088 // This will stop unmapped V6 addresses from talking to mapped V6 addresses. 1089 if (local_normalized.family() == remote_normalized.family()) { 1090 return true; 1091 } 1092 1093 // If ip1 is IPv4 and ip2 is :: and ip2 is not IPV6_V6ONLY. 1094 int remote_v6_only = 0; 1095 remote->GetOption(Socket::OPT_IPV6_V6ONLY, &remote_v6_only); 1096 if (local_ip.family() == AF_INET && !remote_v6_only && IPIsAny(remote_ip)) { 1097 return true; 1098 } 1099 // Same check, backwards. 1100 int local_v6_only = 0; 1101 local->GetOption(Socket::OPT_IPV6_V6ONLY, &local_v6_only); 1102 if (remote_ip.family() == AF_INET && !local_v6_only && IPIsAny(local_ip)) { 1103 return true; 1104 } 1105 1106 // Check to see if either socket was explicitly bound to IPv6-any. 1107 // These sockets can talk with anyone. 1108 if (local_ip.family() == AF_INET6 && local->was_any()) { 1109 return true; 1110 } 1111 if (remote_ip.family() == AF_INET6 && remote->was_any()) { 1112 return true; 1113 } 1114 1115 return false; 1116 } 1117 1118 } // namespace talk_base 1119