blob: 6589ebb5a20fd9b406c934593d79e122f473d3ec [file] [log] [blame]
henrike@webrtc.org0e118e72013-07-10 00:45:36 +00001/*
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>
pbos@webrtc.orgb9518272014-03-07 15:22:04 +000031#include <math.h>
henrike@webrtc.org0e118e72013-07-10 00:45:36 +000032
33#include <algorithm>
henrike@webrtc.org0e118e72013-07-10 00:45:36 +000034#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
44namespace talk_base {
45#ifdef WIN32
46const in_addr kInitialNextIPv4 = { {0x01, 0, 0, 0} };
47#else
48// This value is entirely arbitrary, hence the lack of concern about endianness.
49const in_addr kInitialNextIPv4 = { 0x01000000 };
50#endif
51// Starts at ::2 so as to not cause confusion with ::1.
52const in6_addr kInitialNextIPv6 = { { {
53 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2
54 } } };
55
56const uint16 kFirstEphemeralPort = 49152;
57const uint16 kLastEphemeralPort = 65535;
58const uint16 kEphemeralPortCount = kLastEphemeralPort - kFirstEphemeralPort + 1;
59const uint32 kDefaultNetworkCapacity = 64 * 1024;
60const uint32 kDefaultTcpBufferSize = 32 * 1024;
61
62const uint32 UDP_HEADER_SIZE = 28; // IP + UDP headers
63const uint32 TCP_HEADER_SIZE = 40; // IP + TCP headers
64const uint32 TCP_MSS = 1400; // Maximum segment size
65
66// Note: The current algorithm doesn't work for sample sizes smaller than this.
67const int NUM_SAMPLES = 1000;
68
69enum {
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.
77class 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_];
pbos@webrtc.orgb9518272014-03-07 15:22:04 +000083 memcpy(data_, data, size_);
henrike@webrtc.org0e118e72013-07-10 00:45:36 +000084 }
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
106struct 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.
113class 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),
henrike@webrtc.org7162d282013-07-26 19:17:59 +0000117 state_(CS_CLOSED), error_(0), listen_queue_(NULL),
118 write_enabled_(false),
henrike@webrtc.org0e118e72013-07-10 00:45:36 +0000119 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());
pbos@webrtc.orgb9518272014-03-07 15:22:04 +0000286 memcpy(pv, packet->data(), data_read);
henrike@webrtc.org0e118e72013-07-10 00:45:36 +0000287 *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
555VirtualSocketServer::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
572VirtualSocketServer::~VirtualSocketServer() {
573 delete bindings_;
574 delete connections_;
575 delete delay_dist_;
576 if (server_owned_) {
577 delete server_;
578 }
579}
580
581IPAddress 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
596uint16 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
606Socket* VirtualSocketServer::CreateSocket(int type) {
607 return CreateSocket(AF_INET, type);
608}
609
610Socket* VirtualSocketServer::CreateSocket(int family, int type) {
611 return CreateSocketInternal(family, type);
612}
613
614AsyncSocket* VirtualSocketServer::CreateAsyncSocket(int type) {
615 return CreateAsyncSocket(AF_INET, type);
616}
617
618AsyncSocket* VirtualSocketServer::CreateAsyncSocket(int family, int type) {
619 return CreateSocketInternal(family, type);
620}
621
622VirtualSocket* VirtualSocketServer::CreateSocketInternal(int family, int type) {
623 return new VirtualSocket(this, family, type, true);
624}
625
626void 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
634bool 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
642void VirtualSocketServer::WakeUp() {
643 socketserver()->WakeUp();
644}
645
646bool 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
659int 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
673int 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
696VirtualSocket* 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
703int 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
712void 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
726VirtualSocket* 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
738void 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
748static double Random() {
749 return static_cast<double>(rand()) / RAND_MAX;
750}
751
752int 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
773bool 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
782int 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
837void 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
886void 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
917void 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
927uint32 VirtualSocketServer::SendDelay(uint32 size) {
928 if (bandwidth_ == 0)
929 return 0;
930 else
931 return 1000 * size / bandwidth_;
932}
933
934#if 0
935void 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
955void 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
pbos@webrtc.orgb9518272014-03-07 15:22:04 +0000966static double PI = 4 * atan(1.0);
henrike@webrtc.org0e118e72013-07-10 00:45:36 +0000967
968static double Normal(double x, double mean, double stddev) {
969 double a = (x - mean) * (x - mean) / (2 * stddev * stddev);
pbos@webrtc.orgb9518272014-03-07 15:22:04 +0000970 return exp(-a) / (stddev * sqrt(2 * PI));
henrike@webrtc.org0e118e72013-07-10 00:45:36 +0000971}
972
973#if 0 // static unused gives a warning
974static 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
982VirtualSocketServer::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
1003uint32 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
1010struct 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
1023VirtualSocketServer::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
1036VirtualSocketServer::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
1044VirtualSocketServer::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
1058double 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
1077bool 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