whip-output.cpp 16 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583
  1. #include "whip-output.h"
  2. #include "whip-utils.h"
  3. /*
  4. * Sets the maximum size for a video fragment. Effective range is
  5. * 576-1470, with a lower value equating to more packets created,
  6. * but also better network compatability.
  7. */
  8. static uint16_t MAX_VIDEO_FRAGMENT_SIZE = 1200;
  9. const int signaling_media_id_length = 16;
  10. const char signaling_media_id_valid_char[] = "0123456789"
  11. "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
  12. "abcdefghijklmnopqrstuvwxyz";
  13. const std::string user_agent = generate_user_agent();
  14. const char *audio_mid = "0";
  15. const uint8_t audio_payload_type = 111;
  16. const char *video_mid = "1";
  17. const uint8_t video_payload_type = 96;
  18. WHIPOutput::WHIPOutput(obs_data_t *, obs_output_t *output)
  19. : output(output),
  20. endpoint_url(),
  21. bearer_token(),
  22. resource_url(),
  23. running(false),
  24. start_stop_mutex(),
  25. start_stop_thread(),
  26. base_ssrc(generate_random_u32()),
  27. peer_connection(nullptr),
  28. audio_track(nullptr),
  29. video_track(nullptr),
  30. total_bytes_sent(0),
  31. connect_time_ms(0),
  32. start_time_ns(0),
  33. last_audio_timestamp(0),
  34. last_video_timestamp(0)
  35. {
  36. }
  37. WHIPOutput::~WHIPOutput()
  38. {
  39. Stop();
  40. std::lock_guard<std::mutex> l(start_stop_mutex);
  41. if (start_stop_thread.joinable())
  42. start_stop_thread.join();
  43. }
  44. bool WHIPOutput::Start()
  45. {
  46. std::lock_guard<std::mutex> l(start_stop_mutex);
  47. if (!obs_output_can_begin_data_capture(output, 0))
  48. return false;
  49. if (!obs_output_initialize_encoders(output, 0))
  50. return false;
  51. if (start_stop_thread.joinable())
  52. start_stop_thread.join();
  53. start_stop_thread = std::thread(&WHIPOutput::StartThread, this);
  54. return true;
  55. }
  56. void WHIPOutput::Stop(bool signal)
  57. {
  58. std::lock_guard<std::mutex> l(start_stop_mutex);
  59. if (start_stop_thread.joinable())
  60. start_stop_thread.join();
  61. start_stop_thread = std::thread(&WHIPOutput::StopThread, this, signal);
  62. }
  63. void WHIPOutput::Data(struct encoder_packet *packet)
  64. {
  65. if (!packet) {
  66. Stop(false);
  67. obs_output_signal_stop(output, OBS_OUTPUT_ENCODE_ERROR);
  68. return;
  69. }
  70. if (packet->type == OBS_ENCODER_AUDIO) {
  71. int64_t duration = packet->dts_usec - last_audio_timestamp;
  72. Send(packet->data, packet->size, duration, audio_track,
  73. audio_sr_reporter);
  74. last_audio_timestamp = packet->dts_usec;
  75. } else if (packet->type == OBS_ENCODER_VIDEO) {
  76. int64_t duration = packet->dts_usec - last_video_timestamp;
  77. Send(packet->data, packet->size, duration, video_track,
  78. video_sr_reporter);
  79. last_video_timestamp = packet->dts_usec;
  80. }
  81. }
  82. void WHIPOutput::ConfigureAudioTrack(std::string media_stream_id,
  83. std::string cname)
  84. {
  85. auto media_stream_track_id = std::string(media_stream_id + "-audio");
  86. uint32_t ssrc = base_ssrc;
  87. rtc::Description::Audio audio_description(
  88. audio_mid, rtc::Description::Direction::SendOnly);
  89. audio_description.addOpusCodec(audio_payload_type);
  90. audio_description.addSSRC(ssrc, cname, media_stream_id,
  91. media_stream_track_id);
  92. audio_track = peer_connection->addTrack(audio_description);
  93. auto rtp_config = std::make_shared<rtc::RtpPacketizationConfig>(
  94. ssrc, cname, audio_payload_type,
  95. rtc::OpusRtpPacketizer::DefaultClockRate);
  96. auto packetizer = std::make_shared<rtc::OpusRtpPacketizer>(rtp_config);
  97. audio_sr_reporter = std::make_shared<rtc::RtcpSrReporter>(rtp_config);
  98. auto nack_responder = std::make_shared<rtc::RtcpNackResponder>();
  99. packetizer->addToChain(audio_sr_reporter);
  100. packetizer->addToChain(nack_responder);
  101. audio_track->setMediaHandler(packetizer);
  102. }
  103. void WHIPOutput::ConfigureVideoTrack(std::string media_stream_id,
  104. std::string cname)
  105. {
  106. auto media_stream_track_id = std::string(media_stream_id + "-video");
  107. // More predictable SSRC values between audio and video
  108. uint32_t ssrc = base_ssrc + 1;
  109. rtc::Description::Video video_description(
  110. video_mid, rtc::Description::Direction::SendOnly);
  111. video_description.addH264Codec(video_payload_type);
  112. video_description.addSSRC(ssrc, cname, media_stream_id,
  113. media_stream_track_id);
  114. video_track = peer_connection->addTrack(video_description);
  115. auto rtp_config = std::make_shared<rtc::RtpPacketizationConfig>(
  116. ssrc, cname, video_payload_type,
  117. rtc::H264RtpPacketizer::defaultClockRate);
  118. auto packetizer = std::make_shared<rtc::H264RtpPacketizer>(
  119. rtc::H264RtpPacketizer::Separator::StartSequence, rtp_config,
  120. MAX_VIDEO_FRAGMENT_SIZE);
  121. video_sr_reporter = std::make_shared<rtc::RtcpSrReporter>(rtp_config);
  122. auto nack_responder = std::make_shared<rtc::RtcpNackResponder>();
  123. packetizer->addToChain(video_sr_reporter);
  124. packetizer->addToChain(nack_responder);
  125. video_track->setMediaHandler(packetizer);
  126. }
  127. /**
  128. * @brief Initialize encoders and store connect info provided by the service.
  129. *
  130. * @return bool
  131. */
  132. bool WHIPOutput::Init()
  133. {
  134. if (!obs_output_can_begin_data_capture(output, 0))
  135. return false;
  136. if (!obs_output_initialize_encoders(output, 0))
  137. return false;
  138. obs_service_t *service = obs_output_get_service(output);
  139. if (!service) {
  140. obs_output_signal_stop(output, OBS_OUTPUT_ERROR);
  141. return false;
  142. }
  143. endpoint_url = obs_service_get_connect_info(
  144. service, OBS_SERVICE_CONNECT_INFO_SERVER_URL);
  145. if (endpoint_url.empty()) {
  146. obs_output_signal_stop(output, OBS_OUTPUT_BAD_PATH);
  147. return false;
  148. }
  149. bearer_token = obs_service_get_connect_info(
  150. service, OBS_SERVICE_CONNECT_INFO_BEARER_TOKEN);
  151. return true;
  152. }
  153. /**
  154. * @brief Set up the PeerConnection and media tracks.
  155. *
  156. * @return bool
  157. */
  158. bool WHIPOutput::Setup()
  159. {
  160. peer_connection = std::make_shared<rtc::PeerConnection>();
  161. peer_connection->onStateChange([this](rtc::PeerConnection::State state) {
  162. switch (state) {
  163. case rtc::PeerConnection::State::New:
  164. do_log(LOG_INFO, "PeerConnection state is now: New");
  165. break;
  166. case rtc::PeerConnection::State::Connecting:
  167. do_log(LOG_INFO,
  168. "PeerConnection state is now: Connecting");
  169. start_time_ns = os_gettime_ns();
  170. break;
  171. case rtc::PeerConnection::State::Connected:
  172. do_log(LOG_INFO,
  173. "PeerConnection state is now: Connected");
  174. connect_time_ms =
  175. (int)((os_gettime_ns() - start_time_ns) /
  176. 1000000.0);
  177. do_log(LOG_INFO, "Connect time: %dms",
  178. connect_time_ms.load());
  179. break;
  180. case rtc::PeerConnection::State::Disconnected:
  181. do_log(LOG_INFO,
  182. "PeerConnection state is now: Disconnected");
  183. Stop(false);
  184. obs_output_signal_stop(output, OBS_OUTPUT_DISCONNECTED);
  185. break;
  186. case rtc::PeerConnection::State::Failed:
  187. do_log(LOG_INFO, "PeerConnection state is now: Failed");
  188. Stop(false);
  189. obs_output_signal_stop(output, OBS_OUTPUT_ERROR);
  190. break;
  191. case rtc::PeerConnection::State::Closed:
  192. do_log(LOG_INFO, "PeerConnection state is now: Closed");
  193. break;
  194. }
  195. });
  196. std::string media_stream_id, cname;
  197. media_stream_id.reserve(signaling_media_id_length);
  198. cname.reserve(signaling_media_id_length);
  199. for (int i = 0; i < signaling_media_id_length; ++i) {
  200. media_stream_id += signaling_media_id_valid_char
  201. [rand() % (sizeof(signaling_media_id_valid_char) - 1)];
  202. cname += signaling_media_id_valid_char
  203. [rand() % (sizeof(signaling_media_id_valid_char) - 1)];
  204. }
  205. ConfigureAudioTrack(media_stream_id, cname);
  206. ConfigureVideoTrack(media_stream_id, cname);
  207. peer_connection->setLocalDescription();
  208. return true;
  209. }
  210. bool WHIPOutput::Connect()
  211. {
  212. struct curl_slist *headers = NULL;
  213. headers = curl_slist_append(headers, "Content-Type: application/sdp");
  214. if (!bearer_token.empty()) {
  215. auto bearer_token_header =
  216. std::string("Authorization: Bearer ") + bearer_token;
  217. headers =
  218. curl_slist_append(headers, bearer_token_header.c_str());
  219. }
  220. std::string read_buffer;
  221. std::vector<std::string> location_headers;
  222. auto offer_sdp =
  223. std::string(peer_connection->localDescription().value());
  224. #ifdef DEBUG_SDP
  225. do_log(LOG_DEBUG, "Offer SDP:\n%s", offer_sdp.c_str());
  226. #endif
  227. // Add user-agent to our requests
  228. headers = curl_slist_append(headers, user_agent.c_str());
  229. char error_buffer[CURL_ERROR_SIZE] = {};
  230. CURL *c = curl_easy_init();
  231. curl_easy_setopt(c, CURLOPT_WRITEFUNCTION, curl_writefunction);
  232. curl_easy_setopt(c, CURLOPT_WRITEDATA, (void *)&read_buffer);
  233. curl_easy_setopt(c, CURLOPT_HEADERFUNCTION,
  234. curl_header_location_function);
  235. curl_easy_setopt(c, CURLOPT_HEADERDATA, (void *)&location_headers);
  236. curl_easy_setopt(c, CURLOPT_HTTPHEADER, headers);
  237. curl_easy_setopt(c, CURLOPT_URL, endpoint_url.c_str());
  238. curl_easy_setopt(c, CURLOPT_POST, 1L);
  239. curl_easy_setopt(c, CURLOPT_COPYPOSTFIELDS, offer_sdp.c_str());
  240. curl_easy_setopt(c, CURLOPT_TIMEOUT, 8L);
  241. curl_easy_setopt(c, CURLOPT_FOLLOWLOCATION, 1L);
  242. curl_easy_setopt(c, CURLOPT_UNRESTRICTED_AUTH, 1L);
  243. curl_easy_setopt(c, CURLOPT_ERRORBUFFER, error_buffer);
  244. auto cleanup = [&]() {
  245. curl_easy_cleanup(c);
  246. curl_slist_free_all(headers);
  247. };
  248. CURLcode res = curl_easy_perform(c);
  249. if (res != CURLE_OK) {
  250. do_log(LOG_ERROR, "Connect failed: %s",
  251. error_buffer[0] ? error_buffer
  252. : curl_easy_strerror(res));
  253. cleanup();
  254. obs_output_signal_stop(output, OBS_OUTPUT_CONNECT_FAILED);
  255. return false;
  256. }
  257. long response_code;
  258. curl_easy_getinfo(c, CURLINFO_RESPONSE_CODE, &response_code);
  259. if (response_code != 201) {
  260. do_log(LOG_ERROR,
  261. "Connect failed: HTTP endpoint returned response code %ld",
  262. response_code);
  263. cleanup();
  264. obs_output_signal_stop(output, OBS_OUTPUT_INVALID_STREAM);
  265. return false;
  266. }
  267. if (read_buffer.empty()) {
  268. do_log(LOG_ERROR,
  269. "Connect failed: No data returned from HTTP endpoint request");
  270. cleanup();
  271. obs_output_signal_stop(output, OBS_OUTPUT_CONNECT_FAILED);
  272. return false;
  273. }
  274. long redirect_count = 0;
  275. curl_easy_getinfo(c, CURLINFO_REDIRECT_COUNT, &redirect_count);
  276. if (location_headers.size() < static_cast<size_t>(redirect_count) + 1) {
  277. do_log(LOG_ERROR,
  278. "WHIP server did not provide a resource URL via the Location header");
  279. cleanup();
  280. obs_output_signal_stop(output, OBS_OUTPUT_CONNECT_FAILED);
  281. return false;
  282. }
  283. CURLU *url_builder = curl_url();
  284. auto last_location_header = location_headers.back();
  285. // If Location header doesn't start with `http` it is a relative URL.
  286. // Construct a absolute URL using the host of the effective URL
  287. if (last_location_header.find("http") != 0) {
  288. char *effective_url = nullptr;
  289. curl_easy_getinfo(c, CURLINFO_EFFECTIVE_URL, &effective_url);
  290. if (effective_url == nullptr) {
  291. do_log(LOG_ERROR, "Failed to build Resource URL");
  292. cleanup();
  293. obs_output_signal_stop(output,
  294. OBS_OUTPUT_CONNECT_FAILED);
  295. return false;
  296. }
  297. curl_url_set(url_builder, CURLUPART_URL, effective_url, 0);
  298. curl_url_set(url_builder, CURLUPART_PATH,
  299. last_location_header.c_str(), 0);
  300. curl_url_set(url_builder, CURLUPART_QUERY, "", 0);
  301. } else {
  302. curl_url_set(url_builder, CURLUPART_URL,
  303. last_location_header.c_str(), 0);
  304. }
  305. char *url = nullptr;
  306. CURLUcode rc = curl_url_get(url_builder, CURLUPART_URL, &url,
  307. CURLU_NO_DEFAULT_PORT);
  308. if (rc) {
  309. do_log(LOG_ERROR,
  310. "WHIP server provided a invalid resource URL via the Location header");
  311. cleanup();
  312. obs_output_signal_stop(output, OBS_OUTPUT_CONNECT_FAILED);
  313. return false;
  314. }
  315. resource_url = url;
  316. curl_free(url);
  317. do_log(LOG_DEBUG, "WHIP Resource URL is: %s", resource_url.c_str());
  318. curl_url_cleanup(url_builder);
  319. #ifdef DEBUG_SDP
  320. do_log(LOG_DEBUG, "Answer SDP:\n%s", read_buffer.c_str());
  321. #endif
  322. auto response = std::string(read_buffer);
  323. response.erase(0, response.find("v=0"));
  324. rtc::Description answer(response, "answer");
  325. peer_connection->setRemoteDescription(answer);
  326. cleanup();
  327. return true;
  328. }
  329. void WHIPOutput::StartThread()
  330. {
  331. if (!Init())
  332. return;
  333. if (!Setup())
  334. return;
  335. if (!Connect()) {
  336. peer_connection->close();
  337. peer_connection = nullptr;
  338. audio_track = nullptr;
  339. video_track = nullptr;
  340. return;
  341. }
  342. obs_output_begin_data_capture(output, 0);
  343. running = true;
  344. }
  345. void WHIPOutput::SendDelete()
  346. {
  347. if (resource_url.empty()) {
  348. do_log(LOG_DEBUG,
  349. "No resource URL available, not sending DELETE");
  350. return;
  351. }
  352. struct curl_slist *headers = NULL;
  353. if (!bearer_token.empty()) {
  354. auto bearer_token_header =
  355. std::string("Authorization: Bearer ") + bearer_token;
  356. headers =
  357. curl_slist_append(headers, bearer_token_header.c_str());
  358. }
  359. // Add user-agent to our requests
  360. headers = curl_slist_append(headers, user_agent.c_str());
  361. char error_buffer[CURL_ERROR_SIZE] = {};
  362. CURL *c = curl_easy_init();
  363. curl_easy_setopt(c, CURLOPT_HTTPHEADER, headers);
  364. curl_easy_setopt(c, CURLOPT_URL, resource_url.c_str());
  365. curl_easy_setopt(c, CURLOPT_CUSTOMREQUEST, "DELETE");
  366. curl_easy_setopt(c, CURLOPT_TIMEOUT, 8L);
  367. curl_easy_setopt(c, CURLOPT_ERRORBUFFER, error_buffer);
  368. auto cleanup = [&]() {
  369. curl_easy_cleanup(c);
  370. curl_slist_free_all(headers);
  371. };
  372. CURLcode res = curl_easy_perform(c);
  373. if (res != CURLE_OK) {
  374. do_log(LOG_WARNING,
  375. "DELETE request for resource URL failed: %s",
  376. error_buffer[0] ? error_buffer
  377. : curl_easy_strerror(res));
  378. cleanup();
  379. return;
  380. }
  381. long response_code;
  382. curl_easy_getinfo(c, CURLINFO_RESPONSE_CODE, &response_code);
  383. if (response_code != 200) {
  384. do_log(LOG_WARNING,
  385. "DELETE request for resource URL failed. HTTP Code: %ld",
  386. response_code);
  387. cleanup();
  388. return;
  389. }
  390. do_log(LOG_DEBUG,
  391. "Successfully performed DELETE request for resource URL");
  392. resource_url.clear();
  393. cleanup();
  394. }
  395. void WHIPOutput::StopThread(bool signal)
  396. {
  397. if (peer_connection != nullptr) {
  398. peer_connection->close();
  399. peer_connection = nullptr;
  400. audio_track = nullptr;
  401. video_track = nullptr;
  402. }
  403. SendDelete();
  404. /*
  405. * "signal" exists because we have to preserve the "running" state
  406. * across reconnect attempts. If we don't emit a signal if
  407. * something calls obs_output_stop() and it's reconnecting, you'll
  408. * desync the UI, as the output will be "stopped" and not
  409. * "reconnecting", but the "stop" signal will have never been
  410. * emitted.
  411. */
  412. if (running && signal) {
  413. obs_output_signal_stop(output, OBS_OUTPUT_SUCCESS);
  414. running = false;
  415. }
  416. total_bytes_sent = 0;
  417. connect_time_ms = 0;
  418. start_time_ns = 0;
  419. last_audio_timestamp = 0;
  420. last_video_timestamp = 0;
  421. }
  422. void WHIPOutput::Send(void *data, uintptr_t size, uint64_t duration,
  423. std::shared_ptr<rtc::Track> track,
  424. std::shared_ptr<rtc::RtcpSrReporter> rtcp_sr_reporter)
  425. {
  426. if (track == nullptr || !track->isOpen())
  427. return;
  428. std::vector<rtc::byte> sample{(rtc::byte *)data,
  429. (rtc::byte *)data + size};
  430. auto rtp_config = rtcp_sr_reporter->rtpConfig;
  431. // Sample time is in microseconds, we need to convert it to seconds
  432. auto elapsed_seconds = double(duration) / (1000.0 * 1000.0);
  433. // Get elapsed time in clock rate
  434. uint32_t elapsed_timestamp =
  435. rtp_config->secondsToTimestamp(elapsed_seconds);
  436. // Set new timestamp
  437. rtp_config->timestamp = rtp_config->timestamp + elapsed_timestamp;
  438. // get elapsed time in clock rate from last RTCP sender report
  439. auto report_elapsed_timestamp =
  440. rtp_config->timestamp -
  441. rtcp_sr_reporter->lastReportedTimestamp();
  442. // check if last report was at least 1 second ago
  443. if (rtp_config->timestampToSeconds(report_elapsed_timestamp) > 1)
  444. rtcp_sr_reporter->setNeedsToReport();
  445. try {
  446. track->send(sample);
  447. total_bytes_sent += sample.size();
  448. } catch (const std::exception &e) {
  449. do_log(LOG_ERROR, "error: %s ", e.what());
  450. }
  451. }
  452. void register_whip_output()
  453. {
  454. struct obs_output_info info = {};
  455. info.id = "whip_output";
  456. info.flags = OBS_OUTPUT_AV | OBS_OUTPUT_ENCODED | OBS_OUTPUT_SERVICE;
  457. info.get_name = [](void *) -> const char * {
  458. return obs_module_text("Output.Name");
  459. };
  460. info.create = [](obs_data_t *settings, obs_output_t *output) -> void * {
  461. return new WHIPOutput(settings, output);
  462. };
  463. info.destroy = [](void *priv_data) {
  464. delete static_cast<WHIPOutput *>(priv_data);
  465. };
  466. info.start = [](void *priv_data) -> bool {
  467. return static_cast<WHIPOutput *>(priv_data)->Start();
  468. };
  469. info.stop = [](void *priv_data, uint64_t) {
  470. static_cast<WHIPOutput *>(priv_data)->Stop();
  471. };
  472. info.encoded_packet = [](void *priv_data,
  473. struct encoder_packet *packet) {
  474. static_cast<WHIPOutput *>(priv_data)->Data(packet);
  475. };
  476. info.get_defaults = [](obs_data_t *) {
  477. };
  478. info.get_properties = [](void *) -> obs_properties_t * {
  479. return obs_properties_create();
  480. };
  481. info.get_total_bytes = [](void *priv_data) -> uint64_t {
  482. return (uint64_t) static_cast<WHIPOutput *>(priv_data)
  483. ->GetTotalBytes();
  484. };
  485. info.get_connect_time_ms = [](void *priv_data) -> int {
  486. return static_cast<WHIPOutput *>(priv_data)->GetConnectTime();
  487. };
  488. info.encoded_video_codecs = "h264";
  489. info.encoded_audio_codecs = "opus";
  490. info.protocols = "WHIP";
  491. obs_register_output(&info);
  492. }