whip-output.cpp 15 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 uint32_t audio_clockrate = 48000;
  16. const uint8_t audio_payload_type = 111;
  17. const char *video_mid = "1";
  18. const uint32_t video_clockrate = 90000;
  19. const uint8_t video_payload_type = 96;
  20. WHIPOutput::WHIPOutput(obs_data_t *, obs_output_t *output)
  21. : output(output),
  22. endpoint_url(),
  23. bearer_token(),
  24. resource_url(),
  25. running(false),
  26. start_stop_mutex(),
  27. start_stop_thread(),
  28. base_ssrc(generate_random_u32()),
  29. peer_connection(-1),
  30. audio_track(-1),
  31. video_track(-1),
  32. total_bytes_sent(0),
  33. connect_time_ms(0),
  34. start_time_ns(0),
  35. last_audio_timestamp(0),
  36. last_video_timestamp(0)
  37. {
  38. }
  39. WHIPOutput::~WHIPOutput()
  40. {
  41. Stop();
  42. std::lock_guard<std::mutex> l(start_stop_mutex);
  43. if (start_stop_thread.joinable())
  44. start_stop_thread.join();
  45. }
  46. bool WHIPOutput::Start()
  47. {
  48. std::lock_guard<std::mutex> l(start_stop_mutex);
  49. if (!obs_output_can_begin_data_capture(output, 0))
  50. return false;
  51. if (!obs_output_initialize_encoders(output, 0))
  52. return false;
  53. if (start_stop_thread.joinable())
  54. start_stop_thread.join();
  55. start_stop_thread = std::thread(&WHIPOutput::StartThread, this);
  56. return true;
  57. }
  58. void WHIPOutput::Stop(bool signal)
  59. {
  60. std::lock_guard<std::mutex> l(start_stop_mutex);
  61. if (start_stop_thread.joinable())
  62. start_stop_thread.join();
  63. start_stop_thread = std::thread(&WHIPOutput::StopThread, this, signal);
  64. }
  65. void WHIPOutput::Data(struct encoder_packet *packet)
  66. {
  67. if (!packet) {
  68. Stop(false);
  69. obs_output_signal_stop(output, OBS_OUTPUT_ENCODE_ERROR);
  70. return;
  71. }
  72. if (packet->type == OBS_ENCODER_AUDIO) {
  73. int64_t duration = packet->dts_usec - last_audio_timestamp;
  74. Send(packet->data, packet->size, duration, audio_track);
  75. last_audio_timestamp = packet->dts_usec;
  76. } else if (packet->type == OBS_ENCODER_VIDEO) {
  77. int64_t duration = packet->dts_usec - last_video_timestamp;
  78. Send(packet->data, packet->size, duration, video_track);
  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. rtcTrackInit track_init = {
  88. RTC_DIRECTION_SENDONLY,
  89. RTC_CODEC_OPUS,
  90. audio_payload_type,
  91. ssrc,
  92. audio_mid,
  93. cname.c_str(),
  94. media_stream_id.c_str(),
  95. media_stream_track_id.c_str(),
  96. };
  97. // Generate a random starting timestamp for the audio track
  98. uint32_t rtp_audio_timestamp = generate_random_u32();
  99. rtcPacketizationHandlerInit packetizer_init = {ssrc,
  100. cname.c_str(),
  101. audio_payload_type,
  102. audio_clockrate,
  103. 0,
  104. rtp_audio_timestamp,
  105. RTC_NAL_SEPARATOR_LENGTH,
  106. 0};
  107. audio_track = rtcAddTrackEx(peer_connection, &track_init);
  108. rtcSetOpusPacketizationHandler(audio_track, &packetizer_init);
  109. rtcChainRtcpSrReporter(audio_track);
  110. rtcChainRtcpNackResponder(audio_track, 1000);
  111. }
  112. void WHIPOutput::ConfigureVideoTrack(std::string media_stream_id,
  113. std::string cname)
  114. {
  115. auto media_stream_track_id = std::string(media_stream_id + "-video");
  116. // More predictable SSRC values between audio and video
  117. uint32_t ssrc = base_ssrc + 1;
  118. rtcTrackInit track_init = {
  119. RTC_DIRECTION_SENDONLY,
  120. RTC_CODEC_H264,
  121. video_payload_type,
  122. ssrc,
  123. video_mid,
  124. cname.c_str(),
  125. media_stream_id.c_str(),
  126. media_stream_track_id.c_str(),
  127. };
  128. // Generate a random starting timestamp for the video track
  129. uint32_t rtp_video_timestamp = generate_random_u32();
  130. rtcPacketizationHandlerInit packetizer_init = {
  131. ssrc,
  132. cname.c_str(),
  133. video_payload_type,
  134. video_clockrate,
  135. 0,
  136. rtp_video_timestamp,
  137. RTC_NAL_SEPARATOR_START_SEQUENCE,
  138. MAX_VIDEO_FRAGMENT_SIZE};
  139. video_track = rtcAddTrackEx(peer_connection, &track_init);
  140. rtcSetH264PacketizationHandler(video_track, &packetizer_init);
  141. rtcChainRtcpSrReporter(video_track);
  142. rtcChainRtcpNackResponder(video_track, 1000);
  143. }
  144. /**
  145. * @brief Initialize encoders and store connect info provided by the service.
  146. *
  147. * @return bool
  148. */
  149. bool WHIPOutput::Init()
  150. {
  151. if (!obs_output_can_begin_data_capture(output, 0))
  152. return false;
  153. if (!obs_output_initialize_encoders(output, 0))
  154. return false;
  155. obs_service_t *service = obs_output_get_service(output);
  156. if (!service) {
  157. obs_output_signal_stop(output, OBS_OUTPUT_ERROR);
  158. return false;
  159. }
  160. endpoint_url = obs_service_get_connect_info(
  161. service, OBS_SERVICE_CONNECT_INFO_SERVER_URL);
  162. if (endpoint_url.empty()) {
  163. obs_output_signal_stop(output, OBS_OUTPUT_BAD_PATH);
  164. return false;
  165. }
  166. bearer_token = obs_service_get_connect_info(
  167. service, OBS_SERVICE_CONNECT_INFO_BEARER_TOKEN);
  168. return true;
  169. }
  170. /**
  171. * @brief Set up the PeerConnection and media tracks.
  172. *
  173. * @return bool
  174. */
  175. bool WHIPOutput::Setup()
  176. {
  177. rtcConfiguration config;
  178. memset(&config, 0, sizeof(config));
  179. peer_connection = rtcCreatePeerConnection(&config);
  180. rtcSetUserPointer(peer_connection, this);
  181. rtcSetStateChangeCallback(peer_connection, [](int, rtcState state,
  182. void *ptr) {
  183. auto whipOutput = static_cast<WHIPOutput *>(ptr);
  184. switch (state) {
  185. case RTC_NEW:
  186. do_log_s(LOG_INFO, "PeerConnection state is now: New");
  187. break;
  188. case RTC_CONNECTING:
  189. do_log_s(LOG_INFO,
  190. "PeerConnection state is now: Connecting");
  191. whipOutput->start_time_ns = os_gettime_ns();
  192. break;
  193. case RTC_CONNECTED:
  194. do_log_s(LOG_INFO,
  195. "PeerConnection state is now: Connected");
  196. whipOutput->connect_time_ms =
  197. (int)((os_gettime_ns() -
  198. whipOutput->start_time_ns) /
  199. 1000000.0);
  200. do_log_s(LOG_INFO, "Connect time: %dms",
  201. whipOutput->connect_time_ms.load());
  202. break;
  203. case RTC_DISCONNECTED:
  204. do_log_s(LOG_INFO,
  205. "PeerConnection state is now: Disconnected");
  206. whipOutput->Stop(false);
  207. obs_output_signal_stop(whipOutput->output,
  208. OBS_OUTPUT_DISCONNECTED);
  209. break;
  210. case RTC_FAILED:
  211. do_log_s(LOG_INFO,
  212. "PeerConnection state is now: Failed");
  213. whipOutput->Stop(false);
  214. obs_output_signal_stop(whipOutput->output,
  215. OBS_OUTPUT_ERROR);
  216. break;
  217. case RTC_CLOSED:
  218. do_log_s(LOG_INFO,
  219. "PeerConnection state is now: Closed");
  220. break;
  221. }
  222. });
  223. std::string media_stream_id, cname;
  224. media_stream_id.reserve(signaling_media_id_length);
  225. cname.reserve(signaling_media_id_length);
  226. for (int i = 0; i < signaling_media_id_length; ++i) {
  227. media_stream_id += signaling_media_id_valid_char
  228. [rand() % (sizeof(signaling_media_id_valid_char) - 1)];
  229. cname += signaling_media_id_valid_char
  230. [rand() % (sizeof(signaling_media_id_valid_char) - 1)];
  231. }
  232. ConfigureAudioTrack(media_stream_id, cname);
  233. ConfigureVideoTrack(media_stream_id, cname);
  234. rtcSetLocalDescription(peer_connection, "offer");
  235. return true;
  236. }
  237. bool WHIPOutput::Connect()
  238. {
  239. struct curl_slist *headers = NULL;
  240. headers = curl_slist_append(headers, "Content-Type: application/sdp");
  241. if (!bearer_token.empty()) {
  242. auto bearer_token_header =
  243. std::string("Authorization: Bearer ") + bearer_token;
  244. headers =
  245. curl_slist_append(headers, bearer_token_header.c_str());
  246. }
  247. std::string read_buffer;
  248. std::vector<std::string> location_headers;
  249. char offer_sdp[4096] = {0};
  250. rtcGetLocalDescription(peer_connection, offer_sdp, sizeof(offer_sdp));
  251. #ifdef DEBUG_SDP
  252. do_log(LOG_DEBUG, "Offer SDP:\n%s", offer_sdp);
  253. #endif
  254. // Add user-agent to our requests
  255. headers = curl_slist_append(headers, user_agent.c_str());
  256. CURL *c = curl_easy_init();
  257. curl_easy_setopt(c, CURLOPT_WRITEFUNCTION, curl_writefunction);
  258. curl_easy_setopt(c, CURLOPT_WRITEDATA, (void *)&read_buffer);
  259. curl_easy_setopt(c, CURLOPT_HEADERFUNCTION,
  260. curl_header_location_function);
  261. curl_easy_setopt(c, CURLOPT_HEADERDATA, (void *)&location_headers);
  262. curl_easy_setopt(c, CURLOPT_HTTPHEADER, headers);
  263. curl_easy_setopt(c, CURLOPT_URL, endpoint_url.c_str());
  264. curl_easy_setopt(c, CURLOPT_POST, 1L);
  265. curl_easy_setopt(c, CURLOPT_COPYPOSTFIELDS, offer_sdp);
  266. curl_easy_setopt(c, CURLOPT_TIMEOUT, 8L);
  267. curl_easy_setopt(c, CURLOPT_FOLLOWLOCATION, 1L);
  268. curl_easy_setopt(c, CURLOPT_UNRESTRICTED_AUTH, 1L);
  269. auto cleanup = [&]() {
  270. curl_easy_cleanup(c);
  271. curl_slist_free_all(headers);
  272. };
  273. CURLcode res = curl_easy_perform(c);
  274. if (res != CURLE_OK) {
  275. do_log(LOG_ERROR,
  276. "Connect failed: CURL returned result not CURLE_OK");
  277. cleanup();
  278. obs_output_signal_stop(output, OBS_OUTPUT_CONNECT_FAILED);
  279. return false;
  280. }
  281. long response_code;
  282. curl_easy_getinfo(c, CURLINFO_RESPONSE_CODE, &response_code);
  283. if (response_code != 201) {
  284. do_log(LOG_ERROR,
  285. "Connect failed: HTTP endpoint returned response code %ld",
  286. response_code);
  287. cleanup();
  288. obs_output_signal_stop(output, OBS_OUTPUT_INVALID_STREAM);
  289. return false;
  290. }
  291. if (read_buffer.empty()) {
  292. do_log(LOG_ERROR,
  293. "Connect failed: No data returned from HTTP endpoint request");
  294. cleanup();
  295. obs_output_signal_stop(output, OBS_OUTPUT_CONNECT_FAILED);
  296. return false;
  297. }
  298. long redirect_count = 0;
  299. curl_easy_getinfo(c, CURLINFO_REDIRECT_COUNT, &redirect_count);
  300. if (location_headers.size() < static_cast<size_t>(redirect_count) + 1) {
  301. do_log(LOG_ERROR,
  302. "WHIP server did not provide a resource URL via the Location header");
  303. cleanup();
  304. obs_output_signal_stop(output, OBS_OUTPUT_CONNECT_FAILED);
  305. return false;
  306. }
  307. CURLU *url_builder = curl_url();
  308. auto last_location_header = location_headers.back();
  309. // If Location header doesn't start with `http` it is a relative URL.
  310. // Construct a absolute URL using the host of the effective URL
  311. if (last_location_header.find("http") != 0) {
  312. char *effective_url = nullptr;
  313. curl_easy_getinfo(c, CURLINFO_EFFECTIVE_URL, &effective_url);
  314. if (effective_url == nullptr) {
  315. do_log(LOG_ERROR, "Failed to build Resource URL");
  316. cleanup();
  317. obs_output_signal_stop(output,
  318. OBS_OUTPUT_CONNECT_FAILED);
  319. return false;
  320. }
  321. curl_url_set(url_builder, CURLUPART_URL, effective_url, 0);
  322. curl_url_set(url_builder, CURLUPART_PATH,
  323. last_location_header.c_str(), 0);
  324. curl_url_set(url_builder, CURLUPART_QUERY, "", 0);
  325. } else {
  326. curl_url_set(url_builder, CURLUPART_URL,
  327. last_location_header.c_str(), 0);
  328. }
  329. char *url = nullptr;
  330. CURLUcode rc = curl_url_get(url_builder, CURLUPART_URL, &url,
  331. CURLU_NO_DEFAULT_PORT);
  332. if (rc) {
  333. do_log(LOG_ERROR,
  334. "WHIP server provided a invalid resource URL via the Location header");
  335. cleanup();
  336. obs_output_signal_stop(output, OBS_OUTPUT_CONNECT_FAILED);
  337. return false;
  338. }
  339. resource_url = url;
  340. curl_free(url);
  341. do_log(LOG_DEBUG, "WHIP Resource URL is: %s", resource_url.c_str());
  342. curl_url_cleanup(url_builder);
  343. #ifdef DEBUG_SDP
  344. do_log(LOG_DEBUG, "Answer SDP:\n%s", read_buffer.c_str());
  345. #endif
  346. rtcSetRemoteDescription(peer_connection, read_buffer.c_str(), "answer");
  347. cleanup();
  348. return true;
  349. }
  350. void WHIPOutput::StartThread()
  351. {
  352. if (!Init())
  353. return;
  354. if (!Setup())
  355. return;
  356. if (!Connect()) {
  357. rtcDeletePeerConnection(peer_connection);
  358. peer_connection = -1;
  359. audio_track = -1;
  360. video_track = -1;
  361. return;
  362. }
  363. obs_output_begin_data_capture(output, 0);
  364. running = true;
  365. }
  366. void WHIPOutput::SendDelete()
  367. {
  368. if (resource_url.empty()) {
  369. do_log(LOG_DEBUG,
  370. "No resource URL available, not sending DELETE");
  371. return;
  372. }
  373. struct curl_slist *headers = NULL;
  374. if (!bearer_token.empty()) {
  375. auto bearer_token_header =
  376. std::string("Authorization: Bearer ") + bearer_token;
  377. headers =
  378. curl_slist_append(headers, bearer_token_header.c_str());
  379. }
  380. // Add user-agent to our requests
  381. headers = curl_slist_append(headers, user_agent.c_str());
  382. CURL *c = curl_easy_init();
  383. curl_easy_setopt(c, CURLOPT_HTTPHEADER, headers);
  384. curl_easy_setopt(c, CURLOPT_URL, resource_url.c_str());
  385. curl_easy_setopt(c, CURLOPT_CUSTOMREQUEST, "DELETE");
  386. curl_easy_setopt(c, CURLOPT_TIMEOUT, 8L);
  387. auto cleanup = [&]() {
  388. curl_easy_cleanup(c);
  389. curl_slist_free_all(headers);
  390. };
  391. CURLcode res = curl_easy_perform(c);
  392. if (res != CURLE_OK) {
  393. do_log(LOG_WARNING,
  394. "DELETE request for resource URL failed. Reason: %s",
  395. curl_easy_strerror(res));
  396. cleanup();
  397. return;
  398. }
  399. long response_code;
  400. curl_easy_getinfo(c, CURLINFO_RESPONSE_CODE, &response_code);
  401. if (response_code != 200) {
  402. do_log(LOG_WARNING,
  403. "DELETE request for resource URL failed. HTTP Code: %ld",
  404. response_code);
  405. cleanup();
  406. return;
  407. }
  408. do_log(LOG_DEBUG,
  409. "Successfully performed DELETE request for resource URL");
  410. resource_url.clear();
  411. cleanup();
  412. }
  413. void WHIPOutput::StopThread(bool signal)
  414. {
  415. if (peer_connection != -1) {
  416. rtcDeletePeerConnection(peer_connection);
  417. peer_connection = -1;
  418. audio_track = -1;
  419. video_track = -1;
  420. }
  421. SendDelete();
  422. /*
  423. * "signal" exists because we have to preserve the "running" state
  424. * across reconnect attempts. If we don't emit a signal if
  425. * something calls obs_output_stop() and it's reconnecting, you'll
  426. * desync the UI, as the output will be "stopped" and not
  427. * "reconnecting", but the "stop" signal will have never been
  428. * emitted.
  429. */
  430. if (running && signal) {
  431. obs_output_signal_stop(output, OBS_OUTPUT_SUCCESS);
  432. running = false;
  433. }
  434. total_bytes_sent = 0;
  435. connect_time_ms = 0;
  436. start_time_ns = 0;
  437. last_audio_timestamp = 0;
  438. last_video_timestamp = 0;
  439. }
  440. void WHIPOutput::Send(void *data, uintptr_t size, uint64_t duration, int track)
  441. {
  442. if (!running)
  443. return;
  444. // Sample time is in microseconds, we need to convert it to seconds
  445. auto elapsed_seconds = double(duration) / (1000.0 * 1000.0);
  446. // Get elapsed time in clock rate
  447. uint32_t elapsed_timestamp = 0;
  448. rtcTransformSecondsToTimestamp(track, elapsed_seconds,
  449. &elapsed_timestamp);
  450. // Set new timestamp
  451. uint32_t current_timestamp = 0;
  452. rtcGetCurrentTrackTimestamp(track, &current_timestamp);
  453. rtcSetTrackRtpTimestamp(track, current_timestamp + elapsed_timestamp);
  454. total_bytes_sent += size;
  455. rtcSendMessage(track, reinterpret_cast<const char *>(data), (int)size);
  456. }
  457. void register_whip_output()
  458. {
  459. struct obs_output_info info = {};
  460. info.id = "whip_output";
  461. info.flags = OBS_OUTPUT_AV | OBS_OUTPUT_ENCODED | OBS_OUTPUT_SERVICE;
  462. info.get_name = [](void *) -> const char * {
  463. return obs_module_text("Output.Name");
  464. };
  465. info.create = [](obs_data_t *settings, obs_output_t *output) -> void * {
  466. return new WHIPOutput(settings, output);
  467. };
  468. info.destroy = [](void *priv_data) {
  469. delete static_cast<WHIPOutput *>(priv_data);
  470. };
  471. info.start = [](void *priv_data) -> bool {
  472. return static_cast<WHIPOutput *>(priv_data)->Start();
  473. };
  474. info.stop = [](void *priv_data, uint64_t) {
  475. static_cast<WHIPOutput *>(priv_data)->Stop();
  476. };
  477. info.encoded_packet = [](void *priv_data,
  478. struct encoder_packet *packet) {
  479. static_cast<WHIPOutput *>(priv_data)->Data(packet);
  480. };
  481. info.get_defaults = [](obs_data_t *) {
  482. };
  483. info.get_properties = [](void *) -> obs_properties_t * {
  484. return obs_properties_create();
  485. };
  486. info.get_total_bytes = [](void *priv_data) -> uint64_t {
  487. return (uint64_t) static_cast<WHIPOutput *>(priv_data)
  488. ->GetTotalBytes();
  489. };
  490. info.get_connect_time_ms = [](void *priv_data) -> int {
  491. return static_cast<WHIPOutput *>(priv_data)->GetConnectTime();
  492. };
  493. info.encoded_video_codecs = "h264";
  494. info.encoded_audio_codecs = "opus";
  495. info.protocols = "WHIP";
  496. obs_register_output(&info);
  497. }