sip.c 36 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247124812491250125112521253125412551256125712581259126012611262126312641265126612671268126912701271127212731274127512761277127812791280128112821283128412851286128712881289129012911292129312941295129612971298129913001301130213031304130513061307130813091310131113121313131413151316131713181319132013211322132313241325132613271328132913301331133213331334133513361337133813391340134113421343134413451346134713481349135013511352135313541355135613571358135913601361136213631364136513661367136813691370137113721373137413751376137713781379138013811382138313841385138613871388138913901391139213931394139513961397139813991400140114021403140414051406140714081409141014111412141314141415141614171418141914201421142214231424142514261427142814291430143114321433143414351436143714381439144014411442144314441445144614471448144914501451145214531454145514561457145814591460146114621463146414651466146714681469147014711472147314741475147614771478147914801481148214831484148514861487148814891490149114921493149414951496149714981499150015011502150315041505150615071508150915101511151215131514151515161517151815191520152115221523152415251526152715281529153015311532153315341535153615371538153915401541154215431544154515461547154815491550155115521553155415551556155715581559156015611562156315641565156615671568156915701571157215731574157515761577157815791580158115821583158415851586158715881589159015911592159315941595159615971598159916001601160216031604160516061607160816091610161116121613161416151616161716181619162016211622162316241625162616271628162916301631163216331634163516361637163816391640164116421643164416451646164716481649165016511652165316541655165616571658165916601661166216631664166516661667166816691670167116721673167416751676167716781679168016811682168316841685168616871688168916901691169216931694169516961697169816991700170117021703170417051706170717081709171017111712171317141715171617171718
  1. // in Asterisk pjsip.conf, under 1001, set direct_media=no
  2. #include <stdio.h>
  3. #include <stdlib.h>
  4. #include <string.h>
  5. #include <unistd.h>
  6. #include <math.h>
  7. #include "freertos/FreeRTOS.h"
  8. #include "freertos/task.h"
  9. #include "esp_wifi.h"
  10. #include "esp_event.h"
  11. #include "esp_log.h"
  12. #include "nvs_flash.h"
  13. #include "lwip/sockets.h"
  14. #include <stdbool.h>
  15. #include "esp_rom_md5.h"
  16. #include "bsp_board.h"
  17. #include "tca9555_driver.h"
  18. #include "freertos/semphr.h"
  19. /*
  20. #include "esp_audio_enc_default.h"
  21. #include "esp_audio_enc.h"
  22. #include "esp_audio_dec_default.h"
  23. #include "esp_audio_dec.h"
  24. #include "esp_g711_enc.h"
  25. #include "esp_g711_dec.h"*/
  26. #define ESP32_IP "192.168.68.69"
  27. #define RTP_PORT 4000
  28. #define WIFI_SSID "TOCHTECH"
  29. #define WIFI_PASS "Smarturns2017"
  30. #define ASTERISK_IP "192.168.68.61"
  31. #define SIP_PORT 5060
  32. #define SIP_LOCAL_IP "0.0.0.0"
  33. #define SIP_LOCAL_PORT 5062
  34. #define SIP_USER "1001"
  35. #define SIP_PASSWORD "secret123"
  36. #define SAMPLE_RATE 16000
  37. #define RTP_SAMPLE_RATE 8000
  38. #define SAMPLES_PER_PACKET 160
  39. #define TONE_FREQ 1000
  40. #define AEC_FRAME_SAMPLES 256
  41. //#define AEC_OUTPUT_SAMPLES 320
  42. #define RTP_SAMPLES 160
  43. #define MIC_BLOCK_SAMPLES 320
  44. #define REF_BLOCK_SAMPLES 320
  45. #define REF_BUFFER_SAMPLES 4096
  46. #define MIC_BUFFER_SAMPLES 2048
  47. //
  48. #define MIC_FRAME_SAMPLES 160
  49. #define REF_DELAY_SAMPLES 320
  50. #define AEC_DELAY_SAMPLES 320
  51. static size_t ref_delay_count = AEC_DELAY_SAMPLES;
  52. static uint64_t mic_total_processed = 0;
  53. static uint64_t ref_total_written = 0;
  54. static int16_t ref_delay_buffer[AEC_DELAY_SAMPLES];
  55. static int16_t ref_buffer[REF_BUFFER_SAMPLES];
  56. static size_t ref_write = 0;
  57. static size_t ref_read = 0;
  58. static size_t ref_count = 0;
  59. static size_t ref_delay_write = 0;
  60. //static size_t ref_delay_count = 0;
  61. static int16_t mic_aec_buffer[AEC_FRAME_SAMPLES];
  62. static int16_t ref_aec_buffer[AEC_FRAME_SAMPLES];
  63. static int16_t aec_output_buffer[MIC_BUFFER_SAMPLES];
  64. static size_t aec_output_count = 0;
  65. static int32_t mic_buffer[320];
  66. static int16_t mic_pcm16[RTP_SAMPLES];
  67. static SemaphoreHandle_t aec_mutex = NULL;
  68. static int16_t ref_pending[AEC_FRAME_SAMPLES];
  69. static size_t ref_pending_count = 0;
  70. #include "esp_afe_config.h"
  71. #include "esp_afe_sr_iface.h"
  72. #include "esp_afe_sr_models.h"
  73. #include "esp_afe_aec.h"
  74. static int16_t mic_fifo[MIC_BUFFER_SAMPLES];
  75. static int16_t ref_fifo[MIC_BUFFER_SAMPLES];
  76. static size_t mic_fifo_count = 0;
  77. static size_t ref_fifo_count = 0;
  78. static uint8_t payload[MIC_FRAME_SAMPLES];
  79. static uint8_t rx_packet[172];
  80. static int16_t speaker_pcm[160];
  81. static const char *TAG = "SIP";
  82. volatile bool wifi_ready = false;
  83. static int s_rtp_sock = -1;
  84. static struct sockaddr_in s_rtp_peer = {0};
  85. static uint16_t s_rtp_peer_port = 0;
  86. static volatile bool call_active = false;
  87. static volatile bool s_audio_hw_ready = false;
  88. static afe_aec_handle_t *aec_handle = NULL;
  89. esp_err_t esp_get_feed_data(
  90. bool is_get_raw_channel,
  91. int16_t *buffer,
  92. int buffer_len);
  93. esp_err_t esp_audio_play(const int16_t *data,
  94. int length,
  95. uint32_t ticks_to_wait);
  96. esp_err_t sip_media_init(void);
  97. static void rtp_tx_task(void *pvParameters);
  98. static void rtp_rx_task(void *pvParameters);
  99. static void s_aec_process(
  100. int16_t *mic,
  101. int16_t *ref
  102. //int samples
  103. );
  104. static bool aec_reference_get(
  105. int16_t *out,
  106. size_t count
  107. );
  108. /* ================= AEC ================= */
  109. static bool aec_reference_push(
  110. const int16_t *samples,
  111. size_t count)
  112. {
  113. if (samples == NULL || count == 0)
  114. return false;
  115. if (ref_count + count > REF_BUFFER_SAMPLES)
  116. {
  117. ESP_LOGE(
  118. TAG,
  119. "AEC REF OVERFLOW: count=%u incoming=%u",
  120. (unsigned)ref_count,
  121. (unsigned)count
  122. );
  123. return false;
  124. }
  125. for (size_t i = 0; i < count; i++)
  126. {
  127. ref_buffer[ref_write] = samples[i];
  128. ref_write++;
  129. if (ref_write >= REF_BUFFER_SAMPLES)
  130. ref_write = 0;
  131. ref_count++;
  132. ref_total_written++;
  133. }
  134. ESP_LOGI(
  135. TAG,
  136. "AEC REF PUSH: count=%u fifo=%u total=%llu",
  137. (unsigned)count,
  138. (unsigned)ref_count,
  139. (unsigned long long)ref_total_written
  140. );
  141. return true;
  142. }
  143. static bool aec_reference_get(
  144. int16_t *out,
  145. size_t count)
  146. {
  147. size_t required =
  148. AEC_DELAY_SAMPLES + count;
  149. if (ref_count < AEC_DELAY_SAMPLES + count)
  150. {
  151. return false;
  152. }
  153. for (size_t i = 0; i < count; i++)
  154. {
  155. out[i] = ref_buffer[ref_read];
  156. ref_read++;
  157. if (ref_read >= REF_BUFFER_SAMPLES)
  158. ref_read = 0;
  159. }
  160. ref_count -= count;
  161. return true;
  162. }
  163. bool aec_init(void)
  164. {
  165. memset(ref_buffer, 0, sizeof(ref_buffer));
  166. ref_write = 0;
  167. ref_read = 0;
  168. ref_count = 0;
  169. ref_pending_count = 0;
  170. memset(ref_pending, 0, sizeof(ref_pending));
  171. ref_total_written = 0;
  172. mic_total_processed = 0;
  173. aec_output_count = 0;
  174. aec_handle = afe_aec_create(
  175. "MR",
  176. 4,
  177. AFE_TYPE_VC,
  178. AFE_MODE_HIGH_PERF
  179. );
  180. if (aec_handle == NULL)
  181. {
  182. ESP_LOGE(TAG, "AEC init failed");
  183. return false;
  184. }
  185. ESP_LOGI(
  186. TAG,
  187. "AEC initialized: frame_size=%d sample_rate=%d",
  188. aec_handle->frame_size,
  189. SAMPLE_RATE
  190. );
  191. if (aec_handle->frame_size != AEC_FRAME_SAMPLES)
  192. {
  193. ESP_LOGE(
  194. TAG,
  195. "AEC frame mismatch: handle=%d expected=%d",
  196. aec_handle->frame_size,
  197. AEC_FRAME_SAMPLES
  198. );
  199. afe_aec_destroy(aec_handle);
  200. aec_handle = NULL;
  201. return false;
  202. }
  203. return true;
  204. }
  205. static void s_aec_process(
  206. int16_t *mic,
  207. int16_t *ref)
  208. {
  209. if (aec_handle == NULL)
  210. return;
  211. int16_t input[AEC_FRAME_SAMPLES * 2];
  212. int16_t output[AEC_FRAME_SAMPLES];
  213. for (int i = 0; i < AEC_FRAME_SAMPLES; i++)
  214. {
  215. input[2 * i] = mic[i];
  216. input[2 * i + 1] = ref[i];
  217. }
  218. afe_aec_process(
  219. aec_handle,
  220. input,
  221. output
  222. );
  223. if (aec_output_count + AEC_FRAME_SAMPLES >
  224. MIC_BUFFER_SAMPLES)
  225. {
  226. ESP_LOGE(
  227. TAG,
  228. "AEC OUTPUT FIFO OVERFLOW: "
  229. "count=%u adding=%u max=%u",
  230. (unsigned)aec_output_count,
  231. (unsigned)AEC_FRAME_SAMPLES,
  232. (unsigned)MIC_BUFFER_SAMPLES
  233. );
  234. return;
  235. }
  236. memcpy(
  237. &aec_output_buffer[aec_output_count],
  238. output,
  239. AEC_FRAME_SAMPLES * sizeof(int16_t)
  240. );
  241. aec_output_count += AEC_FRAME_SAMPLES;
  242. }
  243. void aec_deinit()
  244. {
  245. if(aec_handle)
  246. {
  247. afe_aec_destroy(aec_handle);
  248. aec_handle=NULL;
  249. }
  250. }
  251. static bool process_aec_320(
  252. const int16_t *mic320,
  253. int16_t *out320)
  254. {
  255. static int16_t mic_pending[AEC_FRAME_SAMPLES];
  256. static size_t mic_pending_count = 0;
  257. static int16_t ref_pending[AEC_FRAME_SAMPLES];
  258. static size_t ref_pending_count = 0;
  259. size_t input_pos = 0;
  260. while (input_pos < MIC_BLOCK_SAMPLES)
  261. {
  262. if (mic_pending_count == AEC_FRAME_SAMPLES)
  263. {
  264. int16_t ref_frame[AEC_FRAME_SAMPLES];
  265. if (!aec_reference_get(
  266. ref_frame,
  267. AEC_FRAME_SAMPLES))
  268. {
  269. ESP_LOGW(
  270. TAG,
  271. "AEC waiting for reference: "
  272. "fifo=%u need=%u",
  273. (unsigned)ref_count,
  274. (unsigned)(
  275. AEC_DELAY_SAMPLES +
  276. AEC_FRAME_SAMPLES)
  277. );
  278. break;
  279. }
  280. ESP_LOGI(
  281. TAG,
  282. "AEC PROCESS: mic=256 ref=256 "
  283. "ref_remaining=%u",
  284. (unsigned)ref_count
  285. );
  286. s_aec_process(
  287. mic_pending,
  288. ref_frame
  289. );
  290. mic_total_processed += AEC_FRAME_SAMPLES;
  291. mic_pending_count = 0;
  292. continue;
  293. }
  294. size_t needed =
  295. AEC_FRAME_SAMPLES -
  296. mic_pending_count;
  297. size_t available =
  298. MIC_BLOCK_SAMPLES -
  299. input_pos;
  300. size_t copy_count =
  301. (needed < available)
  302. ? needed
  303. : available;
  304. memcpy(
  305. &mic_pending[mic_pending_count],
  306. &mic320[input_pos],
  307. copy_count * sizeof(int16_t)
  308. );
  309. mic_pending_count += copy_count;
  310. input_pos += copy_count;
  311. }
  312. if (aec_output_count < MIC_BLOCK_SAMPLES)
  313. {
  314. return false;
  315. }
  316. memcpy(
  317. out320,
  318. aec_output_buffer,
  319. MIC_BLOCK_SAMPLES * sizeof(int16_t)
  320. );
  321. size_t remaining =
  322. aec_output_count -
  323. MIC_BLOCK_SAMPLES;
  324. if (remaining > 0)
  325. {
  326. memmove(
  327. aec_output_buffer,
  328. &aec_output_buffer[MIC_BLOCK_SAMPLES],
  329. remaining * sizeof(int16_t)
  330. );
  331. }
  332. aec_output_count = remaining;
  333. return true;
  334. }
  335. static void downsample_16k_to_8k(
  336. const int16_t *input,
  337. int16_t *output)
  338. {
  339. for (int i = 0; i < 160; i++)
  340. {
  341. output[i] = input[i * 2];
  342. }
  343. }
  344. /* ================= WIFI ================= */
  345. static void wifi_event_handler(void *arg, esp_event_base_t event_base,
  346. int32_t event_id, void *event_data)
  347. {
  348. if (event_base == WIFI_EVENT && event_id == WIFI_EVENT_STA_START) {
  349. esp_wifi_connect();
  350. }
  351. else if (event_base == WIFI_EVENT && event_id == WIFI_EVENT_STA_DISCONNECTED) {
  352. ESP_LOGI(TAG, "WiFi disconnected, retrying...");
  353. wifi_ready = false;
  354. esp_wifi_connect();
  355. }
  356. else if (event_base == IP_EVENT && event_id == IP_EVENT_STA_GOT_IP) {
  357. ESP_LOGI(TAG, "WiFi connected!");
  358. wifi_ready = true;
  359. }
  360. esp_netif_ip_info_t ip;
  361. esp_netif_get_ip_info(esp_netif_get_handle_from_ifkey("WIFI_STA_DEF"), &ip);
  362. ESP_LOGI(TAG, "ESP32 IP: " IPSTR, IP2STR(&ip.ip));
  363. }
  364. void wifi_init(void)
  365. {
  366. nvs_flash_init();
  367. esp_netif_init();
  368. esp_event_loop_create_default();
  369. esp_netif_create_default_wifi_sta();
  370. wifi_init_config_t cfg = WIFI_INIT_CONFIG_DEFAULT();
  371. esp_wifi_init(&cfg);
  372. esp_event_handler_register(WIFI_EVENT, ESP_EVENT_ANY_ID, &wifi_event_handler, NULL);
  373. esp_event_handler_register(IP_EVENT, IP_EVENT_STA_GOT_IP, &wifi_event_handler, NULL);
  374. wifi_config_t wifi_config = {
  375. .sta = {
  376. .ssid = WIFI_SSID,
  377. .password = WIFI_PASS,
  378. },
  379. };
  380. esp_wifi_set_mode(WIFI_MODE_STA);
  381. esp_wifi_set_config(WIFI_IF_STA, &wifi_config);
  382. esp_wifi_start();
  383. }
  384. /* ================= MD5 HELPERS ================= */
  385. static void md5_calc(const unsigned char *input, size_t len, unsigned char output[16])
  386. {
  387. md5_context_t ctx;
  388. esp_rom_md5_init(&ctx);
  389. esp_rom_md5_update(&ctx, input, len);
  390. esp_rom_md5_final(output, &ctx);
  391. }
  392. static void md5_to_hex(unsigned char *md5, char *out)
  393. {
  394. for (int i = 0; i < 16; i++) {
  395. sprintf(out + i * 2, "%02x", md5[i]);
  396. }
  397. }
  398. /* ================= SIP DIGEST ================= */
  399. void sip_compute_response(
  400. const char *username,
  401. const char *realm,
  402. const char *password,
  403. const char *nonce,
  404. const char *uri,
  405. char *out_response)
  406. {
  407. unsigned char ha1_md5[16], ha2_md5[16], final_md5[16];
  408. char ha1_hex[64], ha2_hex[64], final_str[256];
  409. char ha1[128], ha2[128];
  410. snprintf(ha1, sizeof(ha1), "%s:%s:%s", username, realm, password);
  411. md5_calc((unsigned char*)ha1, strlen(ha1), ha1_md5);
  412. md5_to_hex(ha1_md5, ha1_hex);
  413. snprintf(ha2, sizeof(ha2), "REGISTER:%s", uri);
  414. md5_calc((unsigned char*)ha2, strlen(ha2), ha2_md5);
  415. md5_to_hex(ha2_md5, ha2_hex);
  416. snprintf(final_str, sizeof(final_str),
  417. "%s:%s:%s", ha1_hex, nonce, ha2_hex);
  418. md5_calc((unsigned char*)final_str, strlen(final_str), final_md5);
  419. md5_to_hex(final_md5, out_response);
  420. }
  421. /* ================= SIP HELPERS ================= */
  422. static bool sip_extract_header(const char *msg, const char *header, char *out, size_t out_len)
  423. {
  424. const char *start = strstr(msg, header);
  425. if (!start) {
  426. return false;
  427. }
  428. start += strlen(header);
  429. while (*start == ' ' || *start == '\t') {
  430. start++;
  431. }
  432. const char *end = start;
  433. while (*end != '\0' && *end != '\r' && *end != '\n') {
  434. end++;
  435. }
  436. size_t len = (size_t)(end - start);
  437. if (len >= out_len) {
  438. len = out_len - 1;
  439. }
  440. memcpy(out, start, len);
  441. out[len] = '\0';
  442. return true;
  443. }
  444. static void sip_send_200_ok(int sock, const struct sockaddr_in *remote, const char *sip_msg,
  445. const char *sdp_body, size_t sdp_len)
  446. {
  447. char via[512] = {0};
  448. char from[256] = {0};
  449. char to[256] = {0};
  450. char callid[256] = {0};
  451. char cseq[128] = {0};
  452. char response[4096];
  453. if (!sip_extract_header(sip_msg, "Via:", via, sizeof(via))) {
  454. snprintf(via, sizeof(via), "SIP/2.0/UDP %s:%d", SIP_LOCAL_IP, SIP_LOCAL_PORT);
  455. }
  456. if (!sip_extract_header(sip_msg, "From:", from, sizeof(from))) {
  457. snprintf(from, sizeof(from), "<sip:%s@%s>", SIP_USER, ASTERISK_IP);
  458. }
  459. if (!sip_extract_header(sip_msg, "To:", to, sizeof(to))) {
  460. snprintf(to, sizeof(to), "<sip:%s@%s>", SIP_USER, ASTERISK_IP);
  461. }
  462. if (!sip_extract_header(sip_msg, "Call-ID:", callid, sizeof(callid))) {
  463. snprintf(callid, sizeof(callid), "esp32-call");
  464. }
  465. if (!sip_extract_header(sip_msg, "CSeq:", cseq, sizeof(cseq))) {
  466. snprintf(cseq, sizeof(cseq), "1 REGISTER");
  467. }
  468. int body_len = sdp_body ? (int)sdp_len : 0;
  469. int written = snprintf(response, sizeof(response),
  470. "SIP/2.0 200 OK\r\n"
  471. "Via: %s\r\n"
  472. "From: %s\r\n"
  473. "To: %s;tag=esp32\r\n"
  474. "Call-ID: %s\r\n"
  475. "CSeq: %s\r\n"
  476. "Contact: <sip:%s@%s:%d>\r\n"
  477. "%s"
  478. "Content-Length: %d\r\n"
  479. "\r\n"
  480. "%s",
  481. via,
  482. from,
  483. to,
  484. callid,
  485. cseq,
  486. SIP_USER,
  487. ESP32_IP,
  488. SIP_LOCAL_PORT,
  489. sdp_body ? "Content-Type: application/sdp\r\n" : "",
  490. body_len,
  491. sdp_body ? sdp_body : "");
  492. if (written < 0 || written >= (int)sizeof(response)) {
  493. ESP_LOGW(TAG, "SIP 200 OK response too large");
  494. return;
  495. }
  496. sendto(sock, response, (size_t)written, 0,
  497. (struct sockaddr *)remote, sizeof(*remote));
  498. }
  499. static void sip_build_sdp(char *out, size_t out_len)
  500. {
  501. snprintf(out, out_len,
  502. "v=0\r\n"
  503. "o=ESP32 1234 1234 IN IP4 %s\r\n"
  504. "s=ESP32 SIP Call\r\n"
  505. "c=IN IP4 %s\r\n"
  506. "t=0 0\r\n"
  507. "m=audio %d RTP/AVP 0\r\n"
  508. "a=rtpmap:0 PCMU/8000\r\n"
  509. "a=ptime:20\r\n"
  510. "a=sendrecv\r\n",
  511. ESP32_IP,
  512. ESP32_IP,
  513. RTP_PORT);
  514. }
  515. static bool sip_extract_rtp_port(const char *msg, uint16_t *out_port)
  516. {
  517. const char *m = strstr(msg, "m=audio ");
  518. if (!m) {
  519. return false;
  520. }
  521. m += strlen("m=audio ");
  522. char port_buf[16] = {0};
  523. size_t i = 0;
  524. while (*m && *m != ' ' && *m != '\r' && *m != '\n' && i < sizeof(port_buf) - 1) {
  525. port_buf[i++] = *m++;
  526. }
  527. port_buf[i] = '\0';
  528. *out_port = (uint16_t)atoi(port_buf);
  529. return *out_port != 0;
  530. }
  531. /* ================= RTP TX ================= */
  532. static bool rtp_socket_init(void)
  533. {
  534. if (s_rtp_sock >= 0)
  535. return true;
  536. s_rtp_sock = socket(
  537. AF_INET,
  538. SOCK_DGRAM,
  539. IPPROTO_UDP
  540. );
  541. if (s_rtp_sock < 0)
  542. {
  543. ESP_LOGE(
  544. TAG,
  545. "RTP socket failed errno=%d",
  546. errno
  547. );
  548. return false;
  549. }
  550. struct sockaddr_in local_rtp = {0};
  551. local_rtp.sin_family = AF_INET;
  552. local_rtp.sin_port = htons(RTP_PORT);
  553. local_rtp.sin_addr.s_addr = htonl(INADDR_ANY);
  554. if (bind(
  555. s_rtp_sock,
  556. (struct sockaddr *)&local_rtp,
  557. sizeof(local_rtp)) < 0)
  558. {
  559. ESP_LOGE(
  560. TAG,
  561. "RTP bind failed errno=%d",
  562. errno
  563. );
  564. close(s_rtp_sock);
  565. s_rtp_sock = -1;
  566. return false;
  567. }
  568. ESP_LOGI(
  569. TAG,
  570. "RTP socket ready on port %d",
  571. RTP_PORT
  572. );
  573. return true;
  574. }
  575. static bool microphone_read_320(int16_t *out)
  576. {
  577. int16_t buffer1[320];
  578. int16_t buffer2[320];
  579. esp_err_t ret;
  580. ret = esp_get_feed_data(
  581. false,
  582. buffer1,
  583. sizeof(buffer1));
  584. if (ret != ESP_OK)
  585. {
  586. ESP_LOGE(TAG,
  587. "mic read 1 failed: %s",
  588. esp_err_to_name(ret));
  589. return false;
  590. }
  591. ret = esp_get_feed_data(
  592. false,
  593. buffer2,
  594. sizeof(buffer2));
  595. if (ret != ESP_OK)
  596. {
  597. ESP_LOGE(TAG,
  598. "mic read 2 failed: %s",
  599. esp_err_to_name(ret));
  600. return false;
  601. }
  602. /* ================= MIC CHANNEL DEBUG ================= */
  603. int max_l = 0;
  604. int max_r = 0;
  605. for (int i = 0; i < 160; i++)
  606. {
  607. int l = abs(buffer1[i * 2]);
  608. int r = abs(buffer1[i * 2 + 1]);
  609. if (l > max_l)
  610. max_l = l;
  611. if (r > max_r)
  612. max_r = r;
  613. }
  614. ESP_LOGI(TAG,
  615. "MIC channels: L=%d R=%d",
  616. max_l,
  617. max_r);
  618. for (int i = 0; i < 160; i++)
  619. {
  620. out[i] = buffer1[i * 2];
  621. out[160 + i] = buffer2[i * 2];
  622. }
  623. return true;
  624. }
  625. static uint8_t linear2ulaw(int16_t pcm)
  626. {
  627. const int16_t BIAS = 0x84;
  628. const int16_t CLIP = 32635;
  629. uint8_t mask;
  630. uint8_t seg;
  631. uint8_t uval;
  632. int16_t sample = pcm;
  633. if (sample < 0)
  634. {
  635. sample = -sample;
  636. mask = 0x7F;
  637. }
  638. else
  639. {
  640. mask = 0xFF;
  641. }
  642. if (sample > CLIP)
  643. sample = CLIP;
  644. sample += BIAS;
  645. if (sample <= 0xFF)
  646. seg = 0;
  647. else if (sample <= 0x1FF)
  648. seg = 1;
  649. else if (sample <= 0x3FF)
  650. seg = 2;
  651. else if (sample <= 0x7FF)
  652. seg = 3;
  653. else if (sample <= 0xFFF)
  654. seg = 4;
  655. else if (sample <= 0x1FFF)
  656. seg = 5;
  657. else if (sample <= 0x3FFF)
  658. seg = 6;
  659. else
  660. seg = 7;
  661. uval = (seg << 4) |
  662. ((sample >> (seg + 3)) & 0x0F);
  663. return uval ^ mask;
  664. }
  665. static int16_t ulaw2linear(uint8_t u_val)
  666. {
  667. u_val = ~u_val;
  668. int t = ((u_val & 0x0F) << 3) + 0x84;
  669. t <<= ((unsigned)u_val & 0x70) >> 4;
  670. if (u_val & 0x80)
  671. return 0x84 - t;
  672. else
  673. return t - 0x84;
  674. }
  675. static void sip_send_rtp_packet(void)
  676. {
  677. static uint16_t seq = 0;
  678. static uint32_t timestamp = 0;
  679. if (s_rtp_peer_port == 0)
  680. {
  681. ESP_LOGW(TAG, "No RTP peer");
  682. return;
  683. }
  684. if (s_rtp_sock < 0)
  685. {
  686. ESP_LOGE(TAG, "RTP socket not initialized");
  687. return;
  688. }
  689. static int16_t mic320[320];
  690. static int16_t aec320[320];
  691. if (!microphone_read_320(mic320))
  692. return;
  693. int raw_mic_max = 0;
  694. for (int i = 0; i < 320; i++)
  695. {
  696. int v = abs(mic320[i]);
  697. if (v > raw_mic_max)
  698. raw_mic_max = v;
  699. }
  700. ESP_LOGI(TAG,
  701. "MIC16 max=%d first=%d %d %d %d",
  702. raw_mic_max,
  703. mic320[0],
  704. mic320[1],
  705. mic320[2],
  706. mic320[3]);
  707. xSemaphoreTake(aec_mutex, portMAX_DELAY);
  708. bool ready = process_aec_320(
  709. mic320,
  710. aec320);
  711. xSemaphoreGive(aec_mutex);
  712. if (!ready)
  713. {
  714. return;
  715. }
  716. downsample_16k_to_8k(
  717. aec320,
  718. mic_pcm16);
  719. int max = 0;
  720. for (int i = 0; i < RTP_SAMPLES; i++)
  721. {
  722. int value = abs(mic_pcm16[i]);
  723. if (value > max)
  724. max = value;
  725. }
  726. ESP_LOGI(
  727. "MIC",
  728. "AEC output max=%d first=%d %d %d %d",
  729. max,
  730. mic_pcm16[0],
  731. mic_pcm16[1],
  732. mic_pcm16[2],
  733. mic_pcm16[3]);
  734. uint8_t payload[RTP_SAMPLES];
  735. for (int i = 0; i < RTP_SAMPLES; i++)
  736. {
  737. payload[i] = linear2ulaw(
  738. mic_pcm16[i]);
  739. }
  740. uint8_t packet[12 + RTP_SAMPLES];
  741. memset(packet, 0, sizeof(packet));
  742. packet[0] = 0x80;
  743. packet[1] = 0x00;
  744. packet[2] = seq >> 8;
  745. packet[3] = seq & 0xff;
  746. packet[4] = timestamp >> 24;
  747. packet[5] = timestamp >> 16;
  748. packet[6] = timestamp >> 8;
  749. packet[7] = timestamp;
  750. uint32_t ssrc = 0x12345678;
  751. packet[8] = (ssrc >> 24) & 0xff;
  752. packet[9] = (ssrc >> 16) & 0xff;
  753. packet[10] = (ssrc >> 8) & 0xff;
  754. packet[11] = ssrc & 0xff;
  755. memcpy(
  756. packet + 12,
  757. payload,
  758. RTP_SAMPLES);
  759. struct sockaddr_in peer = s_rtp_peer;
  760. peer.sin_port = htons(
  761. s_rtp_peer_port);
  762. int ret = sendto(
  763. s_rtp_sock,
  764. packet,
  765. sizeof(packet),
  766. 0,
  767. (struct sockaddr *)&peer,
  768. sizeof(peer));
  769. if (ret < 0)
  770. {
  771. ESP_LOGE(
  772. TAG,
  773. "RTP send failed errno=%d",
  774. errno);
  775. }
  776. seq++;
  777. timestamp += RTP_SAMPLES;
  778. }
  779. static void rtp_tx_task(void *pvParameters)
  780. {
  781. ESP_LOGI(TAG, "RTP TX started");
  782. TickType_t last_wake = xTaskGetTickCount();
  783. while(call_active)
  784. {/*ESP_LOGI(TAG, "RTP loop");*/
  785. sip_send_rtp_packet();
  786. vTaskDelayUntil(
  787. &last_wake,
  788. pdMS_TO_TICKS(20)
  789. );
  790. }
  791. ESP_LOGI(TAG, "RTP TX stopped");
  792. vTaskDelete(NULL);
  793. }
  794. // ================= RTP RX =================
  795. static void rtp_rx_task(void *pvParameters)
  796. {
  797. ESP_LOGI(TAG, "RTP RX started");
  798. while (call_active)
  799. {
  800. struct sockaddr_in src;
  801. socklen_t len = sizeof(src);
  802. int n = recvfrom(
  803. s_rtp_sock,
  804. rx_packet,
  805. sizeof(rx_packet),
  806. 0,
  807. (struct sockaddr *)&src,
  808. &len
  809. );
  810. if (n < 0)
  811. {
  812. ESP_LOGE(TAG,
  813. "recvfrom failed errno=%d",
  814. errno);
  815. continue;
  816. }
  817. /*
  818. ESP_LOGI(TAG,
  819. "Received RTP packet (%d bytes)",
  820. n);
  821. */
  822. if (n <= 12)
  823. {
  824. ESP_LOGW(TAG, "Packet too small");
  825. continue;
  826. }
  827. /*
  828. ESP_LOGI(TAG,
  829. "RTP Header: %02X %02X %02X %02X",
  830. rx_packet[0],
  831. rx_packet[1],
  832. rx_packet[2],
  833. rx_packet[3]);
  834. ESP_LOGI(TAG,
  835. "RTP Payload: %02X %02X %02X %02X %02X %02X %02X %02X",
  836. rx_packet[12],
  837. rx_packet[13],
  838. rx_packet[14],
  839. rx_packet[15],
  840. rx_packet[16],
  841. rx_packet[17],
  842. rx_packet[18],
  843. rx_packet[19]);
  844. */
  845. int payload_size = n - 12;
  846. if (payload_size > SAMPLES_PER_PACKET)
  847. {
  848. payload_size = SAMPLES_PER_PACKET;
  849. }
  850. int16_t play_buf[640];
  851. int16_t ref_frame[320];
  852. for (int i = 0; i < payload_size; i++)
  853. {
  854. int16_t s1 =
  855. ulaw2linear(rx_packet[12 + i]);
  856. int16_t s2 = s1;
  857. if (i < payload_size - 1)
  858. {
  859. int16_t next =
  860. ulaw2linear(
  861. rx_packet[12 + i + 1]);
  862. s2 = (s1 + next) / 2;
  863. }
  864. play_buf[4 * i + 0] = s1;
  865. play_buf[4 * i + 1] = s1;
  866. play_buf[4 * i + 2] = s2;
  867. play_buf[4 * i + 3] = s2;
  868. ref_frame[2 * i + 0] = s1;
  869. ref_frame[2 * i + 1] = s2;
  870. }
  871. xSemaphoreTake(
  872. aec_mutex,
  873. portMAX_DELAY);
  874. size_t ref_samples = payload_size * 2;
  875. size_t ref_pos = 0;
  876. while (ref_pos < ref_samples)
  877. {
  878. size_t needed =
  879. AEC_FRAME_SAMPLES - ref_pending_count;
  880. size_t available =
  881. ref_samples - ref_pos;
  882. size_t copy_count =
  883. (needed < available)
  884. ? needed
  885. : available;
  886. memcpy(
  887. &ref_pending[ref_pending_count],
  888. &ref_frame[ref_pos],
  889. copy_count * sizeof(int16_t)
  890. );
  891. ref_pending_count += copy_count;
  892. ref_pos += copy_count;
  893. if (ref_pending_count == AEC_FRAME_SAMPLES)
  894. {
  895. aec_reference_push(
  896. ref_pending,
  897. AEC_FRAME_SAMPLES
  898. );
  899. ref_pending_count = 0;
  900. }
  901. }
  902. uint64_t ref_written =
  903. ref_total_written;
  904. size_t ref_level =
  905. ref_count;
  906. xSemaphoreGive(
  907. aec_mutex);
  908. ESP_LOGI(
  909. TAG,
  910. "AEC REF: pushed=%d fifo=%d",
  911. payload_size * 2,
  912. (int)ref_level);
  913. ESP_LOGI(
  914. TAG,
  915. "AEC REF PUSH: samples=%d "
  916. "ref_count=%d total_written=%llu",
  917. payload_size * 2,
  918. (int)ref_level,
  919. (unsigned long long)ref_written);
  920. esp_err_t ret = esp_audio_play(
  921. play_buf,
  922. payload_size * 4 * sizeof(int16_t),
  923. portMAX_DELAY
  924. );
  925. /*
  926. ESP_LOGI(TAG,
  927. "esp_audio_play() = %s",
  928. esp_err_to_name(ret));*/
  929. }
  930. ESP_LOGI(TAG, "RTP RX stopped");
  931. vTaskDelete(NULL);
  932. }
  933. /* ================= SIP TASK ================= */
  934. void sip_register_task(void *pvParameters)
  935. {
  936. struct sockaddr_in server = {0};
  937. server.sin_family = AF_INET;
  938. server.sin_port = htons(SIP_PORT);
  939. server.sin_addr.s_addr = inet_addr(ASTERISK_IP);
  940. int sock = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP);
  941. struct sockaddr_in local_addr = {0};
  942. local_addr.sin_family = AF_INET;
  943. local_addr.sin_port = htons(SIP_LOCAL_PORT);
  944. local_addr.sin_addr.s_addr = INADDR_ANY;
  945. if (bind(sock,
  946. (struct sockaddr *)&local_addr,
  947. sizeof(local_addr)) < 0) {
  948. ESP_LOGE(TAG, "Failed to bind SIP port %d", SIP_LOCAL_PORT);
  949. close(sock);
  950. vTaskDelete(NULL);
  951. return;
  952. }
  953. ESP_LOGI(TAG, "SIP listening on UDP %d", SIP_LOCAL_PORT);
  954. if (sock < 0) {
  955. ESP_LOGE(TAG, "Socket failed");
  956. vTaskDelete(NULL);
  957. return;
  958. }
  959. struct timeval timeout = {
  960. .tv_sec = 300,
  961. .tv_usec = 0
  962. };
  963. setsockopt(sock, SOL_SOCKET, SO_RCVTIMEO, &timeout, sizeof(timeout));
  964. /* ================= FIRST REGISTER ================= */
  965. char register_msg[512];
  966. snprintf(register_msg, sizeof(register_msg),
  967. "REGISTER sip:%s SIP/2.0\r\n"
  968. "Via: SIP/2.0/UDP %s:%d\r\n"
  969. "From: <sip:%s@%s>;tag=1234\r\n"
  970. "To: <sip:%s@%s>\r\n"
  971. "Call-ID: esp32-0001\r\n"
  972. "CSeq: 1 REGISTER\r\n"
  973. "Contact: <sip:%s@%s:%d>\r\n"
  974. "Expires: 300\r\n"
  975. "Content-Length: 0\r\n"
  976. "\r\n",
  977. ASTERISK_IP,
  978. SIP_LOCAL_IP,
  979. SIP_LOCAL_PORT,
  980. SIP_USER,
  981. ASTERISK_IP,
  982. SIP_USER,
  983. ASTERISK_IP,
  984. SIP_USER,
  985. SIP_LOCAL_IP,
  986. SIP_LOCAL_PORT);
  987. ESP_LOGI(TAG, "Sending REGISTER...");
  988. sendto(sock, register_msg, strlen(register_msg), 0,
  989. (struct sockaddr*)&server, sizeof(server));
  990. /* ================= RECEIVE 401 / 200 ================= */
  991. char response[1024];
  992. socklen_t addr_len = sizeof(server);
  993. bool registered = false;
  994. while (!registered) {
  995. int len = recvfrom(sock, response, sizeof(response) - 1, 0,
  996. (struct sockaddr *)&server, &addr_len);
  997. if (len <= 0) {
  998. ESP_LOGE(TAG, "No response");
  999. close(sock);
  1000. vTaskDelete(NULL);
  1001. return;
  1002. }
  1003. response[len] = '\0';
  1004. // printf("SIP RESPONSE:\n%s\n", response);
  1005. ESP_LOGI(TAG, "Received SIP response (%d bytes)", len);
  1006. if (strstr(response, "SIP/2.0 401")) {
  1007. char sip_nonce[256] = {0};
  1008. char sip_realm[128] = {0};
  1009. char *n = strstr(response, "nonce=\"");
  1010. if (n) {
  1011. n += 7;
  1012. char *end = strchr(n, '"');
  1013. if (end) {
  1014. size_t nlen = end - n;
  1015. if (nlen < sizeof(sip_nonce)) {
  1016. strncpy(sip_nonce, n, nlen);
  1017. sip_nonce[nlen] = '\0';
  1018. }
  1019. }
  1020. }
  1021. char *r = strstr(response, "realm=\"");
  1022. if (r) {
  1023. r += 7;
  1024. char *rend = strchr(r, '"');
  1025. if (rend) {
  1026. size_t rlen = rend - r;
  1027. if (rlen < sizeof(sip_realm)) {
  1028. strncpy(sip_realm, r, rlen);
  1029. sip_realm[rlen] = '\0';
  1030. }
  1031. }
  1032. }
  1033. if (sip_realm[0] == '\0' || sip_nonce[0] == '\0') {
  1034. ESP_LOGE(TAG, "Missing realm or nonce in challenge");
  1035. close(sock);
  1036. vTaskDelete(NULL);
  1037. return;
  1038. }
  1039. ESP_LOGI(TAG, "Nonce: %s", sip_nonce);
  1040. ESP_LOGI(TAG, "Realm: %s", sip_realm);
  1041. char response_hash[64];
  1042. char sip_uri[64];
  1043. snprintf(sip_uri, sizeof(sip_uri), "sip:%s", ASTERISK_IP);
  1044. sip_compute_response(
  1045. SIP_USER,
  1046. sip_realm,
  1047. SIP_PASSWORD,
  1048. sip_nonce,
  1049. sip_uri,
  1050. response_hash
  1051. );
  1052. char auth[1024];
  1053. snprintf(auth, sizeof(auth),
  1054. "REGISTER sip:%s SIP/2.0\r\n"
  1055. "Via: SIP/2.0/UDP %s:%d\r\n"
  1056. "From: <sip:%s@%s>;tag=1234\r\n"
  1057. "To: <sip:%s@%s>\r\n"
  1058. "Call-ID: esp32-0001\r\n"
  1059. "CSeq: 2 REGISTER\r\n"
  1060. "Contact: <sip:%s@%s:%d>\r\n"
  1061. "Authorization: Digest username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"sip:%s\", response=\"%s\"\r\n"
  1062. "Expires: 300\r\n"
  1063. "Content-Length: 0\r\n"
  1064. "\r\n",
  1065. ASTERISK_IP,
  1066. SIP_LOCAL_IP,
  1067. SIP_LOCAL_PORT,
  1068. SIP_USER,
  1069. ASTERISK_IP,
  1070. SIP_USER,
  1071. ASTERISK_IP,
  1072. SIP_USER,
  1073. SIP_LOCAL_IP,
  1074. SIP_LOCAL_PORT,
  1075. SIP_USER,
  1076. sip_realm,
  1077. sip_nonce,
  1078. ASTERISK_IP,
  1079. response_hash);
  1080. ESP_LOGI(TAG, "Sending AUTH REGISTER...");
  1081. sendto(sock, auth, strlen(auth), 0,
  1082. (struct sockaddr *)&server, sizeof(server));
  1083. }
  1084. else if (strstr(response, "SIP/2.0 200")) {
  1085. registered = true;
  1086. }
  1087. else if (strstr(response, "SIP/2.0 486")) {
  1088. ESP_LOGW(TAG, "Server returned 486 Busy");
  1089. }
  1090. else {
  1091. ESP_LOGI(TAG, "Ignoring SIP response");
  1092. }
  1093. }
  1094. if (!registered) {
  1095. ESP_LOGW(TAG, "Registration was not completed");
  1096. close(sock);
  1097. vTaskDelete(NULL);
  1098. return;
  1099. }
  1100. ESP_LOGI(TAG, "🎉 SIP REGISTER SUCCESS");
  1101. ESP_LOGI(TAG, "stack remaining: %u",
  1102. uxTaskGetStackHighWaterMark(NULL));
  1103. /* =====================================================
  1104. * KEEP SOCKET OPEN AND WAIT FOR SIP REQUESTS
  1105. * ===================================================== */
  1106. ESP_LOGI(TAG, "Waiting for incoming SIP requests...");
  1107. while (1) {
  1108. char rx_buf[2048];
  1109. int rx_len = recvfrom(sock,
  1110. rx_buf,
  1111. sizeof(rx_buf) - 1,
  1112. 0,
  1113. (struct sockaddr *)&server,
  1114. &addr_len);
  1115. if (rx_len <= 0) {
  1116. continue;
  1117. }
  1118. rx_buf[rx_len] = '\0';
  1119. ESP_LOGI(TAG, "\n===== RAW INVITE =====\n%s\n=======================\n", rx_buf);
  1120. // printf("\n========================================\n");
  1121. // printf("SIP MESSAGE RECEIVED:\n");
  1122. // printf("%s\n", rx_buf);
  1123. ESP_LOGI(TAG, "Received SIP packet (%d bytes)", rx_len);
  1124. // printf("========================================\n");
  1125. if (strncmp(rx_buf, "REGISTER ", 9) == 0) {
  1126. ESP_LOGI(TAG, "Received REGISTER; sending 200 OK");
  1127. sip_send_200_ok(sock, &server, rx_buf, NULL, 0);
  1128. continue;
  1129. }
  1130. else if (strncmp(rx_buf, "INVITE ", 7) == 0) {
  1131. ESP_LOGI(TAG, "📞 Incoming INVITE");
  1132. char *test = strstr(rx_buf, "m=audio");
  1133. if(test)
  1134. {
  1135. ESP_LOGI(TAG, "Found m=audio SDP:");
  1136. ESP_LOGI(TAG, "%s", test);
  1137. }
  1138. else
  1139. {
  1140. ESP_LOGE(TAG, "No m=audio found");
  1141. }
  1142. /* ================= EXTRACT SIP HEADERS ================= */
  1143. char via[512] = {0};
  1144. char from[256] = {0};
  1145. char to[256] = {0};
  1146. char callid[256] = {0};
  1147. char cseq[128] = {0};
  1148. char *header = NULL;
  1149. header = strstr(rx_buf, "Via:");
  1150. if (header) {
  1151. sscanf(header, "Via: %511[^\r\n]", via);
  1152. }
  1153. header = strstr(rx_buf, "From:");
  1154. if (header) {
  1155. sscanf(header, "From: %255[^\r\n]", from);
  1156. }
  1157. header = strstr(rx_buf, "To:");
  1158. if (header) {
  1159. sscanf(header, "To: %255[^\r\n]", to);
  1160. }
  1161. header = strstr(rx_buf, "Call-ID:");
  1162. if (header) {
  1163. sscanf(header, "Call-ID: %255[^\r\n]", callid);
  1164. }
  1165. header = strstr(rx_buf, "CSeq:");
  1166. if (header) {
  1167. sscanf(header, "CSeq: %127[^\r\n]", cseq);
  1168. }
  1169. /* ================= SEND 100 TRYING ================= */
  1170. char trying[4096];
  1171. snprintf(trying, sizeof(trying),
  1172. "SIP/2.0 100 Trying\r\n"
  1173. "Via: %s\r\n"
  1174. "From: %s\r\n"
  1175. "To: %s\r\n"
  1176. "Call-ID: %s\r\n"
  1177. "CSeq: %s\r\n"
  1178. "Content-Length: 0\r\n"
  1179. "\r\n",
  1180. via,
  1181. from,
  1182. to,
  1183. callid,
  1184. cseq);
  1185. sendto(sock,
  1186. trying,
  1187. strlen(trying),
  1188. 0,
  1189. (struct sockaddr *)&server,
  1190. sizeof(server));
  1191. ESP_LOGI(TAG, "100 Trying sent");
  1192. ESP_LOGI(TAG, "stack remaining: %u",
  1193. uxTaskGetStackHighWaterMark(NULL));
  1194. char ok[4096];
  1195. char sdp[1024];
  1196. sip_build_sdp(sdp, sizeof(sdp));
  1197. if (sip_extract_rtp_port(rx_buf, &s_rtp_peer_port))
  1198. {
  1199. s_rtp_peer.sin_family = AF_INET;
  1200. s_rtp_peer.sin_port =
  1201. htons(s_rtp_peer_port);
  1202. s_rtp_peer.sin_addr.s_addr =
  1203. inet_addr(ASTERISK_IP);
  1204. ESP_LOGI(
  1205. TAG,
  1206. "RTP peer port=%d",
  1207. s_rtp_peer_port
  1208. );
  1209. if (!rtp_socket_init())
  1210. {
  1211. ESP_LOGE(
  1212. TAG,
  1213. "RTP socket initialization failed"
  1214. );
  1215. continue;
  1216. }
  1217. }
  1218. snprintf(ok, sizeof(ok),
  1219. "SIP/2.0 200 OK\r\n"
  1220. "Via: %s\r\n"
  1221. "From: %s\r\n"
  1222. "To: %s;esp32\r\n"
  1223. "Call-ID: %s\r\n"
  1224. "CSeq: %s\r\n"
  1225. "Contact: <sip:%s@%s:%d>\r\n"
  1226. "Content-Type: application/sdp\r\n"
  1227. "Content-Length: %d\r\n"
  1228. "\r\n"
  1229. "%s",
  1230. via,
  1231. from,
  1232. to,
  1233. callid,
  1234. cseq,
  1235. SIP_USER,
  1236. ESP32_IP,
  1237. SIP_LOCAL_PORT,
  1238. strlen(sdp),
  1239. sdp);
  1240. ESP_LOGI(TAG, "\n========== 200 OK ==========\n%s\n============================\n", ok);
  1241. sendto(sock,
  1242. ok,
  1243. strlen(ok),
  1244. 0,
  1245. (struct sockaddr *)&server,
  1246. sizeof(server));
  1247. ESP_LOGI(TAG, "200 OK sent");
  1248. }
  1249. else if (strncmp(rx_buf, "OPTIONS ", 8) == 0) {
  1250. ESP_LOGI(TAG, "Received OPTIONS");
  1251. }
  1252. else if (strncmp(rx_buf, "ACK ", 4) == 0)
  1253. {
  1254. ESP_LOGI(TAG, "Received ACK");
  1255. uint8_t test_values[] = {0xFF, 0x7F, 0x00, 0x80};
  1256. for(int i = 0; i < 4; i++)
  1257. {
  1258. ESP_LOGI(TAG,
  1259. "ulaw %02X -> %d",
  1260. test_values[i],
  1261. ulaw2linear(test_values[i]));
  1262. }
  1263. if (!call_active)
  1264. {
  1265. if (s_rtp_sock < 0)
  1266. {
  1267. ESP_LOGE(
  1268. TAG,
  1269. "Cannot start RTP: socket not initialized"
  1270. );
  1271. continue;
  1272. }
  1273. if (s_rtp_peer_port == 0)
  1274. {
  1275. ESP_LOGE(
  1276. TAG,
  1277. "Cannot start RTP: peer port unknown"
  1278. );
  1279. continue;
  1280. }
  1281. call_active = true;
  1282. BaseType_t tx_ret = xTaskCreate(
  1283. rtp_tx_task,
  1284. "rtp_tx",
  1285. 16384,
  1286. NULL,
  1287. 5,
  1288. NULL
  1289. );
  1290. BaseType_t rx_ret = xTaskCreate(
  1291. rtp_rx_task,
  1292. "rtp_rx",
  1293. 16384,
  1294. NULL,
  1295. 5,
  1296. NULL
  1297. );
  1298. ESP_LOGI(
  1299. TAG,
  1300. "RTP tasks: TX=%d RX=%d",
  1301. tx_ret,
  1302. rx_ret
  1303. );
  1304. }
  1305. }
  1306. else if (strncmp(rx_buf, "BYE ", 4) == 0)
  1307. {
  1308. ESP_LOGI(TAG, "Received BYE");
  1309. call_active = false;
  1310. sip_send_200_ok(sock, &server, rx_buf, NULL, 0);
  1311. }
  1312. else {
  1313. ESP_LOGI(TAG, "Unknown SIP message");
  1314. }
  1315. }
  1316. close(sock);
  1317. vTaskDelete(NULL);
  1318. }
  1319. esp_err_t sip_media_init(void)
  1320. {
  1321. if (s_audio_hw_ready) {
  1322. return ESP_OK;
  1323. }
  1324. aec_mutex = xSemaphoreCreateMutex();
  1325. if (aec_mutex == NULL)
  1326. {
  1327. ESP_LOGE(TAG, "AEC mutex failed");
  1328. return ESP_FAIL;
  1329. }
  1330. ESP_LOGI(TAG, "AEC mutex=%p", aec_mutex);
  1331. esp_err_t ret = esp_board_init(16000, 2, 32);
  1332. if (ret != ESP_OK)
  1333. {
  1334. ESP_LOGE(TAG,
  1335. "esp_board_init failed: %s",
  1336. esp_err_to_name(ret));
  1337. return ret;
  1338. }
  1339. tca9555_driver_init();
  1340. aec_init();
  1341. s_audio_hw_ready = true;
  1342. ESP_LOGI(TAG,
  1343. "Audio board initialized for live RTP media");
  1344. return ESP_OK;
  1345. }