sip.c 27 KB

12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667686970717273747576777879808182838485868788899091929394959697989910010110210310410510610710810911011111211311411511611711811912012112212312412512612712812913013113213313413513613713813914014114214314414514614714814915015115215315415515615715815916016116216316416516616716816917017117217317417517617717817918018118218318418518618718818919019119219319419519619719819920020120220320420520620720820921021121221321421521621721821922022122222322422522622722822923023123223323423523623723823924024124224324424524624724824925025125225325425525625725825926026126226326426526626726826927027127227327427527627727827928028128228328428528628728828929029129229329429529629729829930030130230330430530630730830931031131231331431531631731831932032132232332432532632732832933033133233333433533633733833934034134234334434534634734834935035135235335435535635735835936036136236336436536636736836937037137237337437537637737837938038138238338438538638738838939039139239339439539639739839940040140240340440540640740840941041141241341441541641741841942042142242342442542642742842943043143243343443543643743843944044144244344444544644744844945045145245345445545645745845946046146246346446546646746846947047147247347447547647747847948048148248348448548648748848949049149249349449549649749849950050150250350450550650750850951051151251351451551651751851952052152252352452552652752852953053153253353453553653753853954054154254354454554654754854955055155255355455555655755855956056156256356456556656756856957057157257357457557657757857958058158258358458558658758858959059159259359459559659759859960060160260360460560660760860961061161261361461561661761861962062162262362462562662762862963063163263363463563663763863964064164264364464564664764864965065165265365465565665765865966066166266366466566666766866967067167267367467567667767867968068168268368468568668768868969069169269369469569669769869970070170270370470570670770870971071171271371471571671771871972072172272372472572672772872973073173273373473573673773873974074174274374474574674774874975075175275375475575675775875976076176276376476576676776876977077177277377477577677777877978078178278378478578678778878979079179279379479579679779879980080180280380480580680780880981081181281381481581681781881982082182282382482582682782882983083183283383483583683783883984084184284384484584684784884985085185285385485585685785885986086186286386486586686786886987087187287387487587687787887988088188288388488588688788888989089189289389489589689789889990090190290390490590690790890991091191291391491591691791891992092192292392492592692792892993093193293393493593693793893994094194294394494594694794894995095195295395495595695795895996096196296396496596696796896997097197297397497597697797897998098198298398498598698798898999099199299399499599699799899910001001100210031004100510061007100810091010101110121013101410151016101710181019102010211022102310241025102610271028102910301031103210331034103510361037103810391040104110421043104410451046104710481049105010511052105310541055105610571058105910601061106210631064106510661067106810691070107110721073107410751076107710781079108010811082108310841085108610871088108910901091109210931094109510961097109810991100110111021103110411051106110711081109111011111112111311141115111611171118111911201121112211231124112511261127112811291130113111321133113411351136113711381139114011411142114311441145114611471148114911501151115211531154115511561157115811591160116111621163116411651166116711681169117011711172117311741175117611771178117911801181118211831184118511861187118811891190119111921193119411951196119711981199120012011202120312041205120612071208120912101211121212131214121512161217121812191220122112221223122412251226122712281229123012311232123312341235123612371238123912401241124212431244124512461247
  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. #define ESP32_IP "192.168.68.122"
  18. #define RTP_PORT 4000
  19. #define WIFI_SSID "TOCHTECH"
  20. #define WIFI_PASS "Smarturns2017"
  21. #define ASTERISK_IP "192.168.68.121"
  22. #define SIP_PORT 5060
  23. #define SIP_LOCAL_IP "0.0.0.0"
  24. #define SIP_LOCAL_PORT 5062
  25. #define SIP_USER "1001"
  26. #define SIP_PASSWORD "secret123"
  27. #define SAMPLE_RATE 8000
  28. #define SAMPLES_PER_PACKET 160 // 20 ms
  29. #define TONE_FREQ 1000
  30. #define MIC_FRAME_SAMPLES 160
  31. static int32_t mic_buffer[320];
  32. static int16_t mic_pcm16[MIC_FRAME_SAMPLES];
  33. static uint8_t payload[MIC_FRAME_SAMPLES];
  34. static uint8_t rx_packet[172];
  35. static int16_t speaker_pcm[160];
  36. static const char *TAG = "SIP";
  37. volatile bool wifi_ready = false;
  38. static int s_rtp_sock = -1;
  39. static struct sockaddr_in s_rtp_peer = {0};
  40. static uint16_t s_rtp_peer_port = 0;
  41. static volatile bool call_active = false;
  42. esp_err_t esp_get_feed_data(
  43. bool is_get_raw_channel,
  44. int16_t *buffer,
  45. int buffer_len);
  46. esp_err_t esp_audio_play(const int16_t *data,
  47. int length,
  48. uint32_t ticks_to_wait);
  49. static void rtp_tx_task(void *pvParameters);
  50. static void rtp_rx_task(void *pvParameters);
  51. /* ================= WIFI ================= */
  52. static void wifi_event_handler(void *arg, esp_event_base_t event_base,
  53. int32_t event_id, void *event_data)
  54. {
  55. if (event_base == WIFI_EVENT && event_id == WIFI_EVENT_STA_START) {
  56. esp_wifi_connect();
  57. }
  58. else if (event_base == WIFI_EVENT && event_id == WIFI_EVENT_STA_DISCONNECTED) {
  59. ESP_LOGI(TAG, "WiFi disconnected, retrying...");
  60. wifi_ready = false;
  61. esp_wifi_connect();
  62. }
  63. else if (event_base == IP_EVENT && event_id == IP_EVENT_STA_GOT_IP) {
  64. ESP_LOGI(TAG, "WiFi connected!");
  65. wifi_ready = true;
  66. }
  67. esp_netif_ip_info_t ip;
  68. esp_netif_get_ip_info(esp_netif_get_handle_from_ifkey("WIFI_STA_DEF"), &ip);
  69. ESP_LOGI(TAG, "ESP32 IP: " IPSTR, IP2STR(&ip.ip));
  70. }
  71. void wifi_init(void)
  72. {
  73. nvs_flash_init();
  74. esp_netif_init();
  75. esp_event_loop_create_default();
  76. esp_netif_create_default_wifi_sta();
  77. wifi_init_config_t cfg = WIFI_INIT_CONFIG_DEFAULT();
  78. esp_wifi_init(&cfg);
  79. esp_event_handler_register(WIFI_EVENT, ESP_EVENT_ANY_ID, &wifi_event_handler, NULL);
  80. esp_event_handler_register(IP_EVENT, IP_EVENT_STA_GOT_IP, &wifi_event_handler, NULL);
  81. wifi_config_t wifi_config = {
  82. .sta = {
  83. .ssid = WIFI_SSID,
  84. .password = WIFI_PASS,
  85. },
  86. };
  87. esp_wifi_set_mode(WIFI_MODE_STA);
  88. esp_wifi_set_config(WIFI_IF_STA, &wifi_config);
  89. esp_wifi_start();
  90. }
  91. /* ================= MD5 HELPERS ================= */
  92. static void md5_calc(const unsigned char *input, size_t len, unsigned char output[16])
  93. {
  94. md5_context_t ctx;
  95. esp_rom_md5_init(&ctx);
  96. esp_rom_md5_update(&ctx, input, len);
  97. esp_rom_md5_final(output, &ctx);
  98. }
  99. static void md5_to_hex(unsigned char *md5, char *out)
  100. {
  101. for (int i = 0; i < 16; i++) {
  102. sprintf(out + i * 2, "%02x", md5[i]);
  103. }
  104. }
  105. /* ================= SIP DIGEST ================= */
  106. void sip_compute_response(
  107. const char *username,
  108. const char *realm,
  109. const char *password,
  110. const char *nonce,
  111. const char *uri,
  112. char *out_response)
  113. {
  114. unsigned char ha1_md5[16], ha2_md5[16], final_md5[16];
  115. char ha1_hex[64], ha2_hex[64], final_str[256];
  116. char ha1[128], ha2[128];
  117. snprintf(ha1, sizeof(ha1), "%s:%s:%s", username, realm, password);
  118. md5_calc((unsigned char*)ha1, strlen(ha1), ha1_md5);
  119. md5_to_hex(ha1_md5, ha1_hex);
  120. snprintf(ha2, sizeof(ha2), "REGISTER:%s", uri);
  121. md5_calc((unsigned char*)ha2, strlen(ha2), ha2_md5);
  122. md5_to_hex(ha2_md5, ha2_hex);
  123. snprintf(final_str, sizeof(final_str),
  124. "%s:%s:%s", ha1_hex, nonce, ha2_hex);
  125. md5_calc((unsigned char*)final_str, strlen(final_str), final_md5);
  126. md5_to_hex(final_md5, out_response);
  127. }
  128. /* ================= SIP HELPERS ================= */
  129. static bool sip_extract_header(const char *msg, const char *header, char *out, size_t out_len)
  130. {
  131. const char *start = strstr(msg, header);
  132. if (!start) {
  133. return false;
  134. }
  135. start += strlen(header);
  136. while (*start == ' ' || *start == '\t') {
  137. start++;
  138. }
  139. const char *end = start;
  140. while (*end != '\0' && *end != '\r' && *end != '\n') {
  141. end++;
  142. }
  143. size_t len = (size_t)(end - start);
  144. if (len >= out_len) {
  145. len = out_len - 1;
  146. }
  147. memcpy(out, start, len);
  148. out[len] = '\0';
  149. return true;
  150. }
  151. static void sip_send_200_ok(int sock, const struct sockaddr_in *remote, const char *sip_msg,
  152. const char *sdp_body, size_t sdp_len)
  153. {
  154. char via[512] = {0};
  155. char from[256] = {0};
  156. char to[256] = {0};
  157. char callid[256] = {0};
  158. char cseq[128] = {0};
  159. char response[4096];
  160. if (!sip_extract_header(sip_msg, "Via:", via, sizeof(via))) {
  161. snprintf(via, sizeof(via), "SIP/2.0/UDP %s:%d", SIP_LOCAL_IP, SIP_LOCAL_PORT);
  162. }
  163. if (!sip_extract_header(sip_msg, "From:", from, sizeof(from))) {
  164. snprintf(from, sizeof(from), "<sip:%s@%s>", SIP_USER, ASTERISK_IP);
  165. }
  166. if (!sip_extract_header(sip_msg, "To:", to, sizeof(to))) {
  167. snprintf(to, sizeof(to), "<sip:%s@%s>", SIP_USER, ASTERISK_IP);
  168. }
  169. if (!sip_extract_header(sip_msg, "Call-ID:", callid, sizeof(callid))) {
  170. snprintf(callid, sizeof(callid), "esp32-call");
  171. }
  172. if (!sip_extract_header(sip_msg, "CSeq:", cseq, sizeof(cseq))) {
  173. snprintf(cseq, sizeof(cseq), "1 REGISTER");
  174. }
  175. int body_len = sdp_body ? (int)sdp_len : 0;
  176. int written = snprintf(response, sizeof(response),
  177. "SIP/2.0 200 OK\r\n"
  178. "Via: %s\r\n"
  179. "From: %s\r\n"
  180. "To: %s;tag=esp32\r\n"
  181. "Call-ID: %s\r\n"
  182. "CSeq: %s\r\n"
  183. "Contact: <sip:%s@%s:%d>\r\n"
  184. "%s"
  185. "Content-Length: %d\r\n"
  186. "\r\n"
  187. "%s",
  188. via,
  189. from,
  190. to,
  191. callid,
  192. cseq,
  193. SIP_USER,
  194. ESP32_IP,
  195. SIP_LOCAL_PORT,
  196. sdp_body ? "Content-Type: application/sdp\r\n" : "",
  197. body_len,
  198. sdp_body ? sdp_body : "");
  199. if (written < 0 || written >= (int)sizeof(response)) {
  200. ESP_LOGW(TAG, "SIP 200 OK response too large");
  201. return;
  202. }
  203. sendto(sock, response, (size_t)written, 0,
  204. (struct sockaddr *)remote, sizeof(*remote));
  205. }
  206. static void sip_build_sdp(char *out, size_t out_len)
  207. {
  208. snprintf(out, out_len,
  209. "v=0\r\n"
  210. "o=ESP32 1234 1234 IN IP4 %s\r\n"
  211. "s=ESP32 SIP Call\r\n"
  212. "c=IN IP4 %s\r\n"
  213. "t=0 0\r\n"
  214. "m=audio %d RTP/AVP 0\r\n"
  215. "a=rtpmap:0 PCMU/8000\r\n"
  216. "a=ptime:20\r\n"
  217. "a=sendrecv\r\n",
  218. ESP32_IP,
  219. ESP32_IP,
  220. RTP_PORT);
  221. }
  222. static bool sip_extract_rtp_port(const char *msg, uint16_t *out_port)
  223. {
  224. const char *m = strstr(msg, "m=audio ");
  225. if (!m) {
  226. return false;
  227. }
  228. m += strlen("m=audio ");
  229. char port_buf[16] = {0};
  230. size_t i = 0;
  231. while (*m && *m != ' ' && *m != '\r' && *m != '\n' && i < sizeof(port_buf) - 1) {
  232. port_buf[i++] = *m++;
  233. }
  234. port_buf[i] = '\0';
  235. *out_port = (uint16_t)atoi(port_buf);
  236. return *out_port != 0;
  237. }
  238. /* ================= RTP TX ================= */
  239. static bool microphone_read(int16_t *samples, size_t count)
  240. {
  241. ESP_LOGI(TAG,
  242. "mic ptr=%p samples=%d bytes=%d",
  243. samples,
  244. count,
  245. count * sizeof(int16_t));
  246. esp_err_t ret = esp_get_feed_data(
  247. false,
  248. (int16_t *)mic_buffer,
  249. sizeof(mic_buffer));
  250. if(ret != ESP_OK)
  251. {
  252. ESP_LOGE(TAG,
  253. "esp_get_feed_data failed: %s",
  254. esp_err_to_name(ret));
  255. return false;
  256. }
  257. return true;
  258. }
  259. static uint8_t linear2ulaw(int16_t pcm)
  260. {
  261. const int16_t BIAS = 0x84;
  262. const int16_t CLIP = 32635;
  263. uint8_t mask;
  264. uint8_t seg;
  265. uint8_t uval;
  266. int16_t sample = pcm;
  267. if (sample < 0)
  268. {
  269. sample = -sample;
  270. mask = 0x7F;
  271. }
  272. else
  273. {
  274. mask = 0xFF;
  275. }
  276. if (sample > CLIP)
  277. sample = CLIP;
  278. sample += BIAS;
  279. if (sample <= 0xFF)
  280. seg = 0;
  281. else if (sample <= 0x1FF)
  282. seg = 1;
  283. else if (sample <= 0x3FF)
  284. seg = 2;
  285. else if (sample <= 0x7FF)
  286. seg = 3;
  287. else if (sample <= 0xFFF)
  288. seg = 4;
  289. else if (sample <= 0x1FFF)
  290. seg = 5;
  291. else if (sample <= 0x3FFF)
  292. seg = 6;
  293. else
  294. seg = 7;
  295. uval = (seg << 4) |
  296. ((sample >> (seg + 3)) & 0x0F);
  297. return uval ^ mask;
  298. }
  299. static int16_t ulaw2linear(uint8_t u_val)
  300. {
  301. u_val = ~u_val;
  302. int t = ((u_val & 0x0F) << 3) + 0x84;
  303. t <<= ((unsigned)u_val & 0x70) >> 4;
  304. if (u_val & 0x80)
  305. return 0x84 - t;
  306. else
  307. return t - 0x84;
  308. }
  309. static void sip_send_rtp_packet(void)
  310. {
  311. static uint32_t call_count = 0;
  312. /*
  313. ESP_LOGI(TAG,
  314. "sip_send_rtp_packet() #%lu",
  315. (unsigned long)call_count++);*/
  316. static uint16_t seq = 0;
  317. static uint32_t timestamp = 0;
  318. if(s_rtp_peer_port == 0)
  319. {
  320. ESP_LOGW(TAG,"No RTP peer");
  321. return;
  322. }
  323. if(s_rtp_peer_port == 0)
  324. {
  325. ESP_LOGW(TAG,
  326. "No RTP peer port set!");
  327. return;
  328. }
  329. if(s_rtp_sock < 0)
  330. {
  331. s_rtp_sock = socket(AF_INET,
  332. SOCK_DGRAM,
  333. IPPROTO_UDP);
  334. ESP_LOGI(TAG,
  335. "RTP socket=%d",
  336. s_rtp_sock);
  337. struct sockaddr_in local_rtp = {0};
  338. local_rtp.sin_family = AF_INET;
  339. local_rtp.sin_port = htons(RTP_PORT);
  340. local_rtp.sin_addr.s_addr = INADDR_ANY;
  341. if(bind(s_rtp_sock,
  342. (struct sockaddr *)&local_rtp,
  343. sizeof(local_rtp)) < 0)
  344. {
  345. ESP_LOGE(TAG, "RTP bind failed errno=%d", errno);
  346. }
  347. else
  348. {
  349. ESP_LOGI(TAG,
  350. "RTP listening on port %d",
  351. RTP_PORT);
  352. }
  353. if(s_rtp_sock < 0)
  354. {
  355. ESP_LOGE(TAG,
  356. "RTP socket failed");
  357. return;
  358. }
  359. }
  360. uint8_t payload[SAMPLES_PER_PACKET];
  361. // Read microphone data
  362. if(!microphone_read(
  363. (int16_t *)mic_buffer,
  364. sizeof(mic_buffer)))
  365. {
  366. return;
  367. }
  368. int16_t *mic16 = (int16_t *)mic_buffer;
  369. /*
  370. ESP_LOGI("MIC",
  371. "pcm16=%d %d %d %d",
  372. mic16[0],
  373. mic16[1],
  374. mic16[2],
  375. mic16[3]);*/
  376. // Convert 32-bit audio to PCM16
  377. for(int i = 0; i < SAMPLES_PER_PACKET; i++)
  378. {
  379. // Take microphone channel
  380. int32_t s = mic_buffer[i * 2];
  381. // Trying these one at a time
  382. //mic_pcm16[i] = s >> 16;
  383. //mic_pcm16[i] = s >> 15;
  384. mic_pcm16[i] = s >> 14;
  385. }
  386. // Debug microphone level
  387. int max = 0;
  388. for(int i = 0; i < SAMPLES_PER_PACKET; i++)
  389. {
  390. int value = abs(mic_pcm16[i]);
  391. if(value > max)
  392. max = value;
  393. }
  394. /*
  395. ESP_LOGI("MIC", "max=%d", max);*/
  396. // PCM16 -> PCMU (G711 u-law)
  397. for(int i = 0; i < SAMPLES_PER_PACKET; i++)
  398. {
  399. payload[i] = linear2ulaw(mic_pcm16[i]);
  400. }
  401. //testing multiple packets
  402. /*
  403. ESP_LOGI("MIC",
  404. "%6d %6d %6d %6d %6d %6d %6d %6d "
  405. "%6d %6d %6d %6d %6d %6d %6d %6d "
  406. "%6d %6d %6d %6d",
  407. mic_pcm16[0], mic_pcm16[1], mic_pcm16[2], mic_pcm16[3],
  408. mic_pcm16[4], mic_pcm16[5], mic_pcm16[6], mic_pcm16[7],
  409. mic_pcm16[8], mic_pcm16[9], mic_pcm16[10], mic_pcm16[11],
  410. mic_pcm16[12], mic_pcm16[13], mic_pcm16[14], mic_pcm16[15],
  411. mic_pcm16[16], mic_pcm16[17], mic_pcm16[18], mic_pcm16[19]);
  412. */
  413. uint8_t packet[12 + SAMPLES_PER_PACKET];
  414. memset(packet,0,sizeof(packet));
  415. packet[0] = 0x80; // RTP version
  416. packet[1] = 0x00; // PCMU
  417. packet[2] = seq >> 8;
  418. packet[3] = seq & 0xff;
  419. packet[4] = timestamp >> 24;
  420. packet[5] = timestamp >> 16;
  421. packet[6] = timestamp >> 8;
  422. packet[7] = timestamp;
  423. uint32_t ssrc = 0x12345678;
  424. packet[8] = (ssrc >> 24) & 0xff;
  425. packet[9] = (ssrc >> 16) & 0xff;
  426. packet[10] = (ssrc >> 8) & 0xff;
  427. packet[11] = ssrc & 0xff;
  428. memcpy(packet+12,
  429. payload,
  430. SAMPLES_PER_PACKET);
  431. struct sockaddr_in peer = s_rtp_peer;
  432. peer.sin_port = htons(s_rtp_peer_port);
  433. /*ESP_LOGI(TAG,
  434. "RTP destination IP: %s",
  435. inet_ntoa(peer.sin_addr));
  436. ESP_LOGI(TAG,
  437. "Sending RTP to %s:%d",
  438. inet_ntoa(peer.sin_addr),
  439. ntohs(peer.sin_port));*/
  440. esp_netif_ip_info_t ip;
  441. esp_netif_t *netif = esp_netif_get_handle_from_ifkey("WIFI_STA_DEF");
  442. esp_netif_get_ip_info(netif, &ip);
  443. /*ESP_LOGI(TAG, "Current IP: " IPSTR, IP2STR(&ip.ip));*/
  444. int ret =
  445. sendto(
  446. s_rtp_sock,
  447. packet,
  448. sizeof(packet),
  449. 0,
  450. (struct sockaddr *)&peer,
  451. sizeof(peer));
  452. if(ret < 0)
  453. {
  454. ESP_LOGE(TAG,
  455. "RTP send failed errno=%d",
  456. errno);
  457. }
  458. else
  459. {
  460. /* ESP_LOGI(TAG,
  461. "RTP sent bytes=%d seq=%d",
  462. ret,
  463. seq);*/
  464. }
  465. /*if(ret > 0)
  466. {
  467. ESP_LOGI(TAG,
  468. "RTP sent seq=%d",
  469. seq);
  470. }*/
  471. seq++;
  472. timestamp += SAMPLES_PER_PACKET;
  473. }
  474. static void rtp_tx_task(void *pvParameters)
  475. {
  476. ESP_LOGI(TAG, "RTP TX started");
  477. TickType_t last_wake = xTaskGetTickCount();
  478. while(call_active)
  479. {/*ESP_LOGI(TAG, "RTP loop");*/
  480. sip_send_rtp_packet();
  481. vTaskDelayUntil(
  482. &last_wake,
  483. pdMS_TO_TICKS(20)
  484. );
  485. }
  486. ESP_LOGI(TAG, "RTP TX stopped");
  487. vTaskDelete(NULL);
  488. }
  489. /*static void rtp_rx_task(void *pvParameters)
  490. {
  491. ESP_LOGI(TAG, "RTP RX started");
  492. // TEMP: create RTP socket for RX-only testing
  493. if (s_rtp_sock < 0)
  494. {
  495. s_rtp_sock = socket(AF_INET,
  496. SOCK_DGRAM,
  497. IPPROTO_UDP);
  498. if (s_rtp_sock < 0)
  499. {
  500. ESP_LOGE(TAG,
  501. "RTP socket failed errno=%d",
  502. errno);
  503. vTaskDelete(NULL);
  504. return;
  505. }
  506. struct sockaddr_in local_rtp = {0};
  507. local_rtp.sin_family = AF_INET;
  508. local_rtp.sin_port = htons(RTP_PORT);
  509. local_rtp.sin_addr.s_addr = INADDR_ANY;
  510. if (bind(s_rtp_sock,
  511. (struct sockaddr *)&local_rtp,
  512. sizeof(local_rtp)) < 0)
  513. {
  514. ESP_LOGE(TAG,
  515. "RTP bind failed errno=%d",
  516. errno);
  517. close(s_rtp_sock);
  518. s_rtp_sock = -1;
  519. vTaskDelete(NULL);
  520. return;
  521. }
  522. ESP_LOGI(TAG,
  523. "RTP RX listening on port %d",
  524. RTP_PORT);
  525. }
  526. while (call_active)
  527. {
  528. struct sockaddr_in src;
  529. socklen_t len = sizeof(src);
  530. int n = recvfrom(
  531. s_rtp_sock,
  532. rx_packet,
  533. sizeof(rx_packet),
  534. 0,
  535. (struct sockaddr *)&src,
  536. &len);*/
  537. //comment this block out for testing rx only
  538. static void rtp_rx_task(void *pvParameters)
  539. {
  540. ESP_LOGI(TAG, "RTP RX started");
  541. while (call_active)
  542. {
  543. struct sockaddr_in src;
  544. socklen_t len = sizeof(src);
  545. int n = recvfrom(
  546. s_rtp_sock,
  547. rx_packet,
  548. sizeof(rx_packet),
  549. 0,
  550. (struct sockaddr *)&src,
  551. &len);//
  552. if (n < 0)
  553. {
  554. ESP_LOGE(TAG, "recvfrom failed errno=%d", errno);
  555. continue;
  556. }
  557. ESP_LOGI(TAG, "Received RTP packet (%d bytes)", n);
  558. if (n <= 12)
  559. {
  560. ESP_LOGW(TAG, "Packet too small");
  561. continue;
  562. }
  563. ESP_LOGI(TAG,
  564. "RTP Header: %02X %02X %02X %02X",
  565. rx_packet[0],
  566. rx_packet[1],
  567. rx_packet[2],
  568. rx_packet[3]);
  569. ESP_LOGI(TAG,
  570. "RTP Payload: %02X %02X %02X %02X %02X %02X %02X %02X",
  571. rx_packet[12],
  572. rx_packet[13],
  573. rx_packet[14],
  574. rx_packet[15],
  575. rx_packet[16],
  576. rx_packet[17],
  577. rx_packet[18],
  578. rx_packet[19]);
  579. int payload_size = n - 12;
  580. if (payload_size > SAMPLES_PER_PACKET)
  581. {
  582. payload_size = SAMPLES_PER_PACKET;
  583. }
  584. int16_t play_buf[640];
  585. for (int i = 0; i < payload_size; i++)
  586. {
  587. int16_t s1 = ulaw2linear(rx_packet[12+i]);
  588. int16_t s2 = s1;
  589. if (i < payload_size - 1)
  590. {
  591. int16_t next = ulaw2linear(rx_packet[12+i+1]);
  592. s2 = (s1 + next) / 2;
  593. }
  594. // sample 1
  595. play_buf[4*i + 0] = s1; // Left
  596. play_buf[4*i + 1] = s1; // Right
  597. // interpolated sample
  598. play_buf[4*i + 2] = s2; // Left
  599. play_buf[4*i + 3] = s2; // Right
  600. }
  601. esp_err_t ret = esp_audio_play(
  602. play_buf,
  603. payload_size * 4 * sizeof(int16_t),
  604. portMAX_DELAY);
  605. ESP_LOGI(TAG,
  606. "esp_audio_play() = %s",
  607. esp_err_to_name(ret));
  608. ESP_LOGI(TAG, "esp_audio_play() returned %s",
  609. esp_err_to_name(ret));
  610. }
  611. ESP_LOGI(TAG, "RTP RX stopped");
  612. vTaskDelete(NULL);
  613. }
  614. /* ================= SIP TASK ================= */
  615. void sip_register_task(void *pvParameters)
  616. {
  617. struct sockaddr_in server = {0};
  618. server.sin_family = AF_INET;
  619. server.sin_port = htons(SIP_PORT);
  620. server.sin_addr.s_addr = inet_addr(ASTERISK_IP);
  621. int sock = socket(AF_INET, SOCK_DGRAM, IPPROTO_UDP);
  622. struct sockaddr_in local_addr = {0};
  623. local_addr.sin_family = AF_INET;
  624. local_addr.sin_port = htons(SIP_LOCAL_PORT);
  625. local_addr.sin_addr.s_addr = INADDR_ANY;
  626. if (bind(sock,
  627. (struct sockaddr *)&local_addr,
  628. sizeof(local_addr)) < 0) {
  629. ESP_LOGE(TAG, "Failed to bind SIP port %d", SIP_LOCAL_PORT);
  630. close(sock);
  631. vTaskDelete(NULL);
  632. return;
  633. }
  634. ESP_LOGI(TAG, "SIP listening on UDP %d", SIP_LOCAL_PORT);
  635. if (sock < 0) {
  636. ESP_LOGE(TAG, "Socket failed");
  637. vTaskDelete(NULL);
  638. return;
  639. }
  640. struct timeval timeout = {
  641. .tv_sec = 300,
  642. .tv_usec = 0
  643. };
  644. setsockopt(sock, SOL_SOCKET, SO_RCVTIMEO, &timeout, sizeof(timeout));
  645. /* ================= FIRST REGISTER ================= */
  646. char register_msg[512];
  647. snprintf(register_msg, sizeof(register_msg),
  648. "REGISTER sip:%s SIP/2.0\r\n"
  649. "Via: SIP/2.0/UDP %s:%d\r\n"
  650. "From: <sip:%s@%s>;tag=1234\r\n"
  651. "To: <sip:%s@%s>\r\n"
  652. "Call-ID: esp32-0001\r\n"
  653. "CSeq: 1 REGISTER\r\n"
  654. "Contact: <sip:%s@%s:%d>\r\n"
  655. "Expires: 300\r\n"
  656. "Content-Length: 0\r\n"
  657. "\r\n",
  658. ASTERISK_IP,
  659. SIP_LOCAL_IP,
  660. SIP_LOCAL_PORT,
  661. SIP_USER,
  662. ASTERISK_IP,
  663. SIP_USER,
  664. ASTERISK_IP,
  665. SIP_USER,
  666. SIP_LOCAL_IP,
  667. SIP_LOCAL_PORT);
  668. ESP_LOGI(TAG, "Sending REGISTER...");
  669. sendto(sock, register_msg, strlen(register_msg), 0,
  670. (struct sockaddr*)&server, sizeof(server));
  671. /* ================= RECEIVE 401 / 200 ================= */
  672. char response[1024];
  673. socklen_t addr_len = sizeof(server);
  674. bool registered = false;
  675. while (!registered) {
  676. int len = recvfrom(sock, response, sizeof(response) - 1, 0,
  677. (struct sockaddr *)&server, &addr_len);
  678. if (len <= 0) {
  679. ESP_LOGE(TAG, "No response");
  680. close(sock);
  681. vTaskDelete(NULL);
  682. return;
  683. }
  684. response[len] = '\0';
  685. // printf("SIP RESPONSE:\n%s\n", response);
  686. ESP_LOGI(TAG, "Received SIP response (%d bytes)", len);
  687. if (strstr(response, "SIP/2.0 401")) {
  688. char sip_nonce[256] = {0};
  689. char sip_realm[128] = {0};
  690. char *n = strstr(response, "nonce=\"");
  691. if (n) {
  692. n += 7;
  693. char *end = strchr(n, '"');
  694. if (end) {
  695. size_t nlen = end - n;
  696. if (nlen < sizeof(sip_nonce)) {
  697. strncpy(sip_nonce, n, nlen);
  698. sip_nonce[nlen] = '\0';
  699. }
  700. }
  701. }
  702. char *r = strstr(response, "realm=\"");
  703. if (r) {
  704. r += 7;
  705. char *rend = strchr(r, '"');
  706. if (rend) {
  707. size_t rlen = rend - r;
  708. if (rlen < sizeof(sip_realm)) {
  709. strncpy(sip_realm, r, rlen);
  710. sip_realm[rlen] = '\0';
  711. }
  712. }
  713. }
  714. if (sip_realm[0] == '\0' || sip_nonce[0] == '\0') {
  715. ESP_LOGE(TAG, "Missing realm or nonce in challenge");
  716. close(sock);
  717. vTaskDelete(NULL);
  718. return;
  719. }
  720. ESP_LOGI(TAG, "Nonce: %s", sip_nonce);
  721. ESP_LOGI(TAG, "Realm: %s", sip_realm);
  722. char response_hash[64];
  723. char sip_uri[64];
  724. snprintf(sip_uri, sizeof(sip_uri), "sip:%s", ASTERISK_IP);
  725. sip_compute_response(
  726. SIP_USER,
  727. sip_realm,
  728. SIP_PASSWORD,
  729. sip_nonce,
  730. sip_uri,
  731. response_hash
  732. );
  733. char auth[1024];
  734. snprintf(auth, sizeof(auth),
  735. "REGISTER sip:%s SIP/2.0\r\n"
  736. "Via: SIP/2.0/UDP %s:%d\r\n"
  737. "From: <sip:%s@%s>;tag=1234\r\n"
  738. "To: <sip:%s@%s>\r\n"
  739. "Call-ID: esp32-0001\r\n"
  740. "CSeq: 2 REGISTER\r\n"
  741. "Contact: <sip:%s@%s:%d>\r\n"
  742. "Authorization: Digest username=\"%s\", realm=\"%s\", nonce=\"%s\", uri=\"sip:%s\", response=\"%s\"\r\n"
  743. "Expires: 300\r\n"
  744. "Content-Length: 0\r\n"
  745. "\r\n",
  746. ASTERISK_IP,
  747. SIP_LOCAL_IP,
  748. SIP_LOCAL_PORT,
  749. SIP_USER,
  750. ASTERISK_IP,
  751. SIP_USER,
  752. ASTERISK_IP,
  753. SIP_USER,
  754. SIP_LOCAL_IP,
  755. SIP_LOCAL_PORT,
  756. SIP_USER,
  757. sip_realm,
  758. sip_nonce,
  759. ASTERISK_IP,
  760. response_hash);
  761. ESP_LOGI(TAG, "Sending AUTH REGISTER...");
  762. sendto(sock, auth, strlen(auth), 0,
  763. (struct sockaddr *)&server, sizeof(server));
  764. }
  765. else if (strstr(response, "SIP/2.0 200")) {
  766. registered = true;
  767. }
  768. else if (strstr(response, "SIP/2.0 486")) {
  769. ESP_LOGW(TAG, "Server returned 486 Busy");
  770. }
  771. else {
  772. ESP_LOGI(TAG, "Ignoring SIP response");
  773. }
  774. }
  775. if (!registered) {
  776. ESP_LOGW(TAG, "Registration was not completed");
  777. close(sock);
  778. vTaskDelete(NULL);
  779. return;
  780. }
  781. ESP_LOGI(TAG, "🎉 SIP REGISTER SUCCESS");
  782. ESP_LOGI(TAG, "stack remaining: %u",
  783. uxTaskGetStackHighWaterMark(NULL));
  784. /* =====================================================
  785. * KEEP SOCKET OPEN AND WAIT FOR SIP REQUESTS
  786. * ===================================================== */
  787. ESP_LOGI(TAG, "Waiting for incoming SIP requests...");
  788. while (1) {
  789. char rx_buf[2048];
  790. int rx_len = recvfrom(sock,
  791. rx_buf,
  792. sizeof(rx_buf) - 1,
  793. 0,
  794. (struct sockaddr *)&server,
  795. &addr_len);
  796. if (rx_len <= 0) {
  797. continue;
  798. }
  799. rx_buf[rx_len] = '\0';
  800. ESP_LOGI(TAG, "\n===== RAW INVITE =====\n%s\n=======================\n", rx_buf);
  801. // printf("\n========================================\n");
  802. // printf("SIP MESSAGE RECEIVED:\n");
  803. // printf("%s\n", rx_buf);
  804. ESP_LOGI(TAG, "Received SIP packet (%d bytes)", rx_len);
  805. // printf("========================================\n");
  806. if (strncmp(rx_buf, "REGISTER ", 9) == 0) {
  807. ESP_LOGI(TAG, "Received REGISTER; sending 200 OK");
  808. sip_send_200_ok(sock, &server, rx_buf, NULL, 0);
  809. continue;
  810. }
  811. else if (strncmp(rx_buf, "INVITE ", 7) == 0) {
  812. ESP_LOGI(TAG, "📞 Incoming INVITE");
  813. char *test = strstr(rx_buf, "m=audio");
  814. if(test)
  815. {
  816. ESP_LOGI(TAG, "Found m=audio SDP:");
  817. ESP_LOGI(TAG, "%s", test);
  818. }
  819. else
  820. {
  821. ESP_LOGE(TAG, "No m=audio found");
  822. }
  823. /* ================= EXTRACT SIP HEADERS ================= */
  824. char via[512] = {0};
  825. char from[256] = {0};
  826. char to[256] = {0};
  827. char callid[256] = {0};
  828. char cseq[128] = {0};
  829. char *header = NULL;
  830. header = strstr(rx_buf, "Via:");
  831. if (header) {
  832. sscanf(header, "Via: %511[^\r\n]", via);
  833. }
  834. header = strstr(rx_buf, "From:");
  835. if (header) {
  836. sscanf(header, "From: %255[^\r\n]", from);
  837. }
  838. header = strstr(rx_buf, "To:");
  839. if (header) {
  840. sscanf(header, "To: %255[^\r\n]", to);
  841. }
  842. header = strstr(rx_buf, "Call-ID:");
  843. if (header) {
  844. sscanf(header, "Call-ID: %255[^\r\n]", callid);
  845. }
  846. header = strstr(rx_buf, "CSeq:");
  847. if (header) {
  848. sscanf(header, "CSeq: %127[^\r\n]", cseq);
  849. }
  850. /* ================= SEND 100 TRYING ================= */
  851. char trying[4096];
  852. snprintf(trying, sizeof(trying),
  853. "SIP/2.0 100 Trying\r\n"
  854. "Via: %s\r\n"
  855. "From: %s\r\n"
  856. "To: %s\r\n"
  857. "Call-ID: %s\r\n"
  858. "CSeq: %s\r\n"
  859. "Content-Length: 0\r\n"
  860. "\r\n",
  861. via,
  862. from,
  863. to,
  864. callid,
  865. cseq);
  866. sendto(sock,
  867. trying,
  868. strlen(trying),
  869. 0,
  870. (struct sockaddr *)&server,
  871. sizeof(server));
  872. ESP_LOGI(TAG, "100 Trying sent");
  873. ESP_LOGI(TAG, "stack remaining: %u",
  874. uxTaskGetStackHighWaterMark(NULL));
  875. char ok[4096];
  876. char sdp[1024];
  877. sip_build_sdp(sdp, sizeof(sdp));
  878. if (sip_extract_rtp_port(rx_buf, &s_rtp_peer_port)) {
  879. s_rtp_peer.sin_family = AF_INET;
  880. s_rtp_peer.sin_port = htons(s_rtp_peer_port);
  881. s_rtp_peer.sin_addr.s_addr = inet_addr(ASTERISK_IP);
  882. ESP_LOGI(TAG,
  883. "RTP peer port=%d",
  884. s_rtp_peer_port);
  885. }
  886. snprintf(ok, sizeof(ok),
  887. "SIP/2.0 200 OK\r\n"
  888. "Via: %s\r\n"
  889. "From: %s\r\n"
  890. "To: %s;esp32\r\n"
  891. "Call-ID: %s\r\n"
  892. "CSeq: %s\r\n"
  893. "Contact: <sip:%s@%s:%d>\r\n"
  894. "Content-Type: application/sdp\r\n"
  895. "Content-Length: %d\r\n"
  896. "\r\n"
  897. "%s",
  898. via,
  899. from,
  900. to,
  901. callid,
  902. cseq,
  903. SIP_USER,
  904. ESP32_IP,
  905. SIP_LOCAL_PORT,
  906. strlen(sdp),
  907. sdp);
  908. ESP_LOGI(TAG, "\n========== 200 OK ==========\n%s\n============================\n", ok);
  909. sendto(sock,
  910. ok,
  911. strlen(ok),
  912. 0,
  913. (struct sockaddr *)&server,
  914. sizeof(server));
  915. ESP_LOGI(TAG, "200 OK sent");
  916. }
  917. else if (strncmp(rx_buf, "OPTIONS ", 8) == 0) {
  918. ESP_LOGI(TAG, "Received OPTIONS");
  919. }
  920. else if (strncmp(rx_buf, "ACK ", 4) == 0)
  921. {
  922. ESP_LOGI(TAG, "Received ACK");
  923. uint8_t test_values[] = {0xFF, 0x7F, 0x00, 0x80};
  924. for(int i = 0; i < 4; i++)
  925. {
  926. ESP_LOGI(TAG,
  927. "ulaw %02X -> %d",
  928. test_values[i],
  929. ulaw2linear(test_values[i]));
  930. }
  931. if (!call_active)
  932. {
  933. call_active = true;
  934. BaseType_t ret = xTaskCreate(
  935. rtp_tx_task,
  936. "rtp_tx",
  937. 16384,
  938. NULL,
  939. 5,
  940. NULL);
  941. xTaskCreate(
  942. rtp_rx_task,
  943. "rtp_rx",
  944. 16384,
  945. NULL,
  946. 5,
  947. NULL);
  948. //temp comment out
  949. //ESP_LOGI(TAG, "rtp task create result=%d", ret);
  950. }
  951. }
  952. else if (strncmp(rx_buf, "BYE ", 4) == 0)
  953. {
  954. ESP_LOGI(TAG, "Received BYE");
  955. call_active = false;
  956. sip_send_200_ok(sock, &server, rx_buf, NULL, 0);
  957. }
  958. else {
  959. ESP_LOGI(TAG, "Unknown SIP message");
  960. }
  961. }
  962. close(sock);
  963. vTaskDelete(NULL);
  964. }