James Kuszmaul | 82f6c04 | 2021-01-17 11:30:16 -0800 | [diff] [blame^] | 1 | /** |
| 2 | * @file http/client.c HTTP Client |
| 3 | * |
| 4 | * Copyright (C) 2011 Creytiv.com |
| 5 | */ |
| 6 | |
| 7 | #include <string.h> |
| 8 | #include <re_types.h> |
| 9 | #include <re_mem.h> |
| 10 | #include <re_mbuf.h> |
| 11 | #include <re_sa.h> |
| 12 | #include <re_list.h> |
| 13 | #include <re_hash.h> |
| 14 | #include <re_fmt.h> |
| 15 | #include <re_tmr.h> |
| 16 | #include <re_srtp.h> |
| 17 | #include <re_tcp.h> |
| 18 | #include <re_tls.h> |
| 19 | #include <re_dns.h> |
| 20 | #include <re_msg.h> |
| 21 | #include <re_http.h> |
| 22 | #include "http.h" |
| 23 | |
| 24 | |
| 25 | enum { |
| 26 | CONN_TIMEOUT = 30000, |
| 27 | RECV_TIMEOUT = 60000, |
| 28 | IDLE_TIMEOUT = 900000, |
| 29 | BUFSIZE_MAX = 524288, |
| 30 | CONN_BSIZE = 256, |
| 31 | }; |
| 32 | |
| 33 | struct http_cli { |
| 34 | struct list reql; |
| 35 | struct hash *ht_conn; |
| 36 | struct dnsc *dnsc; |
| 37 | struct tls *tls; |
| 38 | }; |
| 39 | |
| 40 | struct conn; |
| 41 | |
| 42 | struct http_req { |
| 43 | struct http_chunk chunk; |
| 44 | struct sa srvv[16]; |
| 45 | struct le le; |
| 46 | struct http_req **reqp; |
| 47 | struct http_cli *cli; |
| 48 | struct http_msg *msg; |
| 49 | struct dns_query *dq; |
| 50 | struct conn *conn; |
| 51 | struct mbuf *mbreq; |
| 52 | struct mbuf *mb; |
| 53 | char *host; |
| 54 | http_resp_h *resph; |
| 55 | http_data_h *datah; |
| 56 | http_conn_h *connh; |
| 57 | void *arg; |
| 58 | size_t rx_len; |
| 59 | unsigned srvc; |
| 60 | uint16_t port; |
| 61 | bool chunked; |
| 62 | bool secure; |
| 63 | bool close; |
| 64 | }; |
| 65 | |
| 66 | |
| 67 | struct conn { |
| 68 | struct tmr tmr; |
| 69 | struct sa addr; |
| 70 | struct le he; |
| 71 | struct http_req *req; |
| 72 | struct tls_conn *sc; |
| 73 | struct tcp_conn *tc; |
| 74 | uint64_t usec; |
| 75 | }; |
| 76 | |
| 77 | |
| 78 | static void req_close(struct http_req *req, int err, |
| 79 | const struct http_msg *msg); |
| 80 | static int req_connect(struct http_req *req); |
| 81 | static void timeout_handler(void *arg); |
| 82 | |
| 83 | |
| 84 | static void cli_destructor(void *arg) |
| 85 | { |
| 86 | struct http_cli *cli = arg; |
| 87 | struct le *le = cli->reql.head; |
| 88 | |
| 89 | while (le) { |
| 90 | struct http_req *req = le->data; |
| 91 | |
| 92 | le = le->next; |
| 93 | req_close(req, ECONNABORTED, NULL); |
| 94 | } |
| 95 | |
| 96 | hash_flush(cli->ht_conn); |
| 97 | mem_deref(cli->ht_conn); |
| 98 | mem_deref(cli->dnsc); |
| 99 | mem_deref(cli->tls); |
| 100 | } |
| 101 | |
| 102 | |
| 103 | static void req_destructor(void *arg) |
| 104 | { |
| 105 | struct http_req *req = arg; |
| 106 | |
| 107 | list_unlink(&req->le); |
| 108 | mem_deref(req->msg); |
| 109 | mem_deref(req->dq); |
| 110 | mem_deref(req->conn); |
| 111 | mem_deref(req->mbreq); |
| 112 | mem_deref(req->mb); |
| 113 | mem_deref(req->host); |
| 114 | } |
| 115 | |
| 116 | |
| 117 | static void conn_destructor(void *arg) |
| 118 | { |
| 119 | struct conn *conn = arg; |
| 120 | |
| 121 | tmr_cancel(&conn->tmr); |
| 122 | hash_unlink(&conn->he); |
| 123 | mem_deref(conn->sc); |
| 124 | mem_deref(conn->tc); |
| 125 | } |
| 126 | |
| 127 | |
| 128 | static void conn_idle(struct conn *conn) |
| 129 | { |
| 130 | tmr_start(&conn->tmr, IDLE_TIMEOUT, timeout_handler, conn); |
| 131 | conn->req = NULL; |
| 132 | } |
| 133 | |
| 134 | |
| 135 | static void req_close(struct http_req *req, int err, |
| 136 | const struct http_msg *msg) |
| 137 | { |
| 138 | list_unlink(&req->le); |
| 139 | req->dq = mem_deref(req->dq); |
| 140 | req->datah = NULL; |
| 141 | |
| 142 | if (req->conn) { |
| 143 | if (req->connh) |
| 144 | req->connh(req->conn->tc, req->conn->sc, req->arg); |
| 145 | |
| 146 | if (err || req->close || req->connh) |
| 147 | mem_deref(req->conn); |
| 148 | else |
| 149 | conn_idle(req->conn); |
| 150 | |
| 151 | req->conn = NULL; |
| 152 | } |
| 153 | |
| 154 | req->connh = NULL; |
| 155 | |
| 156 | if (req->reqp) { |
| 157 | *req->reqp = NULL; |
| 158 | req->reqp = NULL; |
| 159 | } |
| 160 | |
| 161 | if (req->resph) { |
| 162 | if (msg) |
| 163 | msg->mb->pos = 0; |
| 164 | |
| 165 | req->resph(err, msg, req->arg); |
| 166 | req->resph = NULL; |
| 167 | } |
| 168 | |
| 169 | mem_deref(req); |
| 170 | } |
| 171 | |
| 172 | |
| 173 | static void try_next(struct conn *conn, int err) |
| 174 | { |
| 175 | struct http_req *req = conn->req; |
| 176 | bool retry = conn->usec > 1; |
| 177 | |
| 178 | mem_deref(conn); |
| 179 | |
| 180 | if (!req) |
| 181 | return; |
| 182 | |
| 183 | req->conn = NULL; |
| 184 | |
| 185 | if (retry) |
| 186 | ++req->srvc; |
| 187 | |
| 188 | if (req->srvc > 0 && !req->msg) { |
| 189 | |
| 190 | err = req_connect(req); |
| 191 | if (!err) |
| 192 | return; |
| 193 | } |
| 194 | |
| 195 | req_close(req, err, NULL); |
| 196 | } |
| 197 | |
| 198 | |
| 199 | static int write_body_buf(struct http_msg *msg, const uint8_t *buf, size_t sz) |
| 200 | { |
| 201 | if ((msg->mb->pos + sz) > BUFSIZE_MAX) |
| 202 | return EOVERFLOW; |
| 203 | |
| 204 | return mbuf_write_mem(msg->mb, buf, sz); |
| 205 | } |
| 206 | |
| 207 | |
| 208 | static int write_body(struct http_req *req, struct mbuf *mb) |
| 209 | { |
| 210 | const size_t size = min(mbuf_get_left(mb), req->rx_len); |
| 211 | int err; |
| 212 | |
| 213 | if (size == 0) |
| 214 | return 0; |
| 215 | |
| 216 | if (req->datah) |
| 217 | err = req->datah(mbuf_buf(mb), size, req->msg, req->arg); |
| 218 | else |
| 219 | err = write_body_buf(req->msg, mbuf_buf(mb), size); |
| 220 | |
| 221 | if (err) |
| 222 | return err; |
| 223 | |
| 224 | req->rx_len -= size; |
| 225 | mb->pos += size; |
| 226 | |
| 227 | return 0; |
| 228 | } |
| 229 | |
| 230 | |
| 231 | static int req_recv(struct http_req *req, struct mbuf *mb, bool *last) |
| 232 | { |
| 233 | int err; |
| 234 | |
| 235 | *last = false; |
| 236 | |
| 237 | if (!req->chunked) { |
| 238 | |
| 239 | err = write_body(req, mb); |
| 240 | if (err) |
| 241 | return err; |
| 242 | |
| 243 | if (req->rx_len == 0) |
| 244 | *last = true; |
| 245 | |
| 246 | return 0; |
| 247 | } |
| 248 | |
| 249 | while (mbuf_get_left(mb)) { |
| 250 | |
| 251 | if (req->rx_len == 0) { |
| 252 | |
| 253 | err = http_chunk_decode(&req->chunk, mb, &req->rx_len); |
| 254 | if (err == ENODATA) |
| 255 | return 0; |
| 256 | else if (err) |
| 257 | return err; |
| 258 | else if (req->rx_len == 0) { |
| 259 | *last = true; |
| 260 | return 0; |
| 261 | } |
| 262 | } |
| 263 | |
| 264 | err = write_body(req, mb); |
| 265 | if (err) |
| 266 | return err; |
| 267 | } |
| 268 | |
| 269 | return 0; |
| 270 | } |
| 271 | |
| 272 | |
| 273 | static void timeout_handler(void *arg) |
| 274 | { |
| 275 | struct conn *conn = arg; |
| 276 | |
| 277 | try_next(conn, ETIMEDOUT); |
| 278 | } |
| 279 | |
| 280 | |
| 281 | static void estab_handler(void *arg) |
| 282 | { |
| 283 | struct conn *conn = arg; |
| 284 | struct http_req *req = conn->req; |
| 285 | int err; |
| 286 | |
| 287 | if (!req) |
| 288 | return; |
| 289 | |
| 290 | err = tcp_send(conn->tc, req->mbreq); |
| 291 | if (err) { |
| 292 | try_next(conn, err); |
| 293 | return; |
| 294 | } |
| 295 | |
| 296 | tmr_start(&conn->tmr, RECV_TIMEOUT, timeout_handler, conn); |
| 297 | } |
| 298 | |
| 299 | |
| 300 | static void recv_handler(struct mbuf *mb, void *arg) |
| 301 | { |
| 302 | const struct http_hdr *hdr; |
| 303 | struct conn *conn = arg; |
| 304 | struct http_req *req = conn->req; |
| 305 | size_t pos; |
| 306 | bool last; |
| 307 | int err; |
| 308 | |
| 309 | if (!req) |
| 310 | return; |
| 311 | |
| 312 | if (req->msg) { |
| 313 | err = req_recv(req, mb, &last); |
| 314 | if (err || last) |
| 315 | goto out; |
| 316 | |
| 317 | return; |
| 318 | } |
| 319 | |
| 320 | if (req->mb) { |
| 321 | |
| 322 | const size_t len = mbuf_get_left(mb); |
| 323 | |
| 324 | if ((mbuf_get_left(req->mb) + len) > BUFSIZE_MAX) { |
| 325 | err = EOVERFLOW; |
| 326 | goto out; |
| 327 | } |
| 328 | |
| 329 | pos = req->mb->pos; |
| 330 | req->mb->pos = req->mb->end; |
| 331 | |
| 332 | err = mbuf_write_mem(req->mb, mbuf_buf(mb), len); |
| 333 | if (err) |
| 334 | goto out; |
| 335 | |
| 336 | req->mb->pos = pos; |
| 337 | } |
| 338 | else { |
| 339 | req->mb = mem_ref(mb); |
| 340 | } |
| 341 | |
| 342 | pos = req->mb->pos; |
| 343 | |
| 344 | err = http_msg_decode(&req->msg, req->mb, false); |
| 345 | if (err) { |
| 346 | if (err == ENODATA) { |
| 347 | req->mb->pos = pos; |
| 348 | return; |
| 349 | } |
| 350 | goto out; |
| 351 | } |
| 352 | |
| 353 | if (req->datah) |
| 354 | tmr_cancel(&conn->tmr); |
| 355 | |
| 356 | hdr = http_msg_hdr(req->msg, HTTP_HDR_CONNECTION); |
| 357 | if (hdr && !pl_strcasecmp(&hdr->val, "close")) |
| 358 | req->close = true; |
| 359 | |
| 360 | if (http_msg_hdr_has_value(req->msg, HTTP_HDR_TRANSFER_ENCODING, |
| 361 | "chunked")) |
| 362 | req->chunked = true; |
| 363 | else |
| 364 | req->rx_len = req->msg->clen; |
| 365 | |
| 366 | err = req_recv(req, req->mb, &last); |
| 367 | if (err || last) |
| 368 | goto out; |
| 369 | |
| 370 | return; |
| 371 | |
| 372 | out: |
| 373 | req_close(req, err, req->msg); |
| 374 | } |
| 375 | |
| 376 | |
| 377 | static void close_handler(int err, void *arg) |
| 378 | { |
| 379 | struct conn *conn = arg; |
| 380 | |
| 381 | try_next(conn, err ? err : ECONNRESET); |
| 382 | } |
| 383 | |
| 384 | |
| 385 | static bool conn_cmp(struct le *le, void *arg) |
| 386 | { |
| 387 | const struct conn *conn = le->data; |
| 388 | const struct http_req *req = arg; |
| 389 | |
| 390 | if (!sa_cmp(&req->srvv[req->srvc], &conn->addr, SA_ALL)) |
| 391 | return false; |
| 392 | |
| 393 | if (req->secure != !!conn->sc) |
| 394 | return false; |
| 395 | |
| 396 | return conn->req == NULL; |
| 397 | } |
| 398 | |
| 399 | |
| 400 | static int conn_connect(struct http_req *req) |
| 401 | { |
| 402 | const struct sa *addr = &req->srvv[req->srvc]; |
| 403 | struct conn *conn; |
| 404 | int err; |
| 405 | |
| 406 | conn = list_ledata(hash_lookup(req->cli->ht_conn, |
| 407 | sa_hash(addr, SA_ALL), conn_cmp, req)); |
| 408 | if (conn) { |
| 409 | err = tcp_send(conn->tc, req->mbreq); |
| 410 | if (!err) { |
| 411 | tmr_start(&conn->tmr, RECV_TIMEOUT, |
| 412 | timeout_handler, conn); |
| 413 | |
| 414 | req->conn = conn; |
| 415 | conn->req = req; |
| 416 | |
| 417 | ++conn->usec; |
| 418 | |
| 419 | return 0; |
| 420 | } |
| 421 | |
| 422 | mem_deref(conn); |
| 423 | } |
| 424 | |
| 425 | conn = mem_zalloc(sizeof(*conn), conn_destructor); |
| 426 | if (!conn) |
| 427 | return ENOMEM; |
| 428 | |
| 429 | hash_append(req->cli->ht_conn, sa_hash(addr, SA_ALL), &conn->he, conn); |
| 430 | |
| 431 | conn->addr = *addr; |
| 432 | conn->usec = 1; |
| 433 | |
| 434 | err = tcp_connect(&conn->tc, addr, estab_handler, recv_handler, |
| 435 | close_handler, conn); |
| 436 | if (err) |
| 437 | goto out; |
| 438 | |
| 439 | #ifdef USE_TLS |
| 440 | if (req->secure) { |
| 441 | |
| 442 | err = tls_start_tcp(&conn->sc, req->cli->tls, conn->tc, 0); |
| 443 | if (err) |
| 444 | goto out; |
| 445 | } |
| 446 | #endif |
| 447 | |
| 448 | tmr_start(&conn->tmr, CONN_TIMEOUT, timeout_handler, conn); |
| 449 | |
| 450 | req->conn = conn; |
| 451 | conn->req = req; |
| 452 | |
| 453 | out: |
| 454 | if (err) |
| 455 | mem_deref(conn); |
| 456 | |
| 457 | return err; |
| 458 | } |
| 459 | |
| 460 | |
| 461 | static int req_connect(struct http_req *req) |
| 462 | { |
| 463 | int err = EINVAL; |
| 464 | |
| 465 | while (req->srvc > 0) { |
| 466 | |
| 467 | --req->srvc; |
| 468 | |
| 469 | req->mb = mem_deref(req->mb); |
| 470 | |
| 471 | err = conn_connect(req); |
| 472 | if (!err) |
| 473 | break; |
| 474 | } |
| 475 | |
| 476 | return err; |
| 477 | } |
| 478 | |
| 479 | |
| 480 | static bool rr_handler(struct dnsrr *rr, void *arg) |
| 481 | { |
| 482 | struct http_req *req = arg; |
| 483 | |
| 484 | if (req->srvc >= ARRAY_SIZE(req->srvv)) |
| 485 | return true; |
| 486 | |
| 487 | switch (rr->type) { |
| 488 | |
| 489 | case DNS_TYPE_A: |
| 490 | sa_set_in(&req->srvv[req->srvc++], rr->rdata.a.addr, |
| 491 | req->port); |
| 492 | break; |
| 493 | |
| 494 | case DNS_TYPE_AAAA: |
| 495 | sa_set_in6(&req->srvv[req->srvc++], rr->rdata.aaaa.addr, |
| 496 | req->port); |
| 497 | break; |
| 498 | } |
| 499 | |
| 500 | return false; |
| 501 | } |
| 502 | |
| 503 | |
| 504 | static void query_handler(int err, const struct dnshdr *hdr, struct list *ansl, |
| 505 | struct list *authl, struct list *addl, void *arg) |
| 506 | { |
| 507 | struct http_req *req = arg; |
| 508 | (void)hdr; |
| 509 | (void)authl; |
| 510 | (void)addl; |
| 511 | |
| 512 | dns_rrlist_apply2(ansl, req->host, DNS_TYPE_A, DNS_TYPE_AAAA, |
| 513 | DNS_CLASS_IN, true, rr_handler, req); |
| 514 | if (req->srvc == 0) { |
| 515 | err = err ? err : EDESTADDRREQ; |
| 516 | goto fail; |
| 517 | } |
| 518 | |
| 519 | err = req_connect(req); |
| 520 | if (err) |
| 521 | goto fail; |
| 522 | |
| 523 | return; |
| 524 | |
| 525 | fail: |
| 526 | req_close(req, err, NULL); |
| 527 | } |
| 528 | |
| 529 | |
| 530 | /** |
| 531 | * Send an HTTP request |
| 532 | * |
| 533 | * @param reqp Pointer to allocated HTTP request object |
| 534 | * @param cli HTTP Client |
| 535 | * @param met Request method |
| 536 | * @param uri Request URI |
| 537 | * @param resph Response handler |
| 538 | * @param datah Content handler (optional) |
| 539 | * @param arg Handler argument |
| 540 | * @param fmt Formatted HTTP headers and body (optional) |
| 541 | * |
| 542 | * @return 0 if success, otherwise errorcode |
| 543 | */ |
| 544 | int http_request(struct http_req **reqp, struct http_cli *cli, const char *met, |
| 545 | const char *uri, http_resp_h *resph, http_data_h *datah, |
| 546 | void *arg, const char *fmt, ...) |
| 547 | { |
| 548 | struct pl scheme, host, port, path; |
| 549 | struct http_req *req; |
| 550 | uint16_t defport; |
| 551 | bool secure; |
| 552 | va_list ap; |
| 553 | int err; |
| 554 | |
| 555 | if (!cli || !met || !uri) |
| 556 | return EINVAL; |
| 557 | |
| 558 | if (re_regex(uri, strlen(uri), "[a-z]+://[^:/]+[:]*[0-9]*[^]+", |
| 559 | &scheme, &host, NULL, &port, &path) || scheme.p != uri) |
| 560 | return EINVAL; |
| 561 | |
| 562 | if (!pl_strcasecmp(&scheme, "http") || |
| 563 | !pl_strcasecmp(&scheme, "ws")) { |
| 564 | secure = false; |
| 565 | defport = 80; |
| 566 | } |
| 567 | #ifdef USE_TLS |
| 568 | else if (!pl_strcasecmp(&scheme, "https") || |
| 569 | !pl_strcasecmp(&scheme, "wss")) { |
| 570 | secure = true; |
| 571 | defport = 443; |
| 572 | } |
| 573 | #endif |
| 574 | else |
| 575 | return ENOTSUP; |
| 576 | |
| 577 | req = mem_zalloc(sizeof(*req), req_destructor); |
| 578 | if (!req) |
| 579 | return ENOMEM; |
| 580 | |
| 581 | list_append(&cli->reql, &req->le, req); |
| 582 | |
| 583 | req->cli = cli; |
| 584 | req->secure = secure; |
| 585 | req->port = pl_isset(&port) ? pl_u32(&port) : defport; |
| 586 | req->resph = resph; |
| 587 | req->datah = datah; |
| 588 | req->arg = arg; |
| 589 | |
| 590 | err = pl_strdup(&req->host, &host); |
| 591 | if (err) |
| 592 | goto out; |
| 593 | |
| 594 | req->mbreq = mbuf_alloc(1024); |
| 595 | if (!req->mbreq) { |
| 596 | err = ENOMEM; |
| 597 | goto out; |
| 598 | } |
| 599 | |
| 600 | err = mbuf_printf(req->mbreq, |
| 601 | "%s %r HTTP/1.1\r\n" |
| 602 | "Host: %r\r\n", |
| 603 | met, &path, &host); |
| 604 | if (fmt) { |
| 605 | va_start(ap, fmt); |
| 606 | err |= mbuf_vprintf(req->mbreq, fmt, ap); |
| 607 | va_end(ap); |
| 608 | } |
| 609 | else { |
| 610 | err |= mbuf_write_str(req->mbreq, "\r\n"); |
| 611 | } |
| 612 | if (err) |
| 613 | goto out; |
| 614 | |
| 615 | req->mbreq->pos = 0; |
| 616 | |
| 617 | if (!sa_set_str(&req->srvv[0], req->host, req->port)) { |
| 618 | |
| 619 | req->srvc = 1; |
| 620 | |
| 621 | err = req_connect(req); |
| 622 | if (err) |
| 623 | goto out; |
| 624 | } |
| 625 | else { |
| 626 | err = dnsc_query(&req->dq, cli->dnsc, req->host, |
| 627 | DNS_TYPE_A, DNS_CLASS_IN, true, |
| 628 | query_handler, req); |
| 629 | if (err) |
| 630 | goto out; |
| 631 | } |
| 632 | |
| 633 | out: |
| 634 | if (err) |
| 635 | mem_deref(req); |
| 636 | else if (reqp) { |
| 637 | req->reqp = reqp; |
| 638 | *reqp = req; |
| 639 | } |
| 640 | |
| 641 | return err; |
| 642 | } |
| 643 | |
| 644 | |
| 645 | /** |
| 646 | * Set HTTP request connection handler |
| 647 | * |
| 648 | * @param req HTTP request object |
| 649 | * @param connh Connection handler |
| 650 | */ |
| 651 | void http_req_set_conn_handler(struct http_req *req, http_conn_h *connh) |
| 652 | { |
| 653 | if (!req) |
| 654 | return; |
| 655 | |
| 656 | req->connh = connh; |
| 657 | } |
| 658 | |
| 659 | |
| 660 | /** |
| 661 | * Allocate an HTTP client instance |
| 662 | * |
| 663 | * @param clip Pointer to allocated HTTP client |
| 664 | * @param dnsc DNS Client |
| 665 | * |
| 666 | * @return 0 if success, otherwise errorcode |
| 667 | */ |
| 668 | int http_client_alloc(struct http_cli **clip, struct dnsc *dnsc) |
| 669 | { |
| 670 | struct http_cli *cli; |
| 671 | int err; |
| 672 | |
| 673 | if (!clip || !dnsc) |
| 674 | return EINVAL; |
| 675 | |
| 676 | cli = mem_zalloc(sizeof(*cli), cli_destructor); |
| 677 | if (!cli) |
| 678 | return ENOMEM; |
| 679 | |
| 680 | err = hash_alloc(&cli->ht_conn, CONN_BSIZE); |
| 681 | if (err) |
| 682 | goto out; |
| 683 | |
| 684 | #ifdef USE_TLS |
| 685 | err = tls_alloc(&cli->tls, TLS_METHOD_SSLV23, NULL, NULL); |
| 686 | #else |
| 687 | err = 0; |
| 688 | #endif |
| 689 | if (err) |
| 690 | goto out; |
| 691 | |
| 692 | cli->dnsc = mem_ref(dnsc); |
| 693 | |
| 694 | out: |
| 695 | if (err) |
| 696 | mem_deref(cli); |
| 697 | else |
| 698 | *clip = cli; |
| 699 | |
| 700 | return err; |
| 701 | } |