1 /*- 2 * Copyright (c) 2000-2004 Dag-Erling Co�dan Sm�rgrav 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer 10 * in this position and unchanged. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. The name of the author may not be used to endorse or promote products 15 * derived from this software without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 27 * 28 * $FreeBSD: src/lib/libfetch/http.c,v 1.78 2007/05/08 19:28:03 des Exp $ 29 * $DragonFly: src/lib/libfetch/http.c,v 1.4 2007/08/05 21:48:12 swildner Exp $ 30 */ 31 32 /* 33 * The following copyright applies to the base64 code: 34 * 35 *- 36 * Copyright 1997 Massachusetts Institute of Technology 37 * 38 * Permission to use, copy, modify, and distribute this software and 39 * its documentation for any purpose and without fee is hereby 40 * granted, provided that both the above copyright notice and this 41 * permission notice appear in all copies, that both the above 42 * copyright notice and this permission notice appear in all 43 * supporting documentation, and that the name of M.I.T. not be used 44 * in advertising or publicity pertaining to distribution of the 45 * software without specific, written prior permission. M.I.T. makes 46 * no representations about the suitability of this software for any 47 * purpose. It is provided "as is" without express or implied 48 * warranty. 49 * 50 * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''. M.I.T. DISCLAIMS 51 * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE, 52 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 53 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT 54 * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 55 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 56 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF 57 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND 58 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 59 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 60 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 61 * SUCH DAMAGE. 62 */ 63 64 #include <sys/param.h> 65 #include <sys/socket.h> 66 67 #include <ctype.h> 68 #include <err.h> 69 #include <errno.h> 70 #include <locale.h> 71 #include <netdb.h> 72 #include <stdarg.h> 73 #include <stdio.h> 74 #include <stdlib.h> 75 #include <string.h> 76 #include <time.h> 77 #include <unistd.h> 78 79 #include <netinet/in.h> 80 #include <netinet/tcp.h> 81 82 #include "fetch.h" 83 #include "common.h" 84 #include "httperr.h" 85 86 /* Maximum number of redirects to follow */ 87 #define MAX_REDIRECT 5 88 89 /* Symbolic names for reply codes we care about */ 90 #define HTTP_OK 200 91 #define HTTP_PARTIAL 206 92 #define HTTP_MOVED_PERM 301 93 #define HTTP_MOVED_TEMP 302 94 #define HTTP_SEE_OTHER 303 95 #define HTTP_TEMP_REDIRECT 307 96 #define HTTP_NEED_AUTH 401 97 #define HTTP_NEED_PROXY_AUTH 407 98 #define HTTP_BAD_RANGE 416 99 #define HTTP_PROTOCOL_ERROR 999 100 101 #define HTTP_REDIRECT(xyz) ((xyz) == HTTP_MOVED_PERM \ 102 || (xyz) == HTTP_MOVED_TEMP \ 103 || (xyz) == HTTP_TEMP_REDIRECT \ 104 || (xyz) == HTTP_SEE_OTHER) 105 106 #define HTTP_ERROR(xyz) ((xyz) > 400 && (xyz) < 599) 107 108 109 /***************************************************************************** 110 * I/O functions for decoding chunked streams 111 */ 112 113 struct httpio 114 { 115 conn_t *conn; /* connection */ 116 int chunked; /* chunked mode */ 117 char *buf; /* chunk buffer */ 118 size_t bufsize; /* size of chunk buffer */ 119 ssize_t buflen; /* amount of data currently in buffer */ 120 int bufpos; /* current read offset in buffer */ 121 int eof; /* end-of-file flag */ 122 int error; /* error flag */ 123 size_t chunksize; /* remaining size of current chunk */ 124 #ifndef NDEBUG 125 size_t total; 126 #endif 127 }; 128 129 /* 130 * Get next chunk header 131 */ 132 static int 133 _http_new_chunk(struct httpio *io) 134 { 135 char *p; 136 137 if (_fetch_getln(io->conn) == -1) 138 return (-1); 139 140 if (io->conn->buflen < 2 || !isxdigit(*io->conn->buf)) 141 return (-1); 142 143 for (p = io->conn->buf; *p && !isspace(*p); ++p) { 144 if (*p == ';') 145 break; 146 if (!isxdigit(*p)) 147 return (-1); 148 if (isdigit(*p)) { 149 io->chunksize = io->chunksize * 16 + 150 *p - '0'; 151 } else { 152 io->chunksize = io->chunksize * 16 + 153 10 + tolower(*p) - 'a'; 154 } 155 } 156 157 #ifndef NDEBUG 158 if (fetchDebug) { 159 io->total += io->chunksize; 160 if (io->chunksize == 0) 161 fprintf(stderr, "%s(): end of last chunk\n", __func__); 162 else 163 fprintf(stderr, "%s(): new chunk: %lu (%lu)\n", 164 __func__, (unsigned long)io->chunksize, 165 (unsigned long)io->total); 166 } 167 #endif 168 169 return (io->chunksize); 170 } 171 172 /* 173 * Grow the input buffer to at least len bytes 174 */ 175 static inline int 176 _http_growbuf(struct httpio *io, size_t len) 177 { 178 char *tmp; 179 180 if (io->bufsize >= len) 181 return (0); 182 183 if ((tmp = realloc(io->buf, len)) == NULL) 184 return (-1); 185 io->buf = tmp; 186 io->bufsize = len; 187 return (0); 188 } 189 190 /* 191 * Fill the input buffer, do chunk decoding on the fly 192 */ 193 static int 194 _http_fillbuf(struct httpio *io, size_t len) 195 { 196 if (io->error) 197 return (-1); 198 if (io->eof) 199 return (0); 200 201 if (io->chunked == 0) { 202 if (_http_growbuf(io, len) == -1) 203 return (-1); 204 if ((io->buflen = _fetch_read(io->conn, io->buf, len)) == -1) { 205 io->error = 1; 206 return (-1); 207 } 208 io->bufpos = 0; 209 return (io->buflen); 210 } 211 212 if (io->chunksize == 0) { 213 switch (_http_new_chunk(io)) { 214 case -1: 215 io->error = 1; 216 return (-1); 217 case 0: 218 io->eof = 1; 219 return (0); 220 } 221 } 222 223 if (len > io->chunksize) 224 len = io->chunksize; 225 if (_http_growbuf(io, len) == -1) 226 return (-1); 227 if ((io->buflen = _fetch_read(io->conn, io->buf, len)) == -1) { 228 io->error = 1; 229 return (-1); 230 } 231 io->chunksize -= io->buflen; 232 233 if (io->chunksize == 0) { 234 char endl[2]; 235 236 if (_fetch_read(io->conn, endl, 2) != 2 || 237 endl[0] != '\r' || endl[1] != '\n') 238 return (-1); 239 } 240 241 io->bufpos = 0; 242 243 return (io->buflen); 244 } 245 246 /* 247 * Read function 248 */ 249 static int 250 _http_readfn(void *v, char *buf, int len) 251 { 252 struct httpio *io = (struct httpio *)v; 253 int l, pos; 254 255 if (io->error) 256 return (-1); 257 if (io->eof) 258 return (0); 259 260 for (pos = 0; len > 0; pos += l, len -= l) { 261 /* empty buffer */ 262 if (!io->buf || io->bufpos == io->buflen) 263 if (_http_fillbuf(io, len) < 1) 264 break; 265 l = io->buflen - io->bufpos; 266 if (len < l) 267 l = len; 268 bcopy(io->buf + io->bufpos, buf + pos, l); 269 io->bufpos += l; 270 } 271 272 if (!pos && io->error) 273 return (-1); 274 return (pos); 275 } 276 277 /* 278 * Write function 279 */ 280 static int 281 _http_writefn(void *v, const char *buf, int len) 282 { 283 struct httpio *io = (struct httpio *)v; 284 285 return (_fetch_write(io->conn, buf, len)); 286 } 287 288 /* 289 * Close function 290 */ 291 static int 292 _http_closefn(void *v) 293 { 294 struct httpio *io = (struct httpio *)v; 295 int r; 296 297 r = _fetch_close(io->conn); 298 if (io->buf) 299 free(io->buf); 300 free(io); 301 return (r); 302 } 303 304 /* 305 * Wrap a file descriptor up 306 */ 307 static FILE * 308 _http_funopen(conn_t *conn, int chunked) 309 { 310 struct httpio *io; 311 FILE *f; 312 313 if ((io = calloc(1, sizeof(*io))) == NULL) { 314 _fetch_syserr(); 315 return (NULL); 316 } 317 io->conn = conn; 318 io->chunked = chunked; 319 f = funopen(io, _http_readfn, _http_writefn, NULL, _http_closefn); 320 if (f == NULL) { 321 _fetch_syserr(); 322 free(io); 323 return (NULL); 324 } 325 return (f); 326 } 327 328 329 /***************************************************************************** 330 * Helper functions for talking to the server and parsing its replies 331 */ 332 333 /* Header types */ 334 typedef enum { 335 hdr_syserror = -2, 336 hdr_error = -1, 337 hdr_end = 0, 338 hdr_unknown = 1, 339 hdr_content_length, 340 hdr_content_range, 341 hdr_last_modified, 342 hdr_location, 343 hdr_transfer_encoding, 344 hdr_www_authenticate 345 } hdr_t; 346 347 /* Names of interesting headers */ 348 static struct { 349 hdr_t num; 350 const char *name; 351 } hdr_names[] = { 352 { hdr_content_length, "Content-Length" }, 353 { hdr_content_range, "Content-Range" }, 354 { hdr_last_modified, "Last-Modified" }, 355 { hdr_location, "Location" }, 356 { hdr_transfer_encoding, "Transfer-Encoding" }, 357 { hdr_www_authenticate, "WWW-Authenticate" }, 358 { hdr_unknown, NULL }, 359 }; 360 361 /* 362 * Send a formatted line; optionally echo to terminal 363 */ 364 static int 365 _http_cmd(conn_t *conn, const char *fmt, ...) 366 { 367 va_list ap; 368 size_t len; 369 char *msg; 370 int r; 371 372 va_start(ap, fmt); 373 len = vasprintf(&msg, fmt, ap); 374 va_end(ap); 375 376 if (msg == NULL) { 377 errno = ENOMEM; 378 _fetch_syserr(); 379 return (-1); 380 } 381 382 r = _fetch_putln(conn, msg, len); 383 free(msg); 384 385 if (r == -1) { 386 _fetch_syserr(); 387 return (-1); 388 } 389 390 return (0); 391 } 392 393 /* 394 * Get and parse status line 395 */ 396 static int 397 _http_get_reply(conn_t *conn) 398 { 399 char *p; 400 401 if (_fetch_getln(conn) == -1) 402 return (-1); 403 /* 404 * A valid status line looks like "HTTP/m.n xyz reason" where m 405 * and n are the major and minor protocol version numbers and xyz 406 * is the reply code. 407 * Unfortunately, there are servers out there (NCSA 1.5.1, to name 408 * just one) that do not send a version number, so we can't rely 409 * on finding one, but if we do, insist on it being 1.0 or 1.1. 410 * We don't care about the reason phrase. 411 */ 412 if (strncmp(conn->buf, "HTTP", 4) != 0) 413 return (HTTP_PROTOCOL_ERROR); 414 p = conn->buf + 4; 415 if (*p == '/') { 416 if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1')) 417 return (HTTP_PROTOCOL_ERROR); 418 p += 4; 419 } 420 if (*p != ' ' || !isdigit(p[1]) || !isdigit(p[2]) || !isdigit(p[3])) 421 return (HTTP_PROTOCOL_ERROR); 422 423 conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0'); 424 return (conn->err); 425 } 426 427 /* 428 * Check a header; if the type matches the given string, return a pointer 429 * to the beginning of the value. 430 */ 431 static const char * 432 _http_match(const char *str, const char *hdr) 433 { 434 while (*str && *hdr && tolower(*str++) == tolower(*hdr++)) 435 /* nothing */; 436 if (*str || *hdr != ':') 437 return (NULL); 438 while (*hdr && isspace(*++hdr)) 439 /* nothing */; 440 return (hdr); 441 } 442 443 /* 444 * Get the next header and return the appropriate symbolic code. 445 */ 446 static hdr_t 447 _http_next_header(conn_t *conn, const char **p) 448 { 449 int i; 450 451 if (_fetch_getln(conn) == -1) 452 return (hdr_syserror); 453 while (conn->buflen && isspace(conn->buf[conn->buflen - 1])) 454 conn->buflen--; 455 conn->buf[conn->buflen] = '\0'; 456 if (conn->buflen == 0) 457 return (hdr_end); 458 /* 459 * We could check for malformed headers but we don't really care. 460 * A valid header starts with a token immediately followed by a 461 * colon; a token is any sequence of non-control, non-whitespace 462 * characters except "()<>@,;:\\\"{}". 463 */ 464 for (i = 0; hdr_names[i].num != hdr_unknown; i++) 465 if ((*p = _http_match(hdr_names[i].name, conn->buf)) != NULL) 466 return (hdr_names[i].num); 467 return (hdr_unknown); 468 } 469 470 /* 471 * Parse a last-modified header 472 */ 473 static int 474 _http_parse_mtime(const char *p, time_t *mtime) 475 { 476 char locale[64], *r; 477 struct tm tm; 478 479 strncpy(locale, setlocale(LC_TIME, NULL), sizeof(locale)); 480 setlocale(LC_TIME, "C"); 481 r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm); 482 /* XXX should add support for date-2 and date-3 */ 483 setlocale(LC_TIME, locale); 484 if (r == NULL) 485 return (-1); 486 DEBUG(fprintf(stderr, "last modified: [%04d-%02d-%02d " 487 "%02d:%02d:%02d]\n", 488 tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday, 489 tm.tm_hour, tm.tm_min, tm.tm_sec)); 490 *mtime = timegm(&tm); 491 return (0); 492 } 493 494 /* 495 * Parse a content-length header 496 */ 497 static int 498 _http_parse_length(const char *p, off_t *length) 499 { 500 off_t len; 501 502 for (len = 0; *p && isdigit(*p); ++p) 503 len = len * 10 + (*p - '0'); 504 if (*p) 505 return (-1); 506 DEBUG(fprintf(stderr, "content length: [%lld]\n", 507 (long long)len)); 508 *length = len; 509 return (0); 510 } 511 512 /* 513 * Parse a content-range header 514 */ 515 static int 516 _http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size) 517 { 518 off_t first, last, len; 519 520 if (strncasecmp(p, "bytes ", 6) != 0) 521 return (-1); 522 p += 6; 523 if (*p == '*') { 524 first = last = -1; 525 ++p; 526 } else { 527 for (first = 0; *p && isdigit(*p); ++p) 528 first = first * 10 + *p - '0'; 529 if (*p != '-') 530 return (-1); 531 for (last = 0, ++p; *p && isdigit(*p); ++p) 532 last = last * 10 + *p - '0'; 533 } 534 if (first > last || *p != '/') 535 return (-1); 536 for (len = 0, ++p; *p && isdigit(*p); ++p) 537 len = len * 10 + *p - '0'; 538 if (*p || len < last - first + 1) 539 return (-1); 540 if (first == -1) { 541 DEBUG(fprintf(stderr, "content range: [*/%lld]\n", 542 (long long)len)); 543 *length = 0; 544 } else { 545 DEBUG(fprintf(stderr, "content range: [%lld-%lld/%lld]\n", 546 (long long)first, (long long)last, (long long)len)); 547 *length = last - first + 1; 548 } 549 *offset = first; 550 *size = len; 551 return (0); 552 } 553 554 555 /***************************************************************************** 556 * Helper functions for authorization 557 */ 558 559 /* 560 * Base64 encoding 561 */ 562 static char * 563 _http_base64(const char *src) 564 { 565 static const char base64[] = 566 "ABCDEFGHIJKLMNOPQRSTUVWXYZ" 567 "abcdefghijklmnopqrstuvwxyz" 568 "0123456789+/"; 569 char *str, *dst; 570 size_t l; 571 int t, r; 572 573 l = strlen(src); 574 if ((str = malloc(((l + 2) / 3) * 4 + 1)) == NULL) 575 return (NULL); 576 dst = str; 577 r = 0; 578 579 while (l >= 3) { 580 t = (src[0] << 16) | (src[1] << 8) | src[2]; 581 dst[0] = base64[(t >> 18) & 0x3f]; 582 dst[1] = base64[(t >> 12) & 0x3f]; 583 dst[2] = base64[(t >> 6) & 0x3f]; 584 dst[3] = base64[(t >> 0) & 0x3f]; 585 src += 3; l -= 3; 586 dst += 4; r += 4; 587 } 588 589 switch (l) { 590 case 2: 591 t = (src[0] << 16) | (src[1] << 8); 592 dst[0] = base64[(t >> 18) & 0x3f]; 593 dst[1] = base64[(t >> 12) & 0x3f]; 594 dst[2] = base64[(t >> 6) & 0x3f]; 595 dst[3] = '='; 596 dst += 4; 597 r += 4; 598 break; 599 case 1: 600 t = src[0] << 16; 601 dst[0] = base64[(t >> 18) & 0x3f]; 602 dst[1] = base64[(t >> 12) & 0x3f]; 603 dst[2] = dst[3] = '='; 604 dst += 4; 605 r += 4; 606 break; 607 case 0: 608 break; 609 } 610 611 *dst = 0; 612 return (str); 613 } 614 615 /* 616 * Encode username and password 617 */ 618 static int 619 _http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd) 620 { 621 char *upw, *auth; 622 int r; 623 624 DEBUG(fprintf(stderr, "usr: [%s]\n", usr)); 625 DEBUG(fprintf(stderr, "pwd: [%s]\n", pwd)); 626 if (asprintf(&upw, "%s:%s", usr, pwd) == -1) 627 return (-1); 628 auth = _http_base64(upw); 629 free(upw); 630 if (auth == NULL) 631 return (-1); 632 r = _http_cmd(conn, "%s: Basic %s", hdr, auth); 633 free(auth); 634 return (r); 635 } 636 637 /* 638 * Send an authorization header 639 */ 640 static int 641 _http_authorize(conn_t *conn, const char *hdr, const char *p) 642 { 643 /* basic authorization */ 644 if (strncasecmp(p, "basic:", 6) == 0) { 645 char *user, *pwd, *str; 646 int r; 647 648 /* skip realm */ 649 for (p += 6; *p && *p != ':'; ++p) 650 /* nothing */ ; 651 if (!*p || strchr(++p, ':') == NULL) 652 return (-1); 653 if ((str = strdup(p)) == NULL) 654 return (-1); /* XXX */ 655 user = str; 656 pwd = strchr(str, ':'); 657 *pwd++ = '\0'; 658 r = _http_basic_auth(conn, hdr, user, pwd); 659 free(str); 660 return (r); 661 } 662 return (-1); 663 } 664 665 666 /***************************************************************************** 667 * Helper functions for connecting to a server or proxy 668 */ 669 670 /* 671 * Connect to the correct HTTP server or proxy. 672 */ 673 static conn_t * 674 _http_connect(struct url *URL, struct url *purl, const char *flags) 675 { 676 conn_t *conn; 677 int verbose; 678 int af, val; 679 680 #ifdef INET6 681 af = AF_UNSPEC; 682 #else 683 af = AF_INET; 684 #endif 685 686 verbose = CHECK_FLAG('v'); 687 if (CHECK_FLAG('4')) 688 af = AF_INET; 689 #ifdef INET6 690 else if (CHECK_FLAG('6')) 691 af = AF_INET6; 692 #endif 693 694 if (purl && strcasecmp(URL->scheme, SCHEME_HTTPS) != 0) { 695 URL = purl; 696 } else if (strcasecmp(URL->scheme, SCHEME_FTP) == 0) { 697 /* can't talk http to an ftp server */ 698 /* XXX should set an error code */ 699 return (NULL); 700 } 701 702 if ((conn = _fetch_connect(URL->host, URL->port, af, verbose)) == NULL) 703 /* _fetch_connect() has already set an error code */ 704 return (NULL); 705 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 && 706 _fetch_ssl(conn, verbose) == -1) { 707 _fetch_close(conn); 708 /* grrr */ 709 errno = EAUTH; 710 _fetch_syserr(); 711 return (NULL); 712 } 713 714 val = 1; 715 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val)); 716 717 return (conn); 718 } 719 720 static struct url * 721 _http_get_proxy(const char *flags) 722 { 723 struct url *purl; 724 char *p; 725 726 if (flags != NULL && strchr(flags, 'd') != NULL) 727 return (NULL); 728 if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) && 729 *p && (purl = fetchParseURL(p))) { 730 if (!*purl->scheme) 731 strcpy(purl->scheme, SCHEME_HTTP); 732 if (!purl->port) 733 purl->port = _fetch_default_proxy_port(purl->scheme); 734 if (strcasecmp(purl->scheme, SCHEME_HTTP) == 0) 735 return (purl); 736 fetchFreeURL(purl); 737 } 738 return (NULL); 739 } 740 741 static void 742 _http_print_html(FILE *out, FILE *in) 743 { 744 size_t len; 745 char *line, *p, *q; 746 int comment, tag; 747 748 comment = tag = 0; 749 while ((line = fgetln(in, &len)) != NULL) { 750 while (len && isspace(line[len - 1])) 751 --len; 752 for (p = q = line; q < line + len; ++q) { 753 if (comment && *q == '-') { 754 if (q + 2 < line + len && 755 strcmp(q, "-->") == 0) { 756 tag = comment = 0; 757 q += 2; 758 } 759 } else if (tag && !comment && *q == '>') { 760 p = q + 1; 761 tag = 0; 762 } else if (!tag && *q == '<') { 763 if (q > p) 764 fwrite(p, q - p, 1, out); 765 tag = 1; 766 if (q + 3 < line + len && 767 strcmp(q, "<!--") == 0) { 768 comment = 1; 769 q += 3; 770 } 771 } 772 } 773 if (!tag && q > p) 774 fwrite(p, q - p, 1, out); 775 fputc('\n', out); 776 } 777 } 778 779 780 /***************************************************************************** 781 * Core 782 */ 783 784 /* 785 * Send a request and process the reply 786 * 787 * XXX This function is way too long, the do..while loop should be split 788 * XXX off into a separate function. 789 */ 790 FILE * 791 _http_request(struct url *URL, const char *op, struct url_stat *us, 792 struct url *purl, const char *flags) 793 { 794 conn_t *conn; 795 struct url *url, *new; 796 int chunked, direct, need_auth, noredirect, verbose; 797 int e, i, n, val; 798 off_t offset, clength, length, size; 799 time_t mtime; 800 const char *p; 801 FILE *f; 802 hdr_t h; 803 char hbuf[MAXHOSTNAMELEN + 7], *host; 804 805 direct = CHECK_FLAG('d'); 806 noredirect = CHECK_FLAG('A'); 807 verbose = CHECK_FLAG('v'); 808 809 if (direct && purl) { 810 fetchFreeURL(purl); 811 purl = NULL; 812 } 813 814 /* try the provided URL first */ 815 url = URL; 816 817 /* if the A flag is set, we only get one try */ 818 n = noredirect ? 1 : MAX_REDIRECT; 819 i = 0; 820 821 e = HTTP_PROTOCOL_ERROR; 822 need_auth = 0; 823 do { 824 new = NULL; 825 chunked = 0; 826 offset = 0; 827 clength = -1; 828 length = -1; 829 size = -1; 830 mtime = 0; 831 832 /* check port */ 833 if (!url->port) 834 url->port = _fetch_default_port(url->scheme); 835 836 /* were we redirected to an FTP URL? */ 837 if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) { 838 if (strcmp(op, "GET") == 0) 839 return (_ftp_request(url, "RETR", us, purl, flags)); 840 else if (strcmp(op, "HEAD") == 0) 841 return (_ftp_request(url, "STAT", us, purl, flags)); 842 } 843 844 /* connect to server or proxy */ 845 if ((conn = _http_connect(url, purl, flags)) == NULL) 846 goto ouch; 847 848 host = url->host; 849 #ifdef INET6 850 if (strchr(url->host, ':')) { 851 snprintf(hbuf, sizeof(hbuf), "[%s]", url->host); 852 host = hbuf; 853 } 854 #endif 855 if (url->port != _fetch_default_port(url->scheme)) { 856 if (host != hbuf) { 857 strcpy(hbuf, host); 858 host = hbuf; 859 } 860 snprintf(hbuf + strlen(hbuf), 861 sizeof(hbuf) - strlen(hbuf), ":%d", url->port); 862 } 863 864 /* send request */ 865 if (verbose) 866 _fetch_info("requesting %s://%s%s", 867 url->scheme, host, url->doc); 868 if (purl) { 869 _http_cmd(conn, "%s %s://%s%s HTTP/1.1", 870 op, url->scheme, host, url->doc); 871 } else { 872 _http_cmd(conn, "%s %s HTTP/1.1", 873 op, url->doc); 874 } 875 876 /* virtual host */ 877 _http_cmd(conn, "Host: %s", host); 878 879 /* proxy authorization */ 880 if (purl) { 881 if (*purl->user || *purl->pwd) 882 _http_basic_auth(conn, "Proxy-Authorization", 883 purl->user, purl->pwd); 884 else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL && *p != '\0') 885 _http_authorize(conn, "Proxy-Authorization", p); 886 } 887 888 /* server authorization */ 889 if (need_auth || *url->user || *url->pwd) { 890 if (*url->user || *url->pwd) 891 _http_basic_auth(conn, "Authorization", url->user, url->pwd); 892 else if ((p = getenv("HTTP_AUTH")) != NULL && *p != '\0') 893 _http_authorize(conn, "Authorization", p); 894 else if (fetchAuthMethod && fetchAuthMethod(url) == 0) { 895 _http_basic_auth(conn, "Authorization", url->user, url->pwd); 896 } else { 897 _http_seterr(HTTP_NEED_AUTH); 898 goto ouch; 899 } 900 } 901 902 /* other headers */ 903 if ((p = getenv("HTTP_REFERER")) != NULL && *p != '\0') { 904 if (strcasecmp(p, "auto") == 0) 905 _http_cmd(conn, "Referer: %s://%s%s", 906 url->scheme, host, url->doc); 907 else 908 _http_cmd(conn, "Referer: %s", p); 909 } 910 if ((p = getenv("HTTP_USER_AGENT")) != NULL && *p != '\0') 911 _http_cmd(conn, "User-Agent: %s", p); 912 else 913 _http_cmd(conn, "User-Agent: %s " _LIBFETCH_VER, getprogname()); 914 if (url->offset > 0) 915 _http_cmd(conn, "Range: bytes=%lld-", (long long)url->offset); 916 _http_cmd(conn, "Connection: close"); 917 _http_cmd(conn, ""); 918 919 /* 920 * Force the queued request to be dispatched. Normally, one 921 * would do this with shutdown(2) but squid proxies can be 922 * configured to disallow such half-closed connections. To 923 * be compatible with such configurations, fiddle with socket 924 * options to force the pending data to be written. 925 */ 926 val = 0; 927 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, 928 sizeof(val)); 929 val = 1; 930 setsockopt(conn->sd, IPPROTO_TCP, TCP_NODELAY, &val, 931 sizeof(val)); 932 933 /* get reply */ 934 switch (_http_get_reply(conn)) { 935 case HTTP_OK: 936 case HTTP_PARTIAL: 937 /* fine */ 938 break; 939 case HTTP_MOVED_PERM: 940 case HTTP_MOVED_TEMP: 941 case HTTP_SEE_OTHER: 942 /* 943 * Not so fine, but we still have to read the 944 * headers to get the new location. 945 */ 946 break; 947 case HTTP_NEED_AUTH: 948 if (need_auth) { 949 /* 950 * We already sent out authorization code, 951 * so there's nothing more we can do. 952 */ 953 _http_seterr(conn->err); 954 goto ouch; 955 } 956 /* try again, but send the password this time */ 957 if (verbose) 958 _fetch_info("server requires authorization"); 959 break; 960 case HTTP_NEED_PROXY_AUTH: 961 /* 962 * If we're talking to a proxy, we already sent 963 * our proxy authorization code, so there's 964 * nothing more we can do. 965 */ 966 _http_seterr(conn->err); 967 goto ouch; 968 case HTTP_BAD_RANGE: 969 /* 970 * This can happen if we ask for 0 bytes because 971 * we already have the whole file. Consider this 972 * a success for now, and check sizes later. 973 */ 974 break; 975 case HTTP_PROTOCOL_ERROR: 976 /* fall through */ 977 case -1: 978 _fetch_syserr(); 979 goto ouch; 980 default: 981 _http_seterr(conn->err); 982 if (!verbose) 983 goto ouch; 984 /* fall through so we can get the full error message */ 985 } 986 987 /* get headers */ 988 do { 989 switch ((h = _http_next_header(conn, &p))) { 990 case hdr_syserror: 991 _fetch_syserr(); 992 goto ouch; 993 case hdr_error: 994 _http_seterr(HTTP_PROTOCOL_ERROR); 995 goto ouch; 996 case hdr_content_length: 997 _http_parse_length(p, &clength); 998 break; 999 case hdr_content_range: 1000 _http_parse_range(p, &offset, &length, &size); 1001 break; 1002 case hdr_last_modified: 1003 _http_parse_mtime(p, &mtime); 1004 break; 1005 case hdr_location: 1006 if (!HTTP_REDIRECT(conn->err)) 1007 break; 1008 if (new) 1009 free(new); 1010 if (verbose) 1011 _fetch_info("%d redirect to %s", conn->err, p); 1012 if (*p == '/') 1013 /* absolute path */ 1014 new = fetchMakeURL(url->scheme, url->host, url->port, p, 1015 url->user, url->pwd); 1016 else 1017 new = fetchParseURL(p); 1018 if (new == NULL) { 1019 /* XXX should set an error code */ 1020 DEBUG(fprintf(stderr, "failed to parse new URL\n")); 1021 goto ouch; 1022 } 1023 if (!*new->user && !*new->pwd) { 1024 strcpy(new->user, url->user); 1025 strcpy(new->pwd, url->pwd); 1026 } 1027 new->offset = url->offset; 1028 new->length = url->length; 1029 break; 1030 case hdr_transfer_encoding: 1031 /* XXX weak test*/ 1032 chunked = (strcasecmp(p, "chunked") == 0); 1033 break; 1034 case hdr_www_authenticate: 1035 if (conn->err != HTTP_NEED_AUTH) 1036 break; 1037 /* if we were smarter, we'd check the method and realm */ 1038 break; 1039 case hdr_end: 1040 /* fall through */ 1041 case hdr_unknown: 1042 /* ignore */ 1043 break; 1044 } 1045 } while (h > hdr_end); 1046 1047 /* we need to provide authentication */ 1048 if (conn->err == HTTP_NEED_AUTH) { 1049 e = conn->err; 1050 need_auth = 1; 1051 _fetch_close(conn); 1052 conn = NULL; 1053 continue; 1054 } 1055 1056 /* requested range not satisfiable */ 1057 if (conn->err == HTTP_BAD_RANGE) { 1058 if (url->offset == size && url->length == 0) { 1059 /* asked for 0 bytes; fake it */ 1060 offset = url->offset; 1061 conn->err = HTTP_OK; 1062 break; 1063 } else { 1064 _http_seterr(conn->err); 1065 goto ouch; 1066 } 1067 } 1068 1069 /* we have a hit or an error */ 1070 if (conn->err == HTTP_OK || conn->err == HTTP_PARTIAL || HTTP_ERROR(conn->err)) 1071 break; 1072 1073 /* all other cases: we got a redirect */ 1074 e = conn->err; 1075 need_auth = 0; 1076 _fetch_close(conn); 1077 conn = NULL; 1078 if (!new) { 1079 DEBUG(fprintf(stderr, "redirect with no new location\n")); 1080 break; 1081 } 1082 if (url != URL) 1083 fetchFreeURL(url); 1084 url = new; 1085 } while (++i < n); 1086 1087 /* we failed, or ran out of retries */ 1088 if (conn == NULL) { 1089 _http_seterr(e); 1090 goto ouch; 1091 } 1092 1093 DEBUG(fprintf(stderr, "offset %lld, length %lld," 1094 " size %lld, clength %lld\n", 1095 (long long)offset, (long long)length, 1096 (long long)size, (long long)clength)); 1097 1098 /* check for inconsistencies */ 1099 if (clength != -1 && length != -1 && clength != length) { 1100 _http_seterr(HTTP_PROTOCOL_ERROR); 1101 goto ouch; 1102 } 1103 if (clength == -1) 1104 clength = length; 1105 if (clength != -1) 1106 length = offset + clength; 1107 if (length != -1 && size != -1 && length != size) { 1108 _http_seterr(HTTP_PROTOCOL_ERROR); 1109 goto ouch; 1110 } 1111 if (size == -1) 1112 size = length; 1113 1114 /* fill in stats */ 1115 if (us) { 1116 us->size = size; 1117 us->atime = us->mtime = mtime; 1118 } 1119 1120 /* too far? */ 1121 if (URL->offset > 0 && offset > URL->offset) { 1122 _http_seterr(HTTP_PROTOCOL_ERROR); 1123 goto ouch; 1124 } 1125 1126 /* report back real offset and size */ 1127 URL->offset = offset; 1128 URL->length = clength; 1129 1130 /* wrap it up in a FILE */ 1131 if ((f = _http_funopen(conn, chunked)) == NULL) { 1132 _fetch_syserr(); 1133 goto ouch; 1134 } 1135 1136 if (url != URL) 1137 fetchFreeURL(url); 1138 if (purl) 1139 fetchFreeURL(purl); 1140 1141 if (HTTP_ERROR(conn->err)) { 1142 _http_print_html(stderr, f); 1143 fclose(f); 1144 f = NULL; 1145 } 1146 1147 return (f); 1148 1149 ouch: 1150 if (url != URL) 1151 fetchFreeURL(url); 1152 if (purl) 1153 fetchFreeURL(purl); 1154 if (conn != NULL) 1155 _fetch_close(conn); 1156 return (NULL); 1157 } 1158 1159 1160 /***************************************************************************** 1161 * Entry points 1162 */ 1163 1164 /* 1165 * Retrieve and stat a file by HTTP 1166 */ 1167 FILE * 1168 fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags) 1169 { 1170 return (_http_request(URL, "GET", us, _http_get_proxy(flags), flags)); 1171 } 1172 1173 /* 1174 * Retrieve a file by HTTP 1175 */ 1176 FILE * 1177 fetchGetHTTP(struct url *URL, const char *flags) 1178 { 1179 return (fetchXGetHTTP(URL, NULL, flags)); 1180 } 1181 1182 /* 1183 * Store a file by HTTP 1184 */ 1185 FILE * 1186 fetchPutHTTP(struct url *URL __unused, const char *flags __unused) 1187 { 1188 warnx("fetchPutHTTP(): not implemented"); 1189 return (NULL); 1190 } 1191 1192 /* 1193 * Get an HTTP document's metadata 1194 */ 1195 int 1196 fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags) 1197 { 1198 FILE *f; 1199 1200 f = _http_request(URL, "HEAD", us, _http_get_proxy(flags), flags); 1201 if (f == NULL) 1202 return (-1); 1203 fclose(f); 1204 return (0); 1205 } 1206 1207 /* 1208 * List a directory 1209 */ 1210 struct url_ent * 1211 fetchListHTTP(struct url *url __unused, const char *flags __unused) 1212 { 1213 warnx("fetchListHTTP(): not implemented"); 1214 return (NULL); 1215 } 1216