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 1.86 2008/12/15 08:27:44 murray 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 #include <sys/time.h> 67 68 #include <ctype.h> 69 #include <err.h> 70 #include <errno.h> 71 #include <locale.h> 72 #include <netdb.h> 73 #include <stdarg.h> 74 #include <stdio.h> 75 #include <stdlib.h> 76 #include <string.h> 77 #include <time.h> 78 #include <unistd.h> 79 80 #include <netinet/in.h> 81 #include <netinet/tcp.h> 82 83 #include "fetch.h" 84 #include "common.h" 85 #include "httperr.h" 86 87 /* Maximum number of redirects to follow */ 88 #define MAX_REDIRECT 5 89 90 /* Symbolic names for reply codes we care about */ 91 #define HTTP_OK 200 92 #define HTTP_PARTIAL 206 93 #define HTTP_MOVED_PERM 301 94 #define HTTP_MOVED_TEMP 302 95 #define HTTP_SEE_OTHER 303 96 #define HTTP_NOT_MODIFIED 304 97 #define HTTP_TEMP_REDIRECT 307 98 #define HTTP_NEED_AUTH 401 99 #define HTTP_NEED_PROXY_AUTH 407 100 #define HTTP_BAD_RANGE 416 101 #define HTTP_PROTOCOL_ERROR 999 102 103 #define HTTP_REDIRECT(xyz) ((xyz) == HTTP_MOVED_PERM \ 104 || (xyz) == HTTP_MOVED_TEMP \ 105 || (xyz) == HTTP_TEMP_REDIRECT \ 106 || (xyz) == HTTP_SEE_OTHER) 107 108 #define HTTP_ERROR(xyz) ((xyz) > 400 && (xyz) < 599) 109 110 111 /***************************************************************************** 112 * I/O functions for decoding chunked streams 113 */ 114 115 struct httpio 116 { 117 conn_t *conn; /* connection */ 118 int chunked; /* chunked mode */ 119 char *buf; /* chunk buffer */ 120 size_t bufsize; /* size of chunk buffer */ 121 ssize_t buflen; /* amount of data currently in buffer */ 122 int bufpos; /* current read offset in buffer */ 123 int eof; /* end-of-file flag */ 124 int error; /* error flag */ 125 size_t chunksize; /* remaining size of current chunk */ 126 #ifndef NDEBUG 127 size_t total; 128 #endif 129 }; 130 131 static int http_cmd(conn_t *, const char *, ...) __printflike(2, 3); 132 133 /* 134 * Get next chunk header 135 */ 136 static int 137 http_new_chunk(struct httpio *io) 138 { 139 char *p; 140 141 if (fetch_getln(io->conn) == -1) 142 return (-1); 143 144 if (io->conn->buflen < 2 || !isxdigit((unsigned char)*io->conn->buf)) 145 return (-1); 146 147 for (p = io->conn->buf; *p && !isspace((unsigned char)*p); ++p) { 148 if (*p == ';') 149 break; 150 if (!isxdigit((unsigned char)*p)) 151 return (-1); 152 if (isdigit((unsigned char)*p)) { 153 io->chunksize = io->chunksize * 16 + 154 *p - '0'; 155 } else { 156 io->chunksize = io->chunksize * 16 + 157 10 + tolower((unsigned char)*p) - 'a'; 158 } 159 } 160 161 #ifndef NDEBUG 162 if (fetchDebug) { 163 io->total += io->chunksize; 164 if (io->chunksize == 0) 165 fprintf(stderr, "%s(): end of last chunk\n", __func__); 166 else 167 fprintf(stderr, "%s(): new chunk: %lu (%lu)\n", 168 __func__, (unsigned long)io->chunksize, 169 (unsigned long)io->total); 170 } 171 #endif 172 173 return (io->chunksize); 174 } 175 176 /* 177 * Grow the input buffer to at least len bytes 178 */ 179 static inline int 180 http_growbuf(struct httpio *io, size_t len) 181 { 182 char *tmp; 183 184 if (io->bufsize >= len) 185 return (0); 186 187 if ((tmp = realloc(io->buf, len)) == NULL) 188 return (-1); 189 io->buf = tmp; 190 io->bufsize = len; 191 return (0); 192 } 193 194 /* 195 * Fill the input buffer, do chunk decoding on the fly 196 */ 197 static int 198 http_fillbuf(struct httpio *io, size_t len) 199 { 200 if (io->error) 201 return (-1); 202 if (io->eof) 203 return (0); 204 205 if (io->chunked == 0) { 206 if (http_growbuf(io, len) == -1) 207 return (-1); 208 if ((io->buflen = fetch_read(io->conn, io->buf, len)) == -1) { 209 io->error = 1; 210 return (-1); 211 } 212 io->bufpos = 0; 213 return (io->buflen); 214 } 215 216 if (io->chunksize == 0) { 217 switch (http_new_chunk(io)) { 218 case -1: 219 io->error = 1; 220 return (-1); 221 case 0: 222 io->eof = 1; 223 return (0); 224 } 225 } 226 227 if (len > io->chunksize) 228 len = io->chunksize; 229 if (http_growbuf(io, len) == -1) 230 return (-1); 231 if ((io->buflen = fetch_read(io->conn, io->buf, len)) == -1) { 232 io->error = 1; 233 return (-1); 234 } 235 io->chunksize -= io->buflen; 236 237 if (io->chunksize == 0) { 238 char endl[2]; 239 240 if (fetch_read(io->conn, endl, 2) != 2 || 241 endl[0] != '\r' || endl[1] != '\n') 242 return (-1); 243 } 244 245 io->bufpos = 0; 246 247 return (io->buflen); 248 } 249 250 /* 251 * Read function 252 */ 253 static int 254 http_readfn(void *v, char *buf, int len) 255 { 256 struct httpio *io = (struct httpio *)v; 257 int l, pos; 258 259 if (io->error) 260 return (-1); 261 if (io->eof) 262 return (0); 263 264 for (pos = 0; len > 0; pos += l, len -= l) { 265 /* empty buffer */ 266 if (!io->buf || io->bufpos == io->buflen) 267 if (http_fillbuf(io, len) < 1) 268 break; 269 l = io->buflen - io->bufpos; 270 if (len < l) 271 l = len; 272 memcpy(buf + pos, io->buf + io->bufpos, l); 273 io->bufpos += l; 274 } 275 276 if (!pos && io->error) 277 return (-1); 278 return (pos); 279 } 280 281 /* 282 * Write function 283 */ 284 static int 285 http_writefn(void *v, const char *buf, int len) 286 { 287 struct httpio *io = (struct httpio *)v; 288 289 return (fetch_write(io->conn, buf, len)); 290 } 291 292 /* 293 * Close function 294 */ 295 static int 296 http_closefn(void *v) 297 { 298 struct httpio *io = (struct httpio *)v; 299 int r; 300 301 r = fetch_close(io->conn); 302 if (io->buf) 303 free(io->buf); 304 free(io); 305 return (r); 306 } 307 308 /* 309 * Wrap a file descriptor up 310 */ 311 static FILE * 312 http_funopen(conn_t *conn, int chunked) 313 { 314 struct httpio *io; 315 FILE *f; 316 317 if ((io = calloc(1, sizeof(*io))) == NULL) { 318 fetch_syserr(); 319 return (NULL); 320 } 321 io->conn = conn; 322 io->chunked = chunked; 323 f = funopen(io, http_readfn, http_writefn, NULL, http_closefn); 324 if (f == NULL) { 325 fetch_syserr(); 326 free(io); 327 return (NULL); 328 } 329 return (f); 330 } 331 332 333 /***************************************************************************** 334 * Helper functions for talking to the server and parsing its replies 335 */ 336 337 /* Header types */ 338 typedef enum { 339 hdr_syserror = -2, 340 hdr_error = -1, 341 hdr_end = 0, 342 hdr_unknown = 1, 343 hdr_content_length, 344 hdr_content_range, 345 hdr_last_modified, 346 hdr_location, 347 hdr_transfer_encoding, 348 hdr_www_authenticate 349 } hdr_t; 350 351 /* Names of interesting headers */ 352 static struct { 353 hdr_t num; 354 const char *name; 355 } hdr_names[] = { 356 { hdr_content_length, "Content-Length" }, 357 { hdr_content_range, "Content-Range" }, 358 { hdr_last_modified, "Last-Modified" }, 359 { hdr_location, "Location" }, 360 { hdr_transfer_encoding, "Transfer-Encoding" }, 361 { hdr_www_authenticate, "WWW-Authenticate" }, 362 { hdr_unknown, NULL }, 363 }; 364 365 /* 366 * Send a formatted line; optionally echo to terminal 367 */ 368 static int 369 http_cmd(conn_t *conn, const char *fmt, ...) 370 { 371 va_list ap; 372 size_t len; 373 char *msg; 374 int r; 375 376 va_start(ap, fmt); 377 len = vasprintf(&msg, fmt, ap); 378 va_end(ap); 379 380 if (msg == NULL) { 381 errno = ENOMEM; 382 fetch_syserr(); 383 return (-1); 384 } 385 386 r = fetch_putln(conn, msg, len); 387 free(msg); 388 389 if (r == -1) { 390 fetch_syserr(); 391 return (-1); 392 } 393 394 return (0); 395 } 396 397 /* 398 * Get and parse status line 399 */ 400 static int 401 http_get_reply(conn_t *conn) 402 { 403 char *p; 404 405 if (fetch_getln(conn) == -1) 406 return (-1); 407 /* 408 * A valid status line looks like "HTTP/m.n xyz reason" where m 409 * and n are the major and minor protocol version numbers and xyz 410 * is the reply code. 411 * Unfortunately, there are servers out there (NCSA 1.5.1, to name 412 * just one) that do not send a version number, so we can't rely 413 * on finding one, but if we do, insist on it being 1.0 or 1.1. 414 * We don't care about the reason phrase. 415 */ 416 if (strncmp(conn->buf, "HTTP", 4) != 0) 417 return (HTTP_PROTOCOL_ERROR); 418 p = conn->buf + 4; 419 if (*p == '/') { 420 if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1')) 421 return (HTTP_PROTOCOL_ERROR); 422 p += 4; 423 } 424 if (*p != ' ' || 425 !isdigit((unsigned char)p[1]) || 426 !isdigit((unsigned char)p[2]) || 427 !isdigit((unsigned char)p[3])) 428 return (HTTP_PROTOCOL_ERROR); 429 430 conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0'); 431 return (conn->err); 432 } 433 434 /* 435 * Check a header; if the type matches the given string, return a pointer 436 * to the beginning of the value. 437 */ 438 static const char * 439 http_match(const char *str, const char *hdr) 440 { 441 while (*str && *hdr && 442 tolower((unsigned char)*str++) == tolower((unsigned char)*hdr++)) 443 /* nothing */; 444 if (*str || *hdr != ':') 445 return (NULL); 446 while (*hdr && isspace((unsigned char)*++hdr)) 447 /* nothing */; 448 return (hdr); 449 } 450 451 /* 452 * Get the next header and return the appropriate symbolic code. 453 */ 454 static hdr_t 455 http_next_header(conn_t *conn, const char **p) 456 { 457 int i; 458 459 if (fetch_getln(conn) == -1) 460 return (hdr_syserror); 461 while (conn->buflen && isspace((unsigned char)conn->buf[conn->buflen - 1])) 462 conn->buflen--; 463 conn->buf[conn->buflen] = '\0'; 464 if (conn->buflen == 0) 465 return (hdr_end); 466 /* 467 * We could check for malformed headers but we don't really care. 468 * A valid header starts with a token immediately followed by a 469 * colon; a token is any sequence of non-control, non-whitespace 470 * characters except "()<>@,;:\\\"{}". 471 */ 472 for (i = 0; hdr_names[i].num != hdr_unknown; i++) 473 if ((*p = http_match(hdr_names[i].name, conn->buf)) != NULL) 474 return (hdr_names[i].num); 475 return (hdr_unknown); 476 } 477 478 /* 479 * Parse a last-modified header 480 */ 481 static int 482 http_parse_mtime(const char *p, time_t *mtime) 483 { 484 char locale[64], *r; 485 struct tm tm; 486 487 strncpy(locale, setlocale(LC_TIME, NULL), sizeof(locale)); 488 setlocale(LC_TIME, "C"); 489 r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm); 490 /* XXX should add support for date-2 and date-3 */ 491 setlocale(LC_TIME, locale); 492 if (r == NULL) 493 return (-1); 494 DEBUG(fprintf(stderr, "last modified: [%04d-%02d-%02d " 495 "%02d:%02d:%02d]\n", 496 tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday, 497 tm.tm_hour, tm.tm_min, tm.tm_sec)); 498 *mtime = timegm(&tm); 499 return (0); 500 } 501 502 /* 503 * Parse a content-length header 504 */ 505 static int 506 http_parse_length(const char *p, off_t *length) 507 { 508 off_t len; 509 510 for (len = 0; *p && isdigit((unsigned char)*p); ++p) 511 len = len * 10 + (*p - '0'); 512 if (*p) 513 return (-1); 514 DEBUG(fprintf(stderr, "content length: [%lld]\n", 515 (long long)len)); 516 *length = len; 517 return (0); 518 } 519 520 /* 521 * Parse a content-range header 522 */ 523 static int 524 http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size) 525 { 526 off_t first, last, len; 527 528 if (strncasecmp(p, "bytes ", 6) != 0) 529 return (-1); 530 p += 6; 531 if (*p == '*') { 532 first = last = -1; 533 ++p; 534 } else { 535 for (first = 0; *p && isdigit((unsigned char)*p); ++p) 536 first = first * 10 + *p - '0'; 537 if (*p != '-') 538 return (-1); 539 for (last = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 540 last = last * 10 + *p - '0'; 541 } 542 if (first > last || *p != '/') 543 return (-1); 544 for (len = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 545 len = len * 10 + *p - '0'; 546 if (*p || len < last - first + 1) 547 return (-1); 548 if (first == -1) { 549 DEBUG(fprintf(stderr, "content range: [*/%lld]\n", 550 (long long)len)); 551 *length = 0; 552 } else { 553 DEBUG(fprintf(stderr, "content range: [%lld-%lld/%lld]\n", 554 (long long)first, (long long)last, (long long)len)); 555 *length = last - first + 1; 556 } 557 *offset = first; 558 *size = len; 559 return (0); 560 } 561 562 563 /***************************************************************************** 564 * Helper functions for authorization 565 */ 566 567 /* 568 * Base64 encoding 569 */ 570 static char * 571 http_base64(const char *src) 572 { 573 static const char base64[] = 574 "ABCDEFGHIJKLMNOPQRSTUVWXYZ" 575 "abcdefghijklmnopqrstuvwxyz" 576 "0123456789+/"; 577 char *str, *dst; 578 size_t l; 579 int t, r; 580 581 l = strlen(src); 582 if ((str = malloc(((l + 2) / 3) * 4 + 1)) == NULL) 583 return (NULL); 584 dst = str; 585 r = 0; 586 587 while (l >= 3) { 588 t = (src[0] << 16) | (src[1] << 8) | src[2]; 589 dst[0] = base64[(t >> 18) & 0x3f]; 590 dst[1] = base64[(t >> 12) & 0x3f]; 591 dst[2] = base64[(t >> 6) & 0x3f]; 592 dst[3] = base64[(t >> 0) & 0x3f]; 593 src += 3; l -= 3; 594 dst += 4; r += 4; 595 } 596 597 switch (l) { 598 case 2: 599 t = (src[0] << 16) | (src[1] << 8); 600 dst[0] = base64[(t >> 18) & 0x3f]; 601 dst[1] = base64[(t >> 12) & 0x3f]; 602 dst[2] = base64[(t >> 6) & 0x3f]; 603 dst[3] = '='; 604 dst += 4; 605 r += 4; 606 break; 607 case 1: 608 t = src[0] << 16; 609 dst[0] = base64[(t >> 18) & 0x3f]; 610 dst[1] = base64[(t >> 12) & 0x3f]; 611 dst[2] = dst[3] = '='; 612 dst += 4; 613 r += 4; 614 break; 615 case 0: 616 break; 617 } 618 619 *dst = 0; 620 return (str); 621 } 622 623 /* 624 * Encode username and password 625 */ 626 static int 627 http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd) 628 { 629 char *upw, *auth; 630 int r; 631 632 DEBUG(fprintf(stderr, "usr: [%s]\n", usr)); 633 DEBUG(fprintf(stderr, "pwd: [%s]\n", pwd)); 634 if (asprintf(&upw, "%s:%s", usr, pwd) == -1) 635 return (-1); 636 auth = http_base64(upw); 637 free(upw); 638 if (auth == NULL) 639 return (-1); 640 r = http_cmd(conn, "%s: Basic %s", hdr, auth); 641 free(auth); 642 return (r); 643 } 644 645 /* 646 * Send an authorization header 647 */ 648 static int 649 http_authorize(conn_t *conn, const char *hdr, const char *p) 650 { 651 /* basic authorization */ 652 if (strncasecmp(p, "basic:", 6) == 0) { 653 char *user, *pwd, *str; 654 int r; 655 656 /* skip realm */ 657 for (p += 6; *p && *p != ':'; ++p) 658 /* nothing */ ; 659 if (!*p || strchr(++p, ':') == NULL) 660 return (-1); 661 if ((str = strdup(p)) == NULL) 662 return (-1); /* XXX */ 663 user = str; 664 pwd = strchr(str, ':'); 665 *pwd++ = '\0'; 666 r = http_basic_auth(conn, hdr, user, pwd); 667 free(str); 668 return (r); 669 } 670 return (-1); 671 } 672 673 674 /***************************************************************************** 675 * Helper functions for connecting to a server or proxy 676 */ 677 678 /* 679 * Connect to the correct HTTP server or proxy. 680 */ 681 static conn_t * 682 http_connect(struct url *URL, struct url *purl, const char *flags) 683 { 684 conn_t *conn; 685 int verbose; 686 int af, val; 687 688 #ifdef INET6 689 af = AF_UNSPEC; 690 #else 691 af = AF_INET; 692 #endif 693 694 verbose = CHECK_FLAG('v'); 695 if (CHECK_FLAG('4')) 696 af = AF_INET; 697 #ifdef INET6 698 else if (CHECK_FLAG('6')) 699 af = AF_INET6; 700 #endif 701 702 if (purl && strcasecmp(URL->scheme, SCHEME_HTTPS) != 0) { 703 URL = purl; 704 } else if (strcasecmp(URL->scheme, SCHEME_FTP) == 0) { 705 /* can't talk http to an ftp server */ 706 /* XXX should set an error code */ 707 return (NULL); 708 } 709 710 if ((conn = fetch_connect(URL->host, URL->port, af, verbose)) == NULL) 711 /* fetch_connect() has already set an error code */ 712 return (NULL); 713 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 && 714 fetch_ssl(conn, verbose) == -1) { 715 fetch_close(conn); 716 /* grrr */ 717 errno = EAUTH; 718 fetch_syserr(); 719 return (NULL); 720 } 721 722 val = 1; 723 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val)); 724 725 return (conn); 726 } 727 728 static struct url * 729 http_get_proxy(struct url * url, const char *flags) 730 { 731 struct url *purl; 732 char *p; 733 734 if (flags != NULL && strchr(flags, 'd') != NULL) 735 return (NULL); 736 if (fetch_no_proxy_match(url->host)) 737 return (NULL); 738 if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) && 739 *p && (purl = fetchParseURL(p))) { 740 if (!*purl->scheme) 741 strcpy(purl->scheme, SCHEME_HTTP); 742 if (!purl->port) 743 purl->port = fetch_default_proxy_port(purl->scheme); 744 if (strcasecmp(purl->scheme, SCHEME_HTTP) == 0) 745 return (purl); 746 fetchFreeURL(purl); 747 } 748 return (NULL); 749 } 750 751 static void 752 http_print_html(FILE *out, FILE *in) 753 { 754 size_t len; 755 char *line, *p, *q; 756 int comment, tag; 757 758 comment = tag = 0; 759 while ((line = fgetln(in, &len)) != NULL) { 760 while (len && isspace((unsigned char)line[len - 1])) 761 --len; 762 for (p = q = line; q < line + len; ++q) { 763 if (comment && *q == '-') { 764 if (q + 2 < line + len && 765 strcmp(q, "-->") == 0) { 766 tag = comment = 0; 767 q += 2; 768 } 769 } else if (tag && !comment && *q == '>') { 770 p = q + 1; 771 tag = 0; 772 } else if (!tag && *q == '<') { 773 if (q > p) 774 fwrite(p, q - p, 1, out); 775 tag = 1; 776 if (q + 3 < line + len && 777 strcmp(q, "<!--") == 0) { 778 comment = 1; 779 q += 3; 780 } 781 } 782 } 783 if (!tag && q > p) 784 fwrite(p, q - p, 1, out); 785 fputc('\n', out); 786 } 787 } 788 789 790 /***************************************************************************** 791 * Core 792 */ 793 794 /* 795 * Send a request and process the reply 796 * 797 * XXX This function is way too long, the do..while loop should be split 798 * XXX off into a separate function. 799 */ 800 FILE * 801 http_request(struct url *URL, const char *op, struct url_stat *us, 802 struct url *purl, const char *flags) 803 { 804 char timebuf[80]; 805 char hbuf[MAXHOSTNAMELEN + 7], *host; 806 conn_t *conn; 807 struct url *url, *new; 808 int chunked, direct, ims, need_auth, noredirect, verbose; 809 int e, i, n, val; 810 off_t offset, clength, length, size; 811 time_t mtime; 812 const char *p; 813 FILE *f; 814 hdr_t h; 815 struct tm *timestruct; 816 817 direct = CHECK_FLAG('d'); 818 noredirect = CHECK_FLAG('A'); 819 verbose = CHECK_FLAG('v'); 820 ims = CHECK_FLAG('i'); 821 822 if (direct && purl) { 823 fetchFreeURL(purl); 824 purl = NULL; 825 } 826 827 /* try the provided URL first */ 828 url = URL; 829 830 /* if the A flag is set, we only get one try */ 831 n = noredirect ? 1 : MAX_REDIRECT; 832 i = 0; 833 834 e = HTTP_PROTOCOL_ERROR; 835 need_auth = 0; 836 do { 837 new = NULL; 838 chunked = 0; 839 offset = 0; 840 clength = -1; 841 length = -1; 842 size = -1; 843 mtime = 0; 844 845 /* check port */ 846 if (!url->port) 847 url->port = fetch_default_port(url->scheme); 848 849 /* were we redirected to an FTP URL? */ 850 if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) { 851 if (strcmp(op, "GET") == 0) 852 return (ftp_request(url, "RETR", us, purl, flags)); 853 else if (strcmp(op, "HEAD") == 0) 854 return (ftp_request(url, "STAT", us, purl, flags)); 855 } 856 857 /* connect to server or proxy */ 858 if ((conn = http_connect(url, purl, flags)) == NULL) 859 goto ouch; 860 861 host = url->host; 862 #ifdef INET6 863 if (strchr(url->host, ':')) { 864 snprintf(hbuf, sizeof(hbuf), "[%s]", url->host); 865 host = hbuf; 866 } 867 #endif 868 if (url->port != fetch_default_port(url->scheme)) { 869 if (host != hbuf) { 870 strcpy(hbuf, host); 871 host = hbuf; 872 } 873 snprintf(hbuf + strlen(hbuf), 874 sizeof(hbuf) - strlen(hbuf), ":%d", url->port); 875 } 876 877 /* send request */ 878 if (verbose) 879 fetch_info("requesting %s://%s%s", 880 url->scheme, host, url->doc); 881 if (purl) { 882 http_cmd(conn, "%s %s://%s%s HTTP/1.1", 883 op, url->scheme, host, url->doc); 884 } else { 885 http_cmd(conn, "%s %s HTTP/1.1", 886 op, url->doc); 887 } 888 889 if (ims && url->ims_time) { 890 timestruct = gmtime((time_t *)&url->ims_time); 891 (void)strftime(timebuf, 80, "%a, %d %b %Y %T GMT", 892 timestruct); 893 if (verbose) 894 fetch_info("If-Modified-Since: %s", timebuf); 895 http_cmd(conn, "If-Modified-Since: %s", timebuf); 896 } 897 /* virtual host */ 898 http_cmd(conn, "Host: %s", host); 899 900 /* proxy authorization */ 901 if (purl) { 902 if (*purl->user || *purl->pwd) 903 http_basic_auth(conn, "Proxy-Authorization", 904 purl->user, purl->pwd); 905 else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL && *p != '\0') 906 http_authorize(conn, "Proxy-Authorization", p); 907 } 908 909 /* server authorization */ 910 if (need_auth || *url->user || *url->pwd) { 911 if (*url->user || *url->pwd) 912 http_basic_auth(conn, "Authorization", url->user, url->pwd); 913 else if ((p = getenv("HTTP_AUTH")) != NULL && *p != '\0') 914 http_authorize(conn, "Authorization", p); 915 else if (fetchAuthMethod && fetchAuthMethod(url) == 0) { 916 http_basic_auth(conn, "Authorization", url->user, url->pwd); 917 } else { 918 http_seterr(HTTP_NEED_AUTH); 919 goto ouch; 920 } 921 } 922 923 /* other headers */ 924 if ((p = getenv("HTTP_REFERER")) != NULL && *p != '\0') { 925 if (strcasecmp(p, "auto") == 0) 926 http_cmd(conn, "Referer: %s://%s%s", 927 url->scheme, host, url->doc); 928 else 929 http_cmd(conn, "Referer: %s", p); 930 } 931 if ((p = getenv("HTTP_USER_AGENT")) != NULL && *p != '\0') 932 http_cmd(conn, "User-Agent: %s", p); 933 else 934 http_cmd(conn, "User-Agent: %s " _LIBFETCH_VER, getprogname()); 935 if (url->offset > 0) 936 http_cmd(conn, "Range: bytes=%lld-", (long long)url->offset); 937 http_cmd(conn, "Connection: close"); 938 http_cmd(conn, "%s", ""); 939 940 /* 941 * Force the queued request to be dispatched. Normally, one 942 * would do this with shutdown(2) but squid proxies can be 943 * configured to disallow such half-closed connections. To 944 * be compatible with such configurations, fiddle with socket 945 * options to force the pending data to be written. 946 */ 947 val = 0; 948 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, 949 sizeof(val)); 950 val = 1; 951 setsockopt(conn->sd, IPPROTO_TCP, TCP_NODELAY, &val, 952 sizeof(val)); 953 954 /* get reply */ 955 switch (http_get_reply(conn)) { 956 case HTTP_OK: 957 case HTTP_PARTIAL: 958 case HTTP_NOT_MODIFIED: 959 /* fine */ 960 break; 961 case HTTP_MOVED_PERM: 962 case HTTP_MOVED_TEMP: 963 case HTTP_SEE_OTHER: 964 /* 965 * Not so fine, but we still have to read the 966 * headers to get the new location. 967 */ 968 break; 969 case HTTP_NEED_AUTH: 970 if (need_auth) { 971 /* 972 * We already sent out authorization code, 973 * so there's nothing more we can do. 974 */ 975 http_seterr(conn->err); 976 goto ouch; 977 } 978 /* try again, but send the password this time */ 979 if (verbose) 980 fetch_info("server requires authorization"); 981 break; 982 case HTTP_NEED_PROXY_AUTH: 983 /* 984 * If we're talking to a proxy, we already sent 985 * our proxy authorization code, so there's 986 * nothing more we can do. 987 */ 988 http_seterr(conn->err); 989 goto ouch; 990 case HTTP_BAD_RANGE: 991 /* 992 * This can happen if we ask for 0 bytes because 993 * we already have the whole file. Consider this 994 * a success for now, and check sizes later. 995 */ 996 break; 997 case HTTP_PROTOCOL_ERROR: 998 /* fall through */ 999 case -1: 1000 fetch_syserr(); 1001 goto ouch; 1002 default: 1003 http_seterr(conn->err); 1004 if (!verbose) 1005 goto ouch; 1006 /* fall through so we can get the full error message */ 1007 } 1008 1009 /* get headers */ 1010 do { 1011 switch ((h = http_next_header(conn, &p))) { 1012 case hdr_syserror: 1013 fetch_syserr(); 1014 goto ouch; 1015 case hdr_error: 1016 http_seterr(HTTP_PROTOCOL_ERROR); 1017 goto ouch; 1018 case hdr_content_length: 1019 http_parse_length(p, &clength); 1020 break; 1021 case hdr_content_range: 1022 http_parse_range(p, &offset, &length, &size); 1023 break; 1024 case hdr_last_modified: 1025 http_parse_mtime(p, &mtime); 1026 break; 1027 case hdr_location: 1028 if (!HTTP_REDIRECT(conn->err)) 1029 break; 1030 if (new) 1031 free(new); 1032 if (verbose) 1033 fetch_info("%d redirect to %s", conn->err, p); 1034 if (*p == '/') 1035 /* absolute path */ 1036 new = fetchMakeURL(url->scheme, url->host, url->port, p, 1037 url->user, url->pwd); 1038 else 1039 new = fetchParseURL(p); 1040 if (new == NULL) { 1041 /* XXX should set an error code */ 1042 DEBUG(fprintf(stderr, "failed to parse new URL\n")); 1043 goto ouch; 1044 } 1045 if (!*new->user && !*new->pwd) { 1046 strcpy(new->user, url->user); 1047 strcpy(new->pwd, url->pwd); 1048 } 1049 new->offset = url->offset; 1050 new->length = url->length; 1051 break; 1052 case hdr_transfer_encoding: 1053 /* XXX weak test*/ 1054 chunked = (strcasecmp(p, "chunked") == 0); 1055 break; 1056 case hdr_www_authenticate: 1057 if (conn->err != HTTP_NEED_AUTH) 1058 break; 1059 /* if we were smarter, we'd check the method and realm */ 1060 break; 1061 case hdr_end: 1062 /* fall through */ 1063 case hdr_unknown: 1064 /* ignore */ 1065 break; 1066 } 1067 } while (h > hdr_end); 1068 1069 /* we need to provide authentication */ 1070 if (conn->err == HTTP_NEED_AUTH) { 1071 e = conn->err; 1072 need_auth = 1; 1073 fetch_close(conn); 1074 conn = NULL; 1075 continue; 1076 } 1077 1078 /* requested range not satisfiable */ 1079 if (conn->err == HTTP_BAD_RANGE) { 1080 if (url->offset == size && url->length == 0) { 1081 /* asked for 0 bytes; fake it */ 1082 offset = url->offset; 1083 clength = -1; 1084 conn->err = HTTP_OK; 1085 break; 1086 } else { 1087 http_seterr(conn->err); 1088 goto ouch; 1089 } 1090 } 1091 1092 /* we have a hit or an error */ 1093 if (conn->err == HTTP_OK 1094 || conn->err == HTTP_NOT_MODIFIED 1095 || conn->err == HTTP_PARTIAL 1096 || HTTP_ERROR(conn->err)) 1097 break; 1098 1099 /* all other cases: we got a redirect */ 1100 e = conn->err; 1101 need_auth = 0; 1102 fetch_close(conn); 1103 conn = NULL; 1104 if (!new) { 1105 DEBUG(fprintf(stderr, "redirect with no new location\n")); 1106 break; 1107 } 1108 if (url != URL) 1109 fetchFreeURL(url); 1110 url = new; 1111 } while (++i < n); 1112 1113 /* we failed, or ran out of retries */ 1114 if (conn == NULL) { 1115 http_seterr(e); 1116 goto ouch; 1117 } 1118 1119 DEBUG(fprintf(stderr, "offset %lld, length %lld," 1120 " size %lld, clength %lld\n", 1121 (long long)offset, (long long)length, 1122 (long long)size, (long long)clength)); 1123 1124 if (conn->err == HTTP_NOT_MODIFIED) { 1125 http_seterr(HTTP_NOT_MODIFIED); 1126 return (NULL); 1127 } 1128 1129 /* check for inconsistencies */ 1130 if (clength != -1 && length != -1 && clength != length) { 1131 http_seterr(HTTP_PROTOCOL_ERROR); 1132 goto ouch; 1133 } 1134 if (clength == -1) 1135 clength = length; 1136 if (clength != -1) 1137 length = offset + clength; 1138 if (length != -1 && size != -1 && length != size) { 1139 http_seterr(HTTP_PROTOCOL_ERROR); 1140 goto ouch; 1141 } 1142 if (size == -1) 1143 size = length; 1144 1145 /* fill in stats */ 1146 if (us) { 1147 us->size = size; 1148 us->atime = us->mtime = mtime; 1149 } 1150 1151 /* too far? */ 1152 if (URL->offset > 0 && offset > URL->offset) { 1153 http_seterr(HTTP_PROTOCOL_ERROR); 1154 goto ouch; 1155 } 1156 1157 /* report back real offset and size */ 1158 URL->offset = offset; 1159 URL->length = clength; 1160 1161 /* wrap it up in a FILE */ 1162 if ((f = http_funopen(conn, chunked)) == NULL) { 1163 fetch_syserr(); 1164 goto ouch; 1165 } 1166 1167 if (url != URL) 1168 fetchFreeURL(url); 1169 if (purl) 1170 fetchFreeURL(purl); 1171 1172 if (HTTP_ERROR(conn->err)) { 1173 http_print_html(stderr, f); 1174 fclose(f); 1175 f = NULL; 1176 } 1177 1178 return (f); 1179 1180 ouch: 1181 if (url != URL) 1182 fetchFreeURL(url); 1183 if (purl) 1184 fetchFreeURL(purl); 1185 if (conn != NULL) 1186 fetch_close(conn); 1187 return (NULL); 1188 } 1189 1190 1191 /***************************************************************************** 1192 * Entry points 1193 */ 1194 1195 /* 1196 * Retrieve and stat a file by HTTP 1197 */ 1198 FILE * 1199 fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags) 1200 { 1201 return (http_request(URL, "GET", us, http_get_proxy(URL, flags), flags)); 1202 } 1203 1204 /* 1205 * Retrieve a file by HTTP 1206 */ 1207 FILE * 1208 fetchGetHTTP(struct url *URL, const char *flags) 1209 { 1210 return (fetchXGetHTTP(URL, NULL, flags)); 1211 } 1212 1213 /* 1214 * Store a file by HTTP 1215 */ 1216 FILE * 1217 fetchPutHTTP(struct url *URL __unused, const char *flags __unused) 1218 { 1219 warnx("fetchPutHTTP(): not implemented"); 1220 return (NULL); 1221 } 1222 1223 /* 1224 * Get an HTTP document's metadata 1225 */ 1226 int 1227 fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags) 1228 { 1229 FILE *f; 1230 1231 f = http_request(URL, "HEAD", us, http_get_proxy(URL, flags), flags); 1232 if (f == NULL) 1233 return (-1); 1234 fclose(f); 1235 return (0); 1236 } 1237 1238 /* 1239 * List a directory 1240 */ 1241 struct url_ent * 1242 fetchListHTTP(struct url *url __unused, const char *flags __unused) 1243 { 1244 warnx("fetchListHTTP(): not implemented"); 1245 return (NULL); 1246 } 1247