1 /*- 2 * Copyright (c) 2000-2011 Dag-Erling Smørgrav 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 29 /* 30 * The following copyright applies to the base64 code: 31 * 32 *- 33 * Copyright 1997 Massachusetts Institute of Technology 34 * 35 * Permission to use, copy, modify, and distribute this software and 36 * its documentation for any purpose and without fee is hereby 37 * granted, provided that both the above copyright notice and this 38 * permission notice appear in all copies, that both the above 39 * copyright notice and this permission notice appear in all 40 * supporting documentation, and that the name of M.I.T. not be used 41 * in advertising or publicity pertaining to distribution of the 42 * software without specific, written prior permission. M.I.T. makes 43 * no representations about the suitability of this software for any 44 * purpose. It is provided "as is" without express or implied 45 * warranty. 46 * 47 * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''. M.I.T. DISCLAIMS 48 * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE, 49 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 50 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT 51 * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 52 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 53 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF 54 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND 55 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 56 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 57 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 */ 60 61 #include <sys/param.h> 62 #include <sys/socket.h> 63 #include <sys/time.h> 64 65 #include <ctype.h> 66 #include <err.h> 67 #include <errno.h> 68 #include <locale.h> 69 #include <netdb.h> 70 #include <stdarg.h> 71 #include <stdio.h> 72 #include <stdlib.h> 73 #include <string.h> 74 #include <time.h> 75 #include <unistd.h> 76 77 #ifdef WITH_SSL 78 #include <openssl/md5.h> 79 #define MD5Init(c) MD5_Init(c) 80 #define MD5Update(c, data, len) MD5_Update(c, data, len) 81 #define MD5Final(md, c) MD5_Final(md, c) 82 #else 83 #include <md5.h> 84 #endif 85 86 #include <netinet/in.h> 87 #include <netinet/tcp.h> 88 89 #include "fetch.h" 90 #include "common.h" 91 #include "httperr.h" 92 93 /* Maximum number of redirects to follow */ 94 #define MAX_REDIRECT 20 95 96 /* Symbolic names for reply codes we care about */ 97 #define HTTP_OK 200 98 #define HTTP_PARTIAL 206 99 #define HTTP_MOVED_PERM 301 100 #define HTTP_MOVED_TEMP 302 101 #define HTTP_SEE_OTHER 303 102 #define HTTP_NOT_MODIFIED 304 103 #define HTTP_USE_PROXY 305 104 #define HTTP_TEMP_REDIRECT 307 105 #define HTTP_PERM_REDIRECT 308 106 #define HTTP_NEED_AUTH 401 107 #define HTTP_NEED_PROXY_AUTH 407 108 #define HTTP_BAD_RANGE 416 109 #define HTTP_PROTOCOL_ERROR 999 110 111 #define HTTP_REDIRECT(xyz) ((xyz) == HTTP_MOVED_PERM \ 112 || (xyz) == HTTP_MOVED_TEMP \ 113 || (xyz) == HTTP_TEMP_REDIRECT \ 114 || (xyz) == HTTP_USE_PROXY \ 115 || (xyz) == HTTP_SEE_OTHER) 116 117 #define HTTP_ERROR(xyz) ((xyz) > 400 && (xyz) < 599) 118 119 120 /***************************************************************************** 121 * I/O functions for decoding chunked streams 122 */ 123 124 struct httpio 125 { 126 conn_t *conn; /* connection */ 127 int chunked; /* chunked mode */ 128 char *buf; /* chunk buffer */ 129 size_t bufsize; /* size of chunk buffer */ 130 ssize_t buflen; /* amount of data currently in buffer */ 131 int bufpos; /* current read offset in buffer */ 132 int eof; /* end-of-file flag */ 133 int error; /* error flag */ 134 size_t chunksize; /* remaining size of current chunk */ 135 #ifndef NDEBUG 136 size_t total; 137 #endif 138 }; 139 140 static int http_cmd(conn_t *, const char *, ...) __printflike(2, 3); 141 142 /* 143 * Get next chunk header 144 */ 145 static int 146 http_new_chunk(struct httpio *io) 147 { 148 char *p; 149 150 if (fetch_getln(io->conn) == -1) 151 return (-1); 152 153 if (io->conn->buflen < 2 || !isxdigit((unsigned char)*io->conn->buf)) 154 return (-1); 155 156 for (p = io->conn->buf; *p && !isspace((unsigned char)*p); ++p) { 157 if (*p == ';') 158 break; 159 if (!isxdigit((unsigned char)*p)) 160 return (-1); 161 if (isdigit((unsigned char)*p)) { 162 io->chunksize = io->chunksize * 16 + 163 *p - '0'; 164 } else { 165 io->chunksize = io->chunksize * 16 + 166 10 + tolower((unsigned char)*p) - 'a'; 167 } 168 } 169 170 #ifndef NDEBUG 171 if (fetchDebug) { 172 io->total += io->chunksize; 173 if (io->chunksize == 0) 174 fprintf(stderr, "%s(): end of last chunk\n", __func__); 175 else 176 fprintf(stderr, "%s(): new chunk: %lu (%lu)\n", 177 __func__, (unsigned long)io->chunksize, 178 (unsigned long)io->total); 179 } 180 #endif 181 182 return (io->chunksize); 183 } 184 185 /* 186 * Grow the input buffer to at least len bytes 187 */ 188 static inline int 189 http_growbuf(struct httpio *io, size_t len) 190 { 191 char *tmp; 192 193 if (io->bufsize >= len) 194 return (0); 195 196 if ((tmp = realloc(io->buf, len)) == NULL) 197 return (-1); 198 io->buf = tmp; 199 io->bufsize = len; 200 return (0); 201 } 202 203 /* 204 * Fill the input buffer, do chunk decoding on the fly 205 */ 206 static int 207 http_fillbuf(struct httpio *io, size_t len) 208 { 209 ssize_t nbytes; 210 211 if (io->error) 212 return (-1); 213 if (io->eof) 214 return (0); 215 216 if (io->chunked == 0) { 217 if (http_growbuf(io, len) == -1) 218 return (-1); 219 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) { 220 io->error = errno; 221 return (-1); 222 } 223 io->buflen = nbytes; 224 io->bufpos = 0; 225 return (io->buflen); 226 } 227 228 if (io->chunksize == 0) { 229 switch (http_new_chunk(io)) { 230 case -1: 231 io->error = 1; 232 return (-1); 233 case 0: 234 io->eof = 1; 235 return (0); 236 } 237 } 238 239 if (len > io->chunksize) 240 len = io->chunksize; 241 if (http_growbuf(io, len) == -1) 242 return (-1); 243 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) { 244 io->error = errno; 245 return (-1); 246 } 247 io->buflen = nbytes; 248 io->chunksize -= io->buflen; 249 250 if (io->chunksize == 0) { 251 char endl[2]; 252 253 if (fetch_read(io->conn, endl, 2) != 2 || 254 endl[0] != '\r' || endl[1] != '\n') 255 return (-1); 256 } 257 258 io->bufpos = 0; 259 260 return (io->buflen); 261 } 262 263 /* 264 * Read function 265 */ 266 static int 267 http_readfn(void *v, char *buf, int len) 268 { 269 struct httpio *io = (struct httpio *)v; 270 int l, pos; 271 272 if (io->error) 273 return (-1); 274 if (io->eof) 275 return (0); 276 277 for (pos = 0; len > 0; pos += l, len -= l) { 278 /* empty buffer */ 279 if (!io->buf || io->bufpos == io->buflen) 280 if (http_fillbuf(io, len) < 1) 281 break; 282 l = io->buflen - io->bufpos; 283 if (len < l) 284 l = len; 285 memcpy(buf + pos, io->buf + io->bufpos, l); 286 io->bufpos += l; 287 } 288 289 if (!pos && io->error) { 290 if (io->error == EINTR) 291 io->error = 0; 292 return (-1); 293 } 294 return (pos); 295 } 296 297 /* 298 * Write function 299 */ 300 static int 301 http_writefn(void *v, const char *buf, int len) 302 { 303 struct httpio *io = (struct httpio *)v; 304 305 return (fetch_write(io->conn, buf, len)); 306 } 307 308 /* 309 * Close function 310 */ 311 static int 312 http_closefn(void *v) 313 { 314 struct httpio *io = (struct httpio *)v; 315 int r; 316 317 r = fetch_close(io->conn); 318 if (io->buf) 319 free(io->buf); 320 free(io); 321 return (r); 322 } 323 324 /* 325 * Wrap a file descriptor up 326 */ 327 static FILE * 328 http_funopen(conn_t *conn, int chunked) 329 { 330 struct httpio *io; 331 FILE *f; 332 333 if ((io = calloc(1, sizeof(*io))) == NULL) { 334 fetch_syserr(); 335 return (NULL); 336 } 337 io->conn = conn; 338 io->chunked = chunked; 339 f = funopen(io, http_readfn, http_writefn, NULL, http_closefn); 340 if (f == NULL) { 341 fetch_syserr(); 342 free(io); 343 return (NULL); 344 } 345 return (f); 346 } 347 348 349 /***************************************************************************** 350 * Helper functions for talking to the server and parsing its replies 351 */ 352 353 /* Header types */ 354 typedef enum { 355 hdr_syserror = -2, 356 hdr_error = -1, 357 hdr_end = 0, 358 hdr_unknown = 1, 359 hdr_content_length, 360 hdr_content_range, 361 hdr_last_modified, 362 hdr_location, 363 hdr_transfer_encoding, 364 hdr_www_authenticate, 365 hdr_proxy_authenticate, 366 } hdr_t; 367 368 /* Names of interesting headers */ 369 static struct { 370 hdr_t num; 371 const char *name; 372 } hdr_names[] = { 373 { hdr_content_length, "Content-Length" }, 374 { hdr_content_range, "Content-Range" }, 375 { hdr_last_modified, "Last-Modified" }, 376 { hdr_location, "Location" }, 377 { hdr_transfer_encoding, "Transfer-Encoding" }, 378 { hdr_www_authenticate, "WWW-Authenticate" }, 379 { hdr_proxy_authenticate, "Proxy-Authenticate" }, 380 { hdr_unknown, NULL }, 381 }; 382 383 /* 384 * Send a formatted line; optionally echo to terminal 385 */ 386 static int 387 http_cmd(conn_t *conn, const char *fmt, ...) 388 { 389 va_list ap; 390 size_t len; 391 char *msg; 392 int r; 393 394 va_start(ap, fmt); 395 len = vasprintf(&msg, fmt, ap); 396 va_end(ap); 397 398 if (msg == NULL) { 399 errno = ENOMEM; 400 fetch_syserr(); 401 return (-1); 402 } 403 404 r = fetch_putln(conn, msg, len); 405 free(msg); 406 407 if (r == -1) { 408 fetch_syserr(); 409 return (-1); 410 } 411 412 return (0); 413 } 414 415 /* 416 * Get and parse status line 417 */ 418 static int 419 http_get_reply(conn_t *conn) 420 { 421 char *p; 422 423 if (fetch_getln(conn) == -1) 424 return (-1); 425 /* 426 * A valid status line looks like "HTTP/m.n xyz reason" where m 427 * and n are the major and minor protocol version numbers and xyz 428 * is the reply code. 429 * Unfortunately, there are servers out there (NCSA 1.5.1, to name 430 * just one) that do not send a version number, so we can't rely 431 * on finding one, but if we do, insist on it being 1.0 or 1.1. 432 * We don't care about the reason phrase. 433 */ 434 if (strncmp(conn->buf, "HTTP", 4) != 0) 435 return (HTTP_PROTOCOL_ERROR); 436 p = conn->buf + 4; 437 if (*p == '/') { 438 if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1')) 439 return (HTTP_PROTOCOL_ERROR); 440 p += 4; 441 } 442 if (*p != ' ' || 443 !isdigit((unsigned char)p[1]) || 444 !isdigit((unsigned char)p[2]) || 445 !isdigit((unsigned char)p[3])) 446 return (HTTP_PROTOCOL_ERROR); 447 448 conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0'); 449 return (conn->err); 450 } 451 452 /* 453 * Check a header; if the type matches the given string, return a pointer 454 * to the beginning of the value. 455 */ 456 static const char * 457 http_match(const char *str, const char *hdr) 458 { 459 while (*str && *hdr && 460 tolower((unsigned char)*str++) == tolower((unsigned char)*hdr++)) 461 /* nothing */; 462 if (*str || *hdr != ':') 463 return (NULL); 464 while (*hdr && isspace((unsigned char)*++hdr)) 465 /* nothing */; 466 return (hdr); 467 } 468 469 470 /* 471 * Get the next header and return the appropriate symbolic code. We 472 * need to read one line ahead for checking for a continuation line 473 * belonging to the current header (continuation lines start with 474 * white space). 475 * 476 * We get called with a fresh line already in the conn buffer, either 477 * from the previous http_next_header() invocation, or, the first 478 * time, from a fetch_getln() performed by our caller. 479 * 480 * This stops when we encounter an empty line (we dont read beyond the header 481 * area). 482 * 483 * Note that the "headerbuf" is just a place to return the result. Its 484 * contents are not used for the next call. This means that no cleanup 485 * is needed when ie doing another connection, just call the cleanup when 486 * fully done to deallocate memory. 487 */ 488 489 /* Limit the max number of continuation lines to some reasonable value */ 490 #define HTTP_MAX_CONT_LINES 10 491 492 /* Place into which to build a header from one or several lines */ 493 typedef struct { 494 char *buf; /* buffer */ 495 size_t bufsize; /* buffer size */ 496 size_t buflen; /* length of buffer contents */ 497 } http_headerbuf_t; 498 499 static void 500 init_http_headerbuf(http_headerbuf_t *buf) 501 { 502 buf->buf = NULL; 503 buf->bufsize = 0; 504 buf->buflen = 0; 505 } 506 507 static void 508 clean_http_headerbuf(http_headerbuf_t *buf) 509 { 510 if (buf->buf) 511 free(buf->buf); 512 init_http_headerbuf(buf); 513 } 514 515 /* Remove whitespace at the end of the buffer */ 516 static void 517 http_conn_trimright(conn_t *conn) 518 { 519 while (conn->buflen && 520 isspace((unsigned char)conn->buf[conn->buflen - 1])) 521 conn->buflen--; 522 conn->buf[conn->buflen] = '\0'; 523 } 524 525 static hdr_t 526 http_next_header(conn_t *conn, http_headerbuf_t *hbuf, const char **p) 527 { 528 unsigned int i, len; 529 530 /* 531 * Have to do the stripping here because of the first line. So 532 * it's done twice for the subsequent lines. No big deal 533 */ 534 http_conn_trimright(conn); 535 if (conn->buflen == 0) 536 return (hdr_end); 537 538 /* Copy the line to the headerbuf */ 539 if (hbuf->bufsize < conn->buflen + 1) { 540 if ((hbuf->buf = realloc(hbuf->buf, conn->buflen + 1)) == NULL) 541 return (hdr_syserror); 542 hbuf->bufsize = conn->buflen + 1; 543 } 544 strcpy(hbuf->buf, conn->buf); 545 hbuf->buflen = conn->buflen; 546 547 /* 548 * Fetch possible continuation lines. Stop at 1st non-continuation 549 * and leave it in the conn buffer 550 */ 551 for (i = 0; i < HTTP_MAX_CONT_LINES; i++) { 552 if (fetch_getln(conn) == -1) 553 return (hdr_syserror); 554 555 /* 556 * Note: we carry on the idea from the previous version 557 * that a pure whitespace line is equivalent to an empty 558 * one (so it's not continuation and will be handled when 559 * we are called next) 560 */ 561 http_conn_trimright(conn); 562 if (conn->buf[0] != ' ' && conn->buf[0] != "\t"[0]) 563 break; 564 565 /* Got a continuation line. Concatenate to previous */ 566 len = hbuf->buflen + conn->buflen; 567 if (hbuf->bufsize < len + 1) { 568 len *= 2; 569 if ((hbuf->buf = realloc(hbuf->buf, len + 1)) == NULL) 570 return (hdr_syserror); 571 hbuf->bufsize = len + 1; 572 } 573 strcpy(hbuf->buf + hbuf->buflen, conn->buf); 574 hbuf->buflen += conn->buflen; 575 } 576 577 /* 578 * We could check for malformed headers but we don't really care. 579 * A valid header starts with a token immediately followed by a 580 * colon; a token is any sequence of non-control, non-whitespace 581 * characters except "()<>@,;:\\\"{}". 582 */ 583 for (i = 0; hdr_names[i].num != hdr_unknown; i++) 584 if ((*p = http_match(hdr_names[i].name, hbuf->buf)) != NULL) 585 return (hdr_names[i].num); 586 587 return (hdr_unknown); 588 } 589 590 /************************** 591 * [Proxy-]Authenticate header parsing 592 */ 593 594 /* 595 * Read doublequote-delimited string into output buffer obuf (allocated 596 * by caller, whose responsibility it is to ensure that it's big enough) 597 * cp points to the first char after the initial '"' 598 * Handles \ quoting 599 * Returns pointer to the first char after the terminating double quote, or 600 * NULL for error. 601 */ 602 static const char * 603 http_parse_headerstring(const char *cp, char *obuf) 604 { 605 for (;;) { 606 switch (*cp) { 607 case 0: /* Unterminated string */ 608 *obuf = 0; 609 return (NULL); 610 case '"': /* Ending quote */ 611 *obuf = 0; 612 return (++cp); 613 case '\\': 614 if (*++cp == 0) { 615 *obuf = 0; 616 return (NULL); 617 } 618 /* FALLTHROUGH */ 619 default: 620 *obuf++ = *cp++; 621 } 622 } 623 } 624 625 /* Http auth challenge schemes */ 626 typedef enum {HTTPAS_UNKNOWN, HTTPAS_BASIC,HTTPAS_DIGEST} http_auth_schemes_t; 627 628 /* Data holder for a Basic or Digest challenge. */ 629 typedef struct { 630 http_auth_schemes_t scheme; 631 char *realm; 632 char *qop; 633 char *nonce; 634 char *opaque; 635 char *algo; 636 int stale; 637 int nc; /* Nonce count */ 638 } http_auth_challenge_t; 639 640 static void 641 init_http_auth_challenge(http_auth_challenge_t *b) 642 { 643 b->scheme = HTTPAS_UNKNOWN; 644 b->realm = b->qop = b->nonce = b->opaque = b->algo = NULL; 645 b->stale = b->nc = 0; 646 } 647 648 static void 649 clean_http_auth_challenge(http_auth_challenge_t *b) 650 { 651 if (b->realm) 652 free(b->realm); 653 if (b->qop) 654 free(b->qop); 655 if (b->nonce) 656 free(b->nonce); 657 if (b->opaque) 658 free(b->opaque); 659 if (b->algo) 660 free(b->algo); 661 init_http_auth_challenge(b); 662 } 663 664 /* Data holder for an array of challenges offered in an http response. */ 665 #define MAX_CHALLENGES 10 666 typedef struct { 667 http_auth_challenge_t *challenges[MAX_CHALLENGES]; 668 int count; /* Number of parsed challenges in the array */ 669 int valid; /* We did parse an authenticate header */ 670 } http_auth_challenges_t; 671 672 static void 673 init_http_auth_challenges(http_auth_challenges_t *cs) 674 { 675 int i; 676 for (i = 0; i < MAX_CHALLENGES; i++) 677 cs->challenges[i] = NULL; 678 cs->count = cs->valid = 0; 679 } 680 681 static void 682 clean_http_auth_challenges(http_auth_challenges_t *cs) 683 { 684 int i; 685 /* We rely on non-zero pointers being allocated, not on the count */ 686 for (i = 0; i < MAX_CHALLENGES; i++) { 687 if (cs->challenges[i] != NULL) { 688 clean_http_auth_challenge(cs->challenges[i]); 689 free(cs->challenges[i]); 690 } 691 } 692 init_http_auth_challenges(cs); 693 } 694 695 /* 696 * Enumeration for lexical elements. Separators will be returned as their own 697 * ascii value 698 */ 699 typedef enum {HTTPHL_WORD=256, HTTPHL_STRING=257, HTTPHL_END=258, 700 HTTPHL_ERROR = 259} http_header_lex_t; 701 702 /* 703 * Determine what kind of token comes next and return possible value 704 * in buf, which is supposed to have been allocated big enough by 705 * caller. Advance input pointer and return element type. 706 */ 707 static int 708 http_header_lex(const char **cpp, char *buf) 709 { 710 size_t l; 711 /* Eat initial whitespace */ 712 *cpp += strspn(*cpp, " \t"); 713 if (**cpp == 0) 714 return (HTTPHL_END); 715 716 /* Separator ? */ 717 if (**cpp == ',' || **cpp == '=') 718 return (*((*cpp)++)); 719 720 /* String ? */ 721 if (**cpp == '"') { 722 *cpp = http_parse_headerstring(++*cpp, buf); 723 if (*cpp == NULL) 724 return (HTTPHL_ERROR); 725 return (HTTPHL_STRING); 726 } 727 728 /* Read other token, until separator or whitespace */ 729 l = strcspn(*cpp, " \t,="); 730 memcpy(buf, *cpp, l); 731 buf[l] = 0; 732 *cpp += l; 733 return (HTTPHL_WORD); 734 } 735 736 /* 737 * Read challenges from http xxx-authenticate header and accumulate them 738 * in the challenges list structure. 739 * 740 * Headers with multiple challenges are specified by rfc2617, but 741 * servers (ie: squid) often send them in separate headers instead, 742 * which in turn is forbidden by the http spec (multiple headers with 743 * the same name are only allowed for pure comma-separated lists, see 744 * rfc2616 sec 4.2). 745 * 746 * We support both approaches anyway 747 */ 748 static int 749 http_parse_authenticate(const char *cp, http_auth_challenges_t *cs) 750 { 751 int ret = -1; 752 http_header_lex_t lex; 753 char *key = malloc(strlen(cp) + 1); 754 char *value = malloc(strlen(cp) + 1); 755 char *buf = malloc(strlen(cp) + 1); 756 757 if (key == NULL || value == NULL || buf == NULL) { 758 fetch_syserr(); 759 goto out; 760 } 761 762 /* In any case we've seen the header and we set the valid bit */ 763 cs->valid = 1; 764 765 /* Need word first */ 766 lex = http_header_lex(&cp, key); 767 if (lex != HTTPHL_WORD) 768 goto out; 769 770 /* Loop on challenges */ 771 for (; cs->count < MAX_CHALLENGES; cs->count++) { 772 cs->challenges[cs->count] = 773 malloc(sizeof(http_auth_challenge_t)); 774 if (cs->challenges[cs->count] == NULL) { 775 fetch_syserr(); 776 goto out; 777 } 778 init_http_auth_challenge(cs->challenges[cs->count]); 779 if (!strcasecmp(key, "basic")) { 780 cs->challenges[cs->count]->scheme = HTTPAS_BASIC; 781 } else if (!strcasecmp(key, "digest")) { 782 cs->challenges[cs->count]->scheme = HTTPAS_DIGEST; 783 } else { 784 cs->challenges[cs->count]->scheme = HTTPAS_UNKNOWN; 785 /* 786 * Continue parsing as basic or digest may 787 * follow, and the syntax is the same for 788 * all. We'll just ignore this one when 789 * looking at the list 790 */ 791 } 792 793 /* Loop on attributes */ 794 for (;;) { 795 /* Key */ 796 lex = http_header_lex(&cp, key); 797 if (lex != HTTPHL_WORD) 798 goto out; 799 800 /* Equal sign */ 801 lex = http_header_lex(&cp, buf); 802 if (lex != '=') 803 goto out; 804 805 /* Value */ 806 lex = http_header_lex(&cp, value); 807 if (lex != HTTPHL_WORD && lex != HTTPHL_STRING) 808 goto out; 809 810 if (!strcasecmp(key, "realm")) 811 cs->challenges[cs->count]->realm = 812 strdup(value); 813 else if (!strcasecmp(key, "qop")) 814 cs->challenges[cs->count]->qop = 815 strdup(value); 816 else if (!strcasecmp(key, "nonce")) 817 cs->challenges[cs->count]->nonce = 818 strdup(value); 819 else if (!strcasecmp(key, "opaque")) 820 cs->challenges[cs->count]->opaque = 821 strdup(value); 822 else if (!strcasecmp(key, "algorithm")) 823 cs->challenges[cs->count]->algo = 824 strdup(value); 825 else if (!strcasecmp(key, "stale")) 826 cs->challenges[cs->count]->stale = 827 strcasecmp(value, "no"); 828 /* Else ignore unknown attributes */ 829 830 /* Comma or Next challenge or End */ 831 lex = http_header_lex(&cp, key); 832 /* 833 * If we get a word here, this is the beginning of the 834 * next challenge. Break the attributes loop 835 */ 836 if (lex == HTTPHL_WORD) 837 break; 838 839 if (lex == HTTPHL_END) { 840 /* End while looking for ',' is normal exit */ 841 cs->count++; 842 ret = 0; 843 goto out; 844 } 845 /* Anything else is an error */ 846 if (lex != ',') 847 goto out; 848 849 } /* End attributes loop */ 850 } /* End challenge loop */ 851 852 /* 853 * Challenges max count exceeded. This really can't happen 854 * with normal data, something's fishy -> error 855 */ 856 857 out: 858 if (key) 859 free(key); 860 if (value) 861 free(value); 862 if (buf) 863 free(buf); 864 return (ret); 865 } 866 867 868 /* 869 * Parse a last-modified header 870 */ 871 static int 872 http_parse_mtime(const char *p, time_t *mtime) 873 { 874 char locale[64], *r; 875 struct tm tm; 876 877 strncpy(locale, setlocale(LC_TIME, NULL), sizeof(locale)); 878 setlocale(LC_TIME, "C"); 879 r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm); 880 /* XXX should add support for date-2 and date-3 */ 881 setlocale(LC_TIME, locale); 882 if (r == NULL) 883 return (-1); 884 DEBUG(fprintf(stderr, "last modified: [%04d-%02d-%02d " 885 "%02d:%02d:%02d]\n", 886 tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday, 887 tm.tm_hour, tm.tm_min, tm.tm_sec)); 888 *mtime = timegm(&tm); 889 return (0); 890 } 891 892 /* 893 * Parse a content-length header 894 */ 895 static int 896 http_parse_length(const char *p, off_t *length) 897 { 898 off_t len; 899 900 for (len = 0; *p && isdigit((unsigned char)*p); ++p) 901 len = len * 10 + (*p - '0'); 902 if (*p) 903 return (-1); 904 DEBUG(fprintf(stderr, "content length: [%lld]\n", 905 (long long)len)); 906 *length = len; 907 return (0); 908 } 909 910 /* 911 * Parse a content-range header 912 */ 913 static int 914 http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size) 915 { 916 off_t first, last, len; 917 918 if (strncasecmp(p, "bytes ", 6) != 0) 919 return (-1); 920 p += 6; 921 if (*p == '*') { 922 first = last = -1; 923 ++p; 924 } else { 925 for (first = 0; *p && isdigit((unsigned char)*p); ++p) 926 first = first * 10 + *p - '0'; 927 if (*p != '-') 928 return (-1); 929 for (last = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 930 last = last * 10 + *p - '0'; 931 } 932 if (first > last || *p != '/') 933 return (-1); 934 for (len = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 935 len = len * 10 + *p - '0'; 936 if (*p || len < last - first + 1) 937 return (-1); 938 if (first == -1) { 939 DEBUG(fprintf(stderr, "content range: [*/%lld]\n", 940 (long long)len)); 941 *length = 0; 942 } else { 943 DEBUG(fprintf(stderr, "content range: [%lld-%lld/%lld]\n", 944 (long long)first, (long long)last, (long long)len)); 945 *length = last - first + 1; 946 } 947 *offset = first; 948 *size = len; 949 return (0); 950 } 951 952 953 /***************************************************************************** 954 * Helper functions for authorization 955 */ 956 957 /* 958 * Base64 encoding 959 */ 960 static char * 961 http_base64(const char *src) 962 { 963 static const char base64[] = 964 "ABCDEFGHIJKLMNOPQRSTUVWXYZ" 965 "abcdefghijklmnopqrstuvwxyz" 966 "0123456789+/"; 967 char *str, *dst; 968 size_t l; 969 int t, r; 970 971 l = strlen(src); 972 if ((str = malloc(((l + 2) / 3) * 4 + 1)) == NULL) 973 return (NULL); 974 dst = str; 975 r = 0; 976 977 while (l >= 3) { 978 t = (src[0] << 16) | (src[1] << 8) | src[2]; 979 dst[0] = base64[(t >> 18) & 0x3f]; 980 dst[1] = base64[(t >> 12) & 0x3f]; 981 dst[2] = base64[(t >> 6) & 0x3f]; 982 dst[3] = base64[(t >> 0) & 0x3f]; 983 src += 3; l -= 3; 984 dst += 4; r += 4; 985 } 986 987 switch (l) { 988 case 2: 989 t = (src[0] << 16) | (src[1] << 8); 990 dst[0] = base64[(t >> 18) & 0x3f]; 991 dst[1] = base64[(t >> 12) & 0x3f]; 992 dst[2] = base64[(t >> 6) & 0x3f]; 993 dst[3] = '='; 994 dst += 4; 995 r += 4; 996 break; 997 case 1: 998 t = src[0] << 16; 999 dst[0] = base64[(t >> 18) & 0x3f]; 1000 dst[1] = base64[(t >> 12) & 0x3f]; 1001 dst[2] = dst[3] = '='; 1002 dst += 4; 1003 r += 4; 1004 break; 1005 case 0: 1006 break; 1007 } 1008 1009 *dst = 0; 1010 return (str); 1011 } 1012 1013 1014 /* 1015 * Extract authorization parameters from environment value. 1016 * The value is like scheme:realm:user:pass 1017 */ 1018 typedef struct { 1019 char *scheme; 1020 char *realm; 1021 char *user; 1022 char *password; 1023 } http_auth_params_t; 1024 1025 static void 1026 init_http_auth_params(http_auth_params_t *s) 1027 { 1028 s->scheme = s->realm = s->user = s->password = NULL; 1029 } 1030 1031 static void 1032 clean_http_auth_params(http_auth_params_t *s) 1033 { 1034 if (s->scheme) 1035 free(s->scheme); 1036 if (s->realm) 1037 free(s->realm); 1038 if (s->user) 1039 free(s->user); 1040 if (s->password) 1041 free(s->password); 1042 init_http_auth_params(s); 1043 } 1044 1045 static int 1046 http_authfromenv(const char *p, http_auth_params_t *parms) 1047 { 1048 int ret = -1; 1049 char *v, *ve; 1050 char *str = strdup(p); 1051 1052 if (str == NULL) { 1053 fetch_syserr(); 1054 return (-1); 1055 } 1056 v = str; 1057 1058 if ((ve = strchr(v, ':')) == NULL) 1059 goto out; 1060 1061 *ve = 0; 1062 if ((parms->scheme = strdup(v)) == NULL) { 1063 fetch_syserr(); 1064 goto out; 1065 } 1066 v = ve + 1; 1067 1068 if ((ve = strchr(v, ':')) == NULL) 1069 goto out; 1070 1071 *ve = 0; 1072 if ((parms->realm = strdup(v)) == NULL) { 1073 fetch_syserr(); 1074 goto out; 1075 } 1076 v = ve + 1; 1077 1078 if ((ve = strchr(v, ':')) == NULL) 1079 goto out; 1080 1081 *ve = 0; 1082 if ((parms->user = strdup(v)) == NULL) { 1083 fetch_syserr(); 1084 goto out; 1085 } 1086 v = ve + 1; 1087 1088 1089 if ((parms->password = strdup(v)) == NULL) { 1090 fetch_syserr(); 1091 goto out; 1092 } 1093 ret = 0; 1094 out: 1095 if (ret == -1) 1096 clean_http_auth_params(parms); 1097 if (str) 1098 free(str); 1099 return (ret); 1100 } 1101 1102 1103 /* 1104 * Digest response: the code to compute the digest is taken from the 1105 * sample implementation in RFC2616 1106 */ 1107 #define IN const 1108 #define OUT 1109 1110 #define HASHLEN 16 1111 typedef char HASH[HASHLEN]; 1112 #define HASHHEXLEN 32 1113 typedef char HASHHEX[HASHHEXLEN+1]; 1114 1115 static const char *hexchars = "0123456789abcdef"; 1116 static void 1117 CvtHex(IN HASH Bin, OUT HASHHEX Hex) 1118 { 1119 unsigned short i; 1120 unsigned char j; 1121 1122 for (i = 0; i < HASHLEN; i++) { 1123 j = (Bin[i] >> 4) & 0xf; 1124 Hex[i*2] = hexchars[j]; 1125 j = Bin[i] & 0xf; 1126 Hex[i*2+1] = hexchars[j]; 1127 } 1128 Hex[HASHHEXLEN] = '\0'; 1129 }; 1130 1131 /* calculate H(A1) as per spec */ 1132 static void 1133 DigestCalcHA1( 1134 IN char * pszAlg, 1135 IN char * pszUserName, 1136 IN char * pszRealm, 1137 IN char * pszPassword, 1138 IN char * pszNonce, 1139 IN char * pszCNonce, 1140 OUT HASHHEX SessionKey 1141 ) 1142 { 1143 MD5_CTX Md5Ctx; 1144 HASH HA1; 1145 1146 MD5Init(&Md5Ctx); 1147 MD5Update(&Md5Ctx, pszUserName, strlen(pszUserName)); 1148 MD5Update(&Md5Ctx, ":", 1); 1149 MD5Update(&Md5Ctx, pszRealm, strlen(pszRealm)); 1150 MD5Update(&Md5Ctx, ":", 1); 1151 MD5Update(&Md5Ctx, pszPassword, strlen(pszPassword)); 1152 MD5Final(HA1, &Md5Ctx); 1153 if (strcasecmp(pszAlg, "md5-sess") == 0) { 1154 1155 MD5Init(&Md5Ctx); 1156 MD5Update(&Md5Ctx, HA1, HASHLEN); 1157 MD5Update(&Md5Ctx, ":", 1); 1158 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce)); 1159 MD5Update(&Md5Ctx, ":", 1); 1160 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce)); 1161 MD5Final(HA1, &Md5Ctx); 1162 } 1163 CvtHex(HA1, SessionKey); 1164 } 1165 1166 /* calculate request-digest/response-digest as per HTTP Digest spec */ 1167 static void 1168 DigestCalcResponse( 1169 IN HASHHEX HA1, /* H(A1) */ 1170 IN char * pszNonce, /* nonce from server */ 1171 IN char * pszNonceCount, /* 8 hex digits */ 1172 IN char * pszCNonce, /* client nonce */ 1173 IN char * pszQop, /* qop-value: "", "auth", "auth-int" */ 1174 IN char * pszMethod, /* method from the request */ 1175 IN char * pszDigestUri, /* requested URL */ 1176 IN HASHHEX HEntity, /* H(entity body) if qop="auth-int" */ 1177 OUT HASHHEX Response /* request-digest or response-digest */ 1178 ) 1179 { 1180 /* DEBUG(fprintf(stderr, 1181 "Calc: HA1[%s] Nonce[%s] qop[%s] method[%s] URI[%s]\n", 1182 HA1, pszNonce, pszQop, pszMethod, pszDigestUri));*/ 1183 MD5_CTX Md5Ctx; 1184 HASH HA2; 1185 HASH RespHash; 1186 HASHHEX HA2Hex; 1187 1188 // calculate H(A2) 1189 MD5Init(&Md5Ctx); 1190 MD5Update(&Md5Ctx, pszMethod, strlen(pszMethod)); 1191 MD5Update(&Md5Ctx, ":", 1); 1192 MD5Update(&Md5Ctx, pszDigestUri, strlen(pszDigestUri)); 1193 if (strcasecmp(pszQop, "auth-int") == 0) { 1194 MD5Update(&Md5Ctx, ":", 1); 1195 MD5Update(&Md5Ctx, HEntity, HASHHEXLEN); 1196 } 1197 MD5Final(HA2, &Md5Ctx); 1198 CvtHex(HA2, HA2Hex); 1199 1200 // calculate response 1201 MD5Init(&Md5Ctx); 1202 MD5Update(&Md5Ctx, HA1, HASHHEXLEN); 1203 MD5Update(&Md5Ctx, ":", 1); 1204 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce)); 1205 MD5Update(&Md5Ctx, ":", 1); 1206 if (*pszQop) { 1207 MD5Update(&Md5Ctx, pszNonceCount, strlen(pszNonceCount)); 1208 MD5Update(&Md5Ctx, ":", 1); 1209 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce)); 1210 MD5Update(&Md5Ctx, ":", 1); 1211 MD5Update(&Md5Ctx, pszQop, strlen(pszQop)); 1212 MD5Update(&Md5Ctx, ":", 1); 1213 } 1214 MD5Update(&Md5Ctx, HA2Hex, HASHHEXLEN); 1215 MD5Final(RespHash, &Md5Ctx); 1216 CvtHex(RespHash, Response); 1217 } 1218 1219 /* 1220 * Generate/Send a Digest authorization header 1221 * This looks like: [Proxy-]Authorization: credentials 1222 * 1223 * credentials = "Digest" digest-response 1224 * digest-response = 1#( username | realm | nonce | digest-uri 1225 * | response | [ algorithm ] | [cnonce] | 1226 * [opaque] | [message-qop] | 1227 * [nonce-count] | [auth-param] ) 1228 * username = "username" "=" username-value 1229 * username-value = quoted-string 1230 * digest-uri = "uri" "=" digest-uri-value 1231 * digest-uri-value = request-uri ; As specified by HTTP/1.1 1232 * message-qop = "qop" "=" qop-value 1233 * cnonce = "cnonce" "=" cnonce-value 1234 * cnonce-value = nonce-value 1235 * nonce-count = "nc" "=" nc-value 1236 * nc-value = 8LHEX 1237 * response = "response" "=" request-digest 1238 * request-digest = <"> 32LHEX <"> 1239 */ 1240 static int 1241 http_digest_auth(conn_t *conn, const char *hdr, http_auth_challenge_t *c, 1242 http_auth_params_t *parms, struct url *url) 1243 { 1244 int r; 1245 char noncecount[10]; 1246 char cnonce[40]; 1247 char *options = NULL; 1248 1249 if (!c->realm || !c->nonce) { 1250 DEBUG(fprintf(stderr, "realm/nonce not set in challenge\n")); 1251 return(-1); 1252 } 1253 if (!c->algo) 1254 c->algo = strdup(""); 1255 1256 if (asprintf(&options, "%s%s%s%s", 1257 *c->algo? ",algorithm=" : "", c->algo, 1258 c->opaque? ",opaque=" : "", c->opaque?c->opaque:"")== -1) 1259 return (-1); 1260 1261 if (!c->qop) { 1262 c->qop = strdup(""); 1263 *noncecount = 0; 1264 *cnonce = 0; 1265 } else { 1266 c->nc++; 1267 sprintf(noncecount, "%08x", c->nc); 1268 /* We don't try very hard with the cnonce ... */ 1269 sprintf(cnonce, "%x%lx", getpid(), (unsigned long)time(0)); 1270 } 1271 1272 HASHHEX HA1; 1273 DigestCalcHA1(c->algo, parms->user, c->realm, 1274 parms->password, c->nonce, cnonce, HA1); 1275 DEBUG(fprintf(stderr, "HA1: [%s]\n", HA1)); 1276 HASHHEX digest; 1277 DigestCalcResponse(HA1, c->nonce, noncecount, cnonce, c->qop, 1278 "GET", url->doc, "", digest); 1279 1280 if (c->qop[0]) { 1281 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\"," 1282 "nonce=\"%s\",uri=\"%s\",response=\"%s\"," 1283 "qop=\"auth\", cnonce=\"%s\", nc=%s%s", 1284 hdr, parms->user, c->realm, 1285 c->nonce, url->doc, digest, 1286 cnonce, noncecount, options); 1287 } else { 1288 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\"," 1289 "nonce=\"%s\",uri=\"%s\",response=\"%s\"%s", 1290 hdr, parms->user, c->realm, 1291 c->nonce, url->doc, digest, options); 1292 } 1293 if (options) 1294 free(options); 1295 return (r); 1296 } 1297 1298 /* 1299 * Encode username and password 1300 */ 1301 static int 1302 http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd) 1303 { 1304 char *upw, *auth; 1305 int r; 1306 1307 DEBUG(fprintf(stderr, "basic: usr: [%s]\n", usr)); 1308 DEBUG(fprintf(stderr, "basic: pwd: [%s]\n", pwd)); 1309 if (asprintf(&upw, "%s:%s", usr, pwd) == -1) 1310 return (-1); 1311 auth = http_base64(upw); 1312 free(upw); 1313 if (auth == NULL) 1314 return (-1); 1315 r = http_cmd(conn, "%s: Basic %s", hdr, auth); 1316 free(auth); 1317 return (r); 1318 } 1319 1320 /* 1321 * Chose the challenge to answer and call the appropriate routine to 1322 * produce the header. 1323 */ 1324 static int 1325 http_authorize(conn_t *conn, const char *hdr, http_auth_challenges_t *cs, 1326 http_auth_params_t *parms, struct url *url) 1327 { 1328 http_auth_challenge_t *digest = NULL; 1329 int i; 1330 1331 /* If user or pass are null we're not happy */ 1332 if (!parms->user || !parms->password) { 1333 DEBUG(fprintf(stderr, "NULL usr or pass\n")); 1334 return (-1); 1335 } 1336 1337 /* Look for a Digest */ 1338 for (i = 0; i < cs->count; i++) { 1339 if (cs->challenges[i]->scheme == HTTPAS_DIGEST) 1340 digest = cs->challenges[i]; 1341 } 1342 1343 /* Error if "Digest" was specified and there is no Digest challenge */ 1344 if (!digest && (parms->scheme && 1345 !strcasecmp(parms->scheme, "digest"))) { 1346 DEBUG(fprintf(stderr, 1347 "Digest auth in env, not supported by peer\n")); 1348 return (-1); 1349 } 1350 /* 1351 * If "basic" was specified in the environment, or there is no Digest 1352 * challenge, do the basic thing. Don't need a challenge for this, 1353 * so no need to check basic!=NULL 1354 */ 1355 if (!digest || (parms->scheme && !strcasecmp(parms->scheme,"basic"))) 1356 return (http_basic_auth(conn,hdr,parms->user,parms->password)); 1357 1358 /* Else, prefer digest. We just checked that it's not NULL */ 1359 return (http_digest_auth(conn, hdr, digest, parms, url)); 1360 } 1361 1362 /***************************************************************************** 1363 * Helper functions for connecting to a server or proxy 1364 */ 1365 1366 /* 1367 * Connect to the correct HTTP server or proxy. 1368 */ 1369 static conn_t * 1370 http_connect(struct url *URL, struct url *purl, const char *flags) 1371 { 1372 conn_t *conn; 1373 int verbose; 1374 int af, val; 1375 1376 #ifdef INET6 1377 af = AF_UNSPEC; 1378 #else 1379 af = AF_INET; 1380 #endif 1381 1382 verbose = CHECK_FLAG('v'); 1383 if (CHECK_FLAG('4')) 1384 af = AF_INET; 1385 #ifdef INET6 1386 else if (CHECK_FLAG('6')) 1387 af = AF_INET6; 1388 #endif 1389 1390 if (purl && strcasecmp(URL->scheme, SCHEME_HTTPS) != 0) { 1391 URL = purl; 1392 } else if (strcasecmp(URL->scheme, SCHEME_FTP) == 0) { 1393 /* can't talk http to an ftp server */ 1394 /* XXX should set an error code */ 1395 return (NULL); 1396 } 1397 1398 if ((conn = fetch_connect(URL->host, URL->port, af, verbose)) == NULL) 1399 /* fetch_connect() has already set an error code */ 1400 return (NULL); 1401 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 && 1402 fetch_ssl(conn, verbose) == -1) { 1403 fetch_close(conn); 1404 /* grrr */ 1405 errno = EAUTH; 1406 fetch_syserr(); 1407 return (NULL); 1408 } 1409 1410 val = 1; 1411 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val)); 1412 1413 return (conn); 1414 } 1415 1416 static struct url * 1417 http_get_proxy(struct url * url, const char *flags) 1418 { 1419 struct url *purl; 1420 char *p; 1421 1422 if (flags != NULL && strchr(flags, 'd') != NULL) 1423 return (NULL); 1424 if (fetch_no_proxy_match(url->host)) 1425 return (NULL); 1426 if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) && 1427 *p && (purl = fetchParseURL(p))) { 1428 if (!*purl->scheme) 1429 strcpy(purl->scheme, SCHEME_HTTP); 1430 if (!purl->port) 1431 purl->port = fetch_default_proxy_port(purl->scheme); 1432 if (strcasecmp(purl->scheme, SCHEME_HTTP) == 0) 1433 return (purl); 1434 fetchFreeURL(purl); 1435 } 1436 return (NULL); 1437 } 1438 1439 static void 1440 http_print_html(FILE *out, FILE *in) 1441 { 1442 size_t len; 1443 char *line, *p, *q; 1444 int comment, tag; 1445 1446 comment = tag = 0; 1447 while ((line = fgetln(in, &len)) != NULL) { 1448 while (len && isspace((unsigned char)line[len - 1])) 1449 --len; 1450 for (p = q = line; q < line + len; ++q) { 1451 if (comment && *q == '-') { 1452 if (q + 2 < line + len && 1453 strcmp(q, "-->") == 0) { 1454 tag = comment = 0; 1455 q += 2; 1456 } 1457 } else if (tag && !comment && *q == '>') { 1458 p = q + 1; 1459 tag = 0; 1460 } else if (!tag && *q == '<') { 1461 if (q > p) 1462 fwrite(p, q - p, 1, out); 1463 tag = 1; 1464 if (q + 3 < line + len && 1465 strcmp(q, "<!--") == 0) { 1466 comment = 1; 1467 q += 3; 1468 } 1469 } 1470 } 1471 if (!tag && q > p) 1472 fwrite(p, q - p, 1, out); 1473 fputc('\n', out); 1474 } 1475 } 1476 1477 1478 /***************************************************************************** 1479 * Core 1480 */ 1481 1482 /* 1483 * Send a request and process the reply 1484 * 1485 * XXX This function is way too long, the do..while loop should be split 1486 * XXX off into a separate function. 1487 */ 1488 FILE * 1489 http_request(struct url *URL, const char *op, struct url_stat *us, 1490 struct url *purl, const char *flags) 1491 { 1492 char timebuf[80]; 1493 char hbuf[MAXHOSTNAMELEN + 7], *host; 1494 conn_t *conn; 1495 struct url *url, *new; 1496 int chunked, direct, ims, noredirect, verbose; 1497 int e, i, n, val; 1498 off_t offset, clength, length, size; 1499 time_t mtime; 1500 const char *p; 1501 FILE *f; 1502 hdr_t h; 1503 struct tm *timestruct; 1504 http_headerbuf_t headerbuf; 1505 http_auth_challenges_t server_challenges; 1506 http_auth_challenges_t proxy_challenges; 1507 1508 /* The following calls don't allocate anything */ 1509 init_http_headerbuf(&headerbuf); 1510 init_http_auth_challenges(&server_challenges); 1511 init_http_auth_challenges(&proxy_challenges); 1512 1513 direct = CHECK_FLAG('d'); 1514 noredirect = CHECK_FLAG('A'); 1515 verbose = CHECK_FLAG('v'); 1516 ims = CHECK_FLAG('i'); 1517 1518 if (direct && purl) { 1519 fetchFreeURL(purl); 1520 purl = NULL; 1521 } 1522 1523 /* try the provided URL first */ 1524 url = URL; 1525 1526 n = MAX_REDIRECT; 1527 i = 0; 1528 1529 e = HTTP_PROTOCOL_ERROR; 1530 do { 1531 new = NULL; 1532 chunked = 0; 1533 offset = 0; 1534 clength = -1; 1535 length = -1; 1536 size = -1; 1537 mtime = 0; 1538 1539 /* check port */ 1540 if (!url->port) 1541 url->port = fetch_default_port(url->scheme); 1542 1543 /* were we redirected to an FTP URL? */ 1544 if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) { 1545 if (strcmp(op, "GET") == 0) 1546 return (ftp_request(url, "RETR", us, purl, flags)); 1547 else if (strcmp(op, "HEAD") == 0) 1548 return (ftp_request(url, "STAT", us, purl, flags)); 1549 } 1550 1551 /* connect to server or proxy */ 1552 if ((conn = http_connect(url, purl, flags)) == NULL) 1553 goto ouch; 1554 1555 host = url->host; 1556 #ifdef INET6 1557 if (strchr(url->host, ':')) { 1558 snprintf(hbuf, sizeof(hbuf), "[%s]", url->host); 1559 host = hbuf; 1560 } 1561 #endif 1562 if (url->port != fetch_default_port(url->scheme)) { 1563 if (host != hbuf) { 1564 strcpy(hbuf, host); 1565 host = hbuf; 1566 } 1567 snprintf(hbuf + strlen(hbuf), 1568 sizeof(hbuf) - strlen(hbuf), ":%d", url->port); 1569 } 1570 1571 /* send request */ 1572 if (verbose) 1573 fetch_info("requesting %s://%s%s", 1574 url->scheme, host, url->doc); 1575 if (purl) { 1576 http_cmd(conn, "%s %s://%s%s HTTP/1.1", 1577 op, url->scheme, host, url->doc); 1578 } else { 1579 http_cmd(conn, "%s %s HTTP/1.1", 1580 op, url->doc); 1581 } 1582 1583 if (ims && url->ims_time) { 1584 timestruct = gmtime((time_t *)&url->ims_time); 1585 (void)strftime(timebuf, 80, "%a, %d %b %Y %T GMT", 1586 timestruct); 1587 if (verbose) 1588 fetch_info("If-Modified-Since: %s", timebuf); 1589 http_cmd(conn, "If-Modified-Since: %s", timebuf); 1590 } 1591 /* virtual host */ 1592 http_cmd(conn, "Host: %s", host); 1593 1594 /* 1595 * Proxy authorization: we only send auth after we received 1596 * a 407 error. We do not first try basic anyway (changed 1597 * when support was added for digest-auth) 1598 */ 1599 if (purl && proxy_challenges.valid) { 1600 http_auth_params_t aparams; 1601 init_http_auth_params(&aparams); 1602 if (*purl->user || *purl->pwd) { 1603 aparams.user = purl->user ? 1604 strdup(purl->user) : strdup(""); 1605 aparams.password = purl->pwd? 1606 strdup(purl->pwd) : strdup(""); 1607 } else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL && 1608 *p != '\0') { 1609 if (http_authfromenv(p, &aparams) < 0) { 1610 http_seterr(HTTP_NEED_PROXY_AUTH); 1611 goto ouch; 1612 } 1613 } 1614 http_authorize(conn, "Proxy-Authorization", 1615 &proxy_challenges, &aparams, url); 1616 clean_http_auth_params(&aparams); 1617 } 1618 1619 /* 1620 * Server authorization: we never send "a priori" 1621 * Basic auth, which used to be done if user/pass were 1622 * set in the url. This would be weird because we'd send the 1623 * password in the clear even if Digest is finally to be 1624 * used (it would have made more sense for the 1625 * pre-digest version to do this when Basic was specified 1626 * in the environment) 1627 */ 1628 if (server_challenges.valid) { 1629 http_auth_params_t aparams; 1630 init_http_auth_params(&aparams); 1631 if (*url->user || *url->pwd) { 1632 aparams.user = url->user ? 1633 strdup(url->user) : strdup(""); 1634 aparams.password = url->pwd ? 1635 strdup(url->pwd) : strdup(""); 1636 } else if ((p = getenv("HTTP_AUTH")) != NULL && 1637 *p != '\0') { 1638 if (http_authfromenv(p, &aparams) < 0) { 1639 http_seterr(HTTP_NEED_AUTH); 1640 goto ouch; 1641 } 1642 } else if (fetchAuthMethod && 1643 fetchAuthMethod(url) == 0) { 1644 aparams.user = url->user ? 1645 strdup(url->user) : strdup(""); 1646 aparams.password = url->pwd ? 1647 strdup(url->pwd) : strdup(""); 1648 } else { 1649 http_seterr(HTTP_NEED_AUTH); 1650 goto ouch; 1651 } 1652 http_authorize(conn, "Authorization", 1653 &server_challenges, &aparams, url); 1654 clean_http_auth_params(&aparams); 1655 } 1656 1657 /* other headers */ 1658 if ((p = getenv("HTTP_REFERER")) != NULL && *p != '\0') { 1659 if (strcasecmp(p, "auto") == 0) 1660 http_cmd(conn, "Referer: %s://%s%s", 1661 url->scheme, host, url->doc); 1662 else 1663 http_cmd(conn, "Referer: %s", p); 1664 } 1665 if ((p = getenv("HTTP_USER_AGENT")) != NULL && *p != '\0') 1666 http_cmd(conn, "User-Agent: %s", p); 1667 else 1668 http_cmd(conn, "User-Agent: %s " _LIBFETCH_VER, getprogname()); 1669 if (url->offset > 0) 1670 http_cmd(conn, "Range: bytes=%lld-", (long long)url->offset); 1671 http_cmd(conn, "Connection: close"); 1672 http_cmd(conn, "%s", ""); 1673 1674 /* 1675 * Force the queued request to be dispatched. Normally, one 1676 * would do this with shutdown(2) but squid proxies can be 1677 * configured to disallow such half-closed connections. To 1678 * be compatible with such configurations, fiddle with socket 1679 * options to force the pending data to be written. 1680 */ 1681 val = 0; 1682 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, 1683 sizeof(val)); 1684 val = 1; 1685 setsockopt(conn->sd, IPPROTO_TCP, TCP_NODELAY, &val, 1686 sizeof(val)); 1687 1688 /* get reply */ 1689 switch (http_get_reply(conn)) { 1690 case HTTP_OK: 1691 case HTTP_PARTIAL: 1692 case HTTP_NOT_MODIFIED: 1693 /* fine */ 1694 break; 1695 case HTTP_MOVED_PERM: 1696 case HTTP_MOVED_TEMP: 1697 case HTTP_SEE_OTHER: 1698 case HTTP_USE_PROXY: 1699 /* 1700 * Not so fine, but we still have to read the 1701 * headers to get the new location. 1702 */ 1703 break; 1704 case HTTP_NEED_AUTH: 1705 if (server_challenges.valid) { 1706 /* 1707 * We already sent out authorization code, 1708 * so there's nothing more we can do. 1709 */ 1710 http_seterr(conn->err); 1711 goto ouch; 1712 } 1713 /* try again, but send the password this time */ 1714 if (verbose) 1715 fetch_info("server requires authorization"); 1716 break; 1717 case HTTP_NEED_PROXY_AUTH: 1718 if (proxy_challenges.valid) { 1719 /* 1720 * We already sent our proxy 1721 * authorization code, so there's 1722 * nothing more we can do. */ 1723 http_seterr(conn->err); 1724 goto ouch; 1725 } 1726 /* try again, but send the password this time */ 1727 if (verbose) 1728 fetch_info("proxy requires authorization"); 1729 break; 1730 case HTTP_BAD_RANGE: 1731 /* 1732 * This can happen if we ask for 0 bytes because 1733 * we already have the whole file. Consider this 1734 * a success for now, and check sizes later. 1735 */ 1736 break; 1737 case HTTP_PROTOCOL_ERROR: 1738 /* fall through */ 1739 case -1: 1740 fetch_syserr(); 1741 goto ouch; 1742 default: 1743 http_seterr(conn->err); 1744 if (!verbose) 1745 goto ouch; 1746 /* fall through so we can get the full error message */ 1747 } 1748 1749 /* get headers. http_next_header expects one line readahead */ 1750 if (fetch_getln(conn) == -1) { 1751 fetch_syserr(); 1752 goto ouch; 1753 } 1754 do { 1755 switch ((h = http_next_header(conn, &headerbuf, &p))) { 1756 case hdr_syserror: 1757 fetch_syserr(); 1758 goto ouch; 1759 case hdr_error: 1760 http_seterr(HTTP_PROTOCOL_ERROR); 1761 goto ouch; 1762 case hdr_content_length: 1763 http_parse_length(p, &clength); 1764 break; 1765 case hdr_content_range: 1766 http_parse_range(p, &offset, &length, &size); 1767 break; 1768 case hdr_last_modified: 1769 http_parse_mtime(p, &mtime); 1770 break; 1771 case hdr_location: 1772 if (!HTTP_REDIRECT(conn->err)) 1773 break; 1774 /* 1775 * if the A flag is set, we don't follow 1776 * temporary redirects. 1777 */ 1778 if (noredirect && 1779 conn->err != HTTP_MOVED_PERM && 1780 conn->err != HTTP_PERM_REDIRECT && 1781 conn->err != HTTP_USE_PROXY) { 1782 n = 1; 1783 break; 1784 } 1785 if (new) 1786 free(new); 1787 if (verbose) 1788 fetch_info("%d redirect to %s", conn->err, p); 1789 if (*p == '/') 1790 /* absolute path */ 1791 new = fetchMakeURL(url->scheme, url->host, url->port, p, 1792 url->user, url->pwd); 1793 else 1794 new = fetchParseURL(p); 1795 if (new == NULL) { 1796 /* XXX should set an error code */ 1797 DEBUG(fprintf(stderr, "failed to parse new URL\n")); 1798 goto ouch; 1799 } 1800 1801 /* Only copy credentials if the host matches */ 1802 if (!strcmp(new->host, url->host) && !*new->user && !*new->pwd) { 1803 strcpy(new->user, url->user); 1804 strcpy(new->pwd, url->pwd); 1805 } 1806 new->offset = url->offset; 1807 new->length = url->length; 1808 break; 1809 case hdr_transfer_encoding: 1810 /* XXX weak test*/ 1811 chunked = (strcasecmp(p, "chunked") == 0); 1812 break; 1813 case hdr_www_authenticate: 1814 if (conn->err != HTTP_NEED_AUTH) 1815 break; 1816 if (http_parse_authenticate(p, &server_challenges) == 0) 1817 ++n; 1818 break; 1819 case hdr_proxy_authenticate: 1820 if (conn->err != HTTP_NEED_PROXY_AUTH) 1821 break; 1822 if (http_parse_authenticate(p, &proxy_challenges) == 0) 1823 ++n; 1824 break; 1825 case hdr_end: 1826 /* fall through */ 1827 case hdr_unknown: 1828 /* ignore */ 1829 break; 1830 } 1831 } while (h > hdr_end); 1832 1833 /* we need to provide authentication */ 1834 if (conn->err == HTTP_NEED_AUTH || 1835 conn->err == HTTP_NEED_PROXY_AUTH) { 1836 e = conn->err; 1837 if ((conn->err == HTTP_NEED_AUTH && 1838 !server_challenges.valid) || 1839 (conn->err == HTTP_NEED_PROXY_AUTH && 1840 !proxy_challenges.valid)) { 1841 /* 401/7 but no www/proxy-authenticate ?? */ 1842 DEBUG(fprintf(stderr, "401/7 and no auth header\n")); 1843 goto ouch; 1844 } 1845 fetch_close(conn); 1846 conn = NULL; 1847 continue; 1848 } 1849 1850 /* requested range not satisfiable */ 1851 if (conn->err == HTTP_BAD_RANGE) { 1852 if (url->offset == size && url->length == 0) { 1853 /* asked for 0 bytes; fake it */ 1854 offset = url->offset; 1855 clength = -1; 1856 conn->err = HTTP_OK; 1857 break; 1858 } else { 1859 http_seterr(conn->err); 1860 goto ouch; 1861 } 1862 } 1863 1864 /* we have a hit or an error */ 1865 if (conn->err == HTTP_OK 1866 || conn->err == HTTP_NOT_MODIFIED 1867 || conn->err == HTTP_PARTIAL 1868 || HTTP_ERROR(conn->err)) 1869 break; 1870 1871 /* all other cases: we got a redirect */ 1872 e = conn->err; 1873 clean_http_auth_challenges(&server_challenges); 1874 fetch_close(conn); 1875 conn = NULL; 1876 if (!new) { 1877 DEBUG(fprintf(stderr, "redirect with no new location\n")); 1878 break; 1879 } 1880 if (url != URL) 1881 fetchFreeURL(url); 1882 url = new; 1883 } while (++i < n); 1884 1885 /* we failed, or ran out of retries */ 1886 if (conn == NULL) { 1887 http_seterr(e); 1888 goto ouch; 1889 } 1890 1891 DEBUG(fprintf(stderr, "offset %lld, length %lld," 1892 " size %lld, clength %lld\n", 1893 (long long)offset, (long long)length, 1894 (long long)size, (long long)clength)); 1895 1896 if (conn->err == HTTP_NOT_MODIFIED) { 1897 http_seterr(HTTP_NOT_MODIFIED); 1898 return (NULL); 1899 } 1900 1901 /* check for inconsistencies */ 1902 if (clength != -1 && length != -1 && clength != length) { 1903 http_seterr(HTTP_PROTOCOL_ERROR); 1904 goto ouch; 1905 } 1906 if (clength == -1) 1907 clength = length; 1908 if (clength != -1) 1909 length = offset + clength; 1910 if (length != -1 && size != -1 && length != size) { 1911 http_seterr(HTTP_PROTOCOL_ERROR); 1912 goto ouch; 1913 } 1914 if (size == -1) 1915 size = length; 1916 1917 /* fill in stats */ 1918 if (us) { 1919 us->size = size; 1920 us->atime = us->mtime = mtime; 1921 } 1922 1923 /* too far? */ 1924 if (URL->offset > 0 && offset > URL->offset) { 1925 http_seterr(HTTP_PROTOCOL_ERROR); 1926 goto ouch; 1927 } 1928 1929 /* report back real offset and size */ 1930 URL->offset = offset; 1931 URL->length = clength; 1932 1933 /* wrap it up in a FILE */ 1934 if ((f = http_funopen(conn, chunked)) == NULL) { 1935 fetch_syserr(); 1936 goto ouch; 1937 } 1938 1939 if (url != URL) 1940 fetchFreeURL(url); 1941 if (purl) 1942 fetchFreeURL(purl); 1943 1944 if (HTTP_ERROR(conn->err)) { 1945 http_print_html(stderr, f); 1946 fclose(f); 1947 f = NULL; 1948 } 1949 clean_http_headerbuf(&headerbuf); 1950 clean_http_auth_challenges(&server_challenges); 1951 clean_http_auth_challenges(&proxy_challenges); 1952 return (f); 1953 1954 ouch: 1955 if (url != URL) 1956 fetchFreeURL(url); 1957 if (purl) 1958 fetchFreeURL(purl); 1959 if (conn != NULL) 1960 fetch_close(conn); 1961 clean_http_headerbuf(&headerbuf); 1962 clean_http_auth_challenges(&server_challenges); 1963 clean_http_auth_challenges(&proxy_challenges); 1964 return (NULL); 1965 } 1966 1967 1968 /***************************************************************************** 1969 * Entry points 1970 */ 1971 1972 /* 1973 * Retrieve and stat a file by HTTP 1974 */ 1975 FILE * 1976 fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags) 1977 { 1978 return (http_request(URL, "GET", us, http_get_proxy(URL, flags), flags)); 1979 } 1980 1981 /* 1982 * Retrieve a file by HTTP 1983 */ 1984 FILE * 1985 fetchGetHTTP(struct url *URL, const char *flags) 1986 { 1987 return (fetchXGetHTTP(URL, NULL, flags)); 1988 } 1989 1990 /* 1991 * Store a file by HTTP 1992 */ 1993 FILE * 1994 fetchPutHTTP(struct url *URL __unused, const char *flags __unused) 1995 { 1996 warnx("fetchPutHTTP(): not implemented"); 1997 return (NULL); 1998 } 1999 2000 /* 2001 * Get an HTTP document's metadata 2002 */ 2003 int 2004 fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags) 2005 { 2006 FILE *f; 2007 2008 f = http_request(URL, "HEAD", us, http_get_proxy(URL, flags), flags); 2009 if (f == NULL) 2010 return (-1); 2011 fclose(f); 2012 return (0); 2013 } 2014 2015 /* 2016 * List a directory 2017 */ 2018 struct url_ent * 2019 fetchListHTTP(struct url *url __unused, const char *flags __unused) 2020 { 2021 warnx("fetchListHTTP(): not implemented"); 2022 return (NULL); 2023 } 2024