1 /*- 2 * Copyright (c) 2000-2014 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 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 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 #ifdef WITH_SSL 81 #include <openssl/md5.h> 82 #define MD5Init(c) MD5_Init(c) 83 #define MD5Update(c, data, len) MD5_Update(c, data, len) 84 #define MD5Final(md, c) MD5_Final(md, c) 85 #else 86 #include <md5.h> 87 #endif 88 89 #include <netinet/in.h> 90 #include <netinet/tcp.h> 91 92 #include "fetch.h" 93 #include "common.h" 94 #include "httperr.h" 95 96 /* Maximum number of redirects to follow */ 97 #define MAX_REDIRECT 20 98 99 /* Symbolic names for reply codes we care about */ 100 #define HTTP_OK 200 101 #define HTTP_PARTIAL 206 102 #define HTTP_MOVED_PERM 301 103 #define HTTP_MOVED_TEMP 302 104 #define HTTP_SEE_OTHER 303 105 #define HTTP_NOT_MODIFIED 304 106 #define HTTP_USE_PROXY 305 107 #define HTTP_TEMP_REDIRECT 307 108 #define HTTP_PERM_REDIRECT 308 109 #define HTTP_NEED_AUTH 401 110 #define HTTP_NEED_PROXY_AUTH 407 111 #define HTTP_BAD_RANGE 416 112 #define HTTP_PROTOCOL_ERROR 999 113 114 #define HTTP_REDIRECT(xyz) ((xyz) == HTTP_MOVED_PERM \ 115 || (xyz) == HTTP_MOVED_TEMP \ 116 || (xyz) == HTTP_TEMP_REDIRECT \ 117 || (xyz) == HTTP_USE_PROXY \ 118 || (xyz) == HTTP_SEE_OTHER) 119 120 #define HTTP_ERROR(xyz) ((xyz) > 400 && (xyz) < 599) 121 122 123 /***************************************************************************** 124 * I/O functions for decoding chunked streams 125 */ 126 127 struct httpio 128 { 129 conn_t *conn; /* connection */ 130 int chunked; /* chunked mode */ 131 char *buf; /* chunk buffer */ 132 size_t bufsize; /* size of chunk buffer */ 133 size_t buflen; /* amount of data currently in buffer */ 134 size_t bufpos; /* current read offset in buffer */ 135 int eof; /* end-of-file flag */ 136 int error; /* error flag */ 137 size_t chunksize; /* remaining size of current chunk */ 138 #ifndef NDEBUG 139 size_t total; 140 #endif 141 }; 142 143 /* 144 * Get next chunk header 145 */ 146 static int 147 http_new_chunk(struct httpio *io) 148 { 149 char *p; 150 151 if (fetch_getln(io->conn) == -1) 152 return (-1); 153 154 if (io->conn->buflen < 2 || !isxdigit((unsigned char)*io->conn->buf)) 155 return (-1); 156 157 for (p = io->conn->buf; *p && !isspace((unsigned char)*p); ++p) { 158 if (*p == ';') 159 break; 160 if (!isxdigit((unsigned char)*p)) 161 return (-1); 162 if (isdigit((unsigned char)*p)) { 163 io->chunksize = io->chunksize * 16 + 164 *p - '0'; 165 } else { 166 io->chunksize = io->chunksize * 16 + 167 10 + tolower((unsigned char)*p) - 'a'; 168 } 169 } 170 171 #ifndef NDEBUG 172 if (fetchDebug) { 173 io->total += io->chunksize; 174 if (io->chunksize == 0) 175 fprintf(stderr, "%s(): end of last chunk\n", __func__); 176 else 177 fprintf(stderr, "%s(): new chunk: %lu (%lu)\n", 178 __func__, (unsigned long)io->chunksize, 179 (unsigned long)io->total); 180 } 181 #endif 182 183 return (io->chunksize); 184 } 185 186 /* 187 * Grow the input buffer to at least len bytes 188 */ 189 static inline int 190 http_growbuf(struct httpio *io, size_t len) 191 { 192 char *tmp; 193 194 if (io->bufsize >= len) 195 return (0); 196 197 if ((tmp = realloc(io->buf, len)) == NULL) 198 return (-1); 199 io->buf = tmp; 200 io->bufsize = len; 201 return (0); 202 } 203 204 /* 205 * Fill the input buffer, do chunk decoding on the fly 206 */ 207 static ssize_t 208 http_fillbuf(struct httpio *io, size_t len) 209 { 210 ssize_t nbytes; 211 char ch; 212 213 if (io->error) 214 return (-1); 215 if (io->eof) 216 return (0); 217 218 /* not chunked: just fetch the requested amount */ 219 if (io->chunked == 0) { 220 if (http_growbuf(io, len) == -1) 221 return (-1); 222 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) { 223 io->error = errno; 224 return (-1); 225 } 226 io->buflen = nbytes; 227 io->bufpos = 0; 228 return (io->buflen); 229 } 230 231 /* chunked, but we ran out: get the next chunk header */ 232 if (io->chunksize == 0) { 233 switch (http_new_chunk(io)) { 234 case -1: 235 io->error = EPROTO; 236 return (-1); 237 case 0: 238 io->eof = 1; 239 return (0); 240 } 241 } 242 243 /* fetch the requested amount, but no more than the current chunk */ 244 if (len > io->chunksize) 245 len = io->chunksize; 246 if (http_growbuf(io, len) == -1) 247 return (-1); 248 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) { 249 io->error = errno; 250 return (-1); 251 } 252 io->bufpos = 0; 253 io->buflen = nbytes; 254 io->chunksize -= nbytes; 255 256 if (io->chunksize == 0) { 257 if (fetch_read(io->conn, &ch, 1) != 1 || ch != '\r' || 258 fetch_read(io->conn, &ch, 1) != 1 || ch != '\n') 259 return (-1); 260 } 261 262 return (io->buflen); 263 } 264 265 /* 266 * Read function 267 */ 268 static int 269 http_readfn(void *v, char *buf, int len) 270 { 271 struct httpio *io = (struct httpio *)v; 272 int rlen; 273 274 if (io->error) 275 return (-1); 276 if (io->eof) 277 return (0); 278 279 /* empty buffer */ 280 if (!io->buf || io->bufpos == io->buflen) { 281 if ((rlen = http_fillbuf(io, len)) < 0) { 282 if ((errno = io->error) == EINTR) 283 io->error = 0; 284 return (-1); 285 } else if (rlen == 0) { 286 return (0); 287 } 288 } 289 290 rlen = io->buflen - io->bufpos; 291 if (len < rlen) 292 rlen = len; 293 memcpy(buf, io->buf + io->bufpos, rlen); 294 io->bufpos += rlen; 295 return (rlen); 296 } 297 298 /* 299 * Write function 300 */ 301 static int 302 http_writefn(void *v, const char *buf, int len) 303 { 304 struct httpio *io = (struct httpio *)v; 305 306 return (fetch_write(io->conn, buf, len)); 307 } 308 309 /* 310 * Close function 311 */ 312 static int 313 http_closefn(void *v) 314 { 315 struct httpio *io = (struct httpio *)v; 316 int r; 317 318 r = fetch_close(io->conn); 319 if (io->buf) 320 free(io->buf); 321 free(io); 322 return (r); 323 } 324 325 /* 326 * Wrap a file descriptor up 327 */ 328 static FILE * 329 http_funopen(conn_t *conn, int chunked) 330 { 331 struct httpio *io; 332 FILE *f; 333 334 if ((io = calloc(1, sizeof(*io))) == NULL) { 335 fetch_syserr(); 336 return (NULL); 337 } 338 io->conn = conn; 339 io->chunked = chunked; 340 f = funopen(io, http_readfn, http_writefn, NULL, http_closefn); 341 if (f == NULL) { 342 fetch_syserr(); 343 free(io); 344 return (NULL); 345 } 346 return (f); 347 } 348 349 350 /***************************************************************************** 351 * Helper functions for talking to the server and parsing its replies 352 */ 353 354 /* Header types */ 355 typedef enum { 356 hdr_syserror = -2, 357 hdr_error = -1, 358 hdr_end = 0, 359 hdr_unknown = 1, 360 hdr_content_length, 361 hdr_content_range, 362 hdr_last_modified, 363 hdr_location, 364 hdr_transfer_encoding, 365 hdr_www_authenticate, 366 hdr_proxy_authenticate, 367 } hdr_t; 368 369 /* Names of interesting headers */ 370 static struct { 371 hdr_t num; 372 const char *name; 373 } hdr_names[] = { 374 { hdr_content_length, "Content-Length" }, 375 { hdr_content_range, "Content-Range" }, 376 { hdr_last_modified, "Last-Modified" }, 377 { hdr_location, "Location" }, 378 { hdr_transfer_encoding, "Transfer-Encoding" }, 379 { hdr_www_authenticate, "WWW-Authenticate" }, 380 { hdr_proxy_authenticate, "Proxy-Authenticate" }, 381 { hdr_unknown, NULL }, 382 }; 383 384 /* 385 * Send a formatted line; optionally echo to terminal 386 */ 387 static int 388 http_cmd(conn_t *conn, const char *fmt, ...) 389 { 390 va_list ap; 391 size_t len; 392 char *msg; 393 int r; 394 395 va_start(ap, fmt); 396 len = vasprintf(&msg, fmt, ap); 397 va_end(ap); 398 399 if (msg == NULL) { 400 errno = ENOMEM; 401 fetch_syserr(); 402 return (-1); 403 } 404 405 r = fetch_putln(conn, msg, len); 406 free(msg); 407 408 if (r == -1) { 409 fetch_syserr(); 410 return (-1); 411 } 412 413 return (0); 414 } 415 416 /* 417 * Get and parse status line 418 */ 419 static int 420 http_get_reply(conn_t *conn) 421 { 422 char *p; 423 424 if (fetch_getln(conn) == -1) 425 return (-1); 426 /* 427 * A valid status line looks like "HTTP/m.n xyz reason" where m 428 * and n are the major and minor protocol version numbers and xyz 429 * is the reply code. 430 * Unfortunately, there are servers out there (NCSA 1.5.1, to name 431 * just one) that do not send a version number, so we can't rely 432 * on finding one, but if we do, insist on it being 1.0 or 1.1. 433 * We don't care about the reason phrase. 434 */ 435 if (strncmp(conn->buf, "HTTP", 4) != 0) 436 return (HTTP_PROTOCOL_ERROR); 437 p = conn->buf + 4; 438 if (*p == '/') { 439 if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1')) 440 return (HTTP_PROTOCOL_ERROR); 441 p += 4; 442 } 443 if (*p != ' ' || 444 !isdigit((unsigned char)p[1]) || 445 !isdigit((unsigned char)p[2]) || 446 !isdigit((unsigned char)p[3])) 447 return (HTTP_PROTOCOL_ERROR); 448 449 conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0'); 450 return (conn->err); 451 } 452 453 /* 454 * Check a header; if the type matches the given string, return a pointer 455 * to the beginning of the value. 456 */ 457 static const char * 458 http_match(const char *str, const char *hdr) 459 { 460 while (*str && *hdr && 461 tolower((unsigned char)*str++) == tolower((unsigned char)*hdr++)) 462 /* nothing */; 463 if (*str || *hdr != ':') 464 return (NULL); 465 while (*hdr && isspace((unsigned char)*++hdr)) 466 /* nothing */; 467 return (hdr); 468 } 469 470 471 /* 472 * Get the next header and return the appropriate symbolic code. We 473 * need to read one line ahead for checking for a continuation line 474 * belonging to the current header (continuation lines start with 475 * white space). 476 * 477 * We get called with a fresh line already in the conn buffer, either 478 * from the previous http_next_header() invocation, or, the first 479 * time, from a fetch_getln() performed by our caller. 480 * 481 * This stops when we encounter an empty line (we dont read beyond the header 482 * area). 483 * 484 * Note that the "headerbuf" is just a place to return the result. Its 485 * contents are not used for the next call. This means that no cleanup 486 * is needed when ie doing another connection, just call the cleanup when 487 * fully done to deallocate memory. 488 */ 489 490 /* Limit the max number of continuation lines to some reasonable value */ 491 #define HTTP_MAX_CONT_LINES 10 492 493 /* Place into which to build a header from one or several lines */ 494 typedef struct { 495 char *buf; /* buffer */ 496 size_t bufsize; /* buffer size */ 497 size_t buflen; /* length of buffer contents */ 498 } http_headerbuf_t; 499 500 static void 501 init_http_headerbuf(http_headerbuf_t *buf) 502 { 503 buf->buf = NULL; 504 buf->bufsize = 0; 505 buf->buflen = 0; 506 } 507 508 static void 509 clean_http_headerbuf(http_headerbuf_t *buf) 510 { 511 if (buf->buf) 512 free(buf->buf); 513 init_http_headerbuf(buf); 514 } 515 516 /* Remove whitespace at the end of the buffer */ 517 static void 518 http_conn_trimright(conn_t *conn) 519 { 520 while (conn->buflen && 521 isspace((unsigned char)conn->buf[conn->buflen - 1])) 522 conn->buflen--; 523 conn->buf[conn->buflen] = '\0'; 524 } 525 526 static hdr_t 527 http_next_header(conn_t *conn, http_headerbuf_t *hbuf, const char **p) 528 { 529 unsigned int i, len; 530 531 /* 532 * Have to do the stripping here because of the first line. So 533 * it's done twice for the subsequent lines. No big deal 534 */ 535 http_conn_trimright(conn); 536 if (conn->buflen == 0) 537 return (hdr_end); 538 539 /* Copy the line to the headerbuf */ 540 if (hbuf->bufsize < conn->buflen + 1) { 541 if ((hbuf->buf = realloc(hbuf->buf, conn->buflen + 1)) == NULL) 542 return (hdr_syserror); 543 hbuf->bufsize = conn->buflen + 1; 544 } 545 strcpy(hbuf->buf, conn->buf); 546 hbuf->buflen = conn->buflen; 547 548 /* 549 * Fetch possible continuation lines. Stop at 1st non-continuation 550 * and leave it in the conn buffer 551 */ 552 for (i = 0; i < HTTP_MAX_CONT_LINES; i++) { 553 if (fetch_getln(conn) == -1) 554 return (hdr_syserror); 555 556 /* 557 * Note: we carry on the idea from the previous version 558 * that a pure whitespace line is equivalent to an empty 559 * one (so it's not continuation and will be handled when 560 * we are called next) 561 */ 562 http_conn_trimright(conn); 563 if (conn->buf[0] != ' ' && conn->buf[0] != "\t"[0]) 564 break; 565 566 /* Got a continuation line. Concatenate to previous */ 567 len = hbuf->buflen + conn->buflen; 568 if (hbuf->bufsize < len + 1) { 569 len *= 2; 570 if ((hbuf->buf = realloc(hbuf->buf, len + 1)) == NULL) 571 return (hdr_syserror); 572 hbuf->bufsize = len + 1; 573 } 574 strcpy(hbuf->buf + hbuf->buflen, conn->buf); 575 hbuf->buflen += conn->buflen; 576 } 577 578 /* 579 * We could check for malformed headers but we don't really care. 580 * A valid header starts with a token immediately followed by a 581 * colon; a token is any sequence of non-control, non-whitespace 582 * characters except "()<>@,;:\\\"{}". 583 */ 584 for (i = 0; hdr_names[i].num != hdr_unknown; i++) 585 if ((*p = http_match(hdr_names[i].name, hbuf->buf)) != NULL) 586 return (hdr_names[i].num); 587 588 return (hdr_unknown); 589 } 590 591 /************************** 592 * [Proxy-]Authenticate header parsing 593 */ 594 595 /* 596 * Read doublequote-delimited string into output buffer obuf (allocated 597 * by caller, whose responsibility it is to ensure that it's big enough) 598 * cp points to the first char after the initial '"' 599 * Handles \ quoting 600 * Returns pointer to the first char after the terminating double quote, or 601 * NULL for error. 602 */ 603 static const char * 604 http_parse_headerstring(const char *cp, char *obuf) 605 { 606 for (;;) { 607 switch (*cp) { 608 case 0: /* Unterminated string */ 609 *obuf = 0; 610 return (NULL); 611 case '"': /* Ending quote */ 612 *obuf = 0; 613 return (++cp); 614 case '\\': 615 if (*++cp == 0) { 616 *obuf = 0; 617 return (NULL); 618 } 619 /* FALLTHROUGH */ 620 default: 621 *obuf++ = *cp++; 622 } 623 } 624 } 625 626 /* Http auth challenge schemes */ 627 typedef enum {HTTPAS_UNKNOWN, HTTPAS_BASIC,HTTPAS_DIGEST} http_auth_schemes_t; 628 629 /* Data holder for a Basic or Digest challenge. */ 630 typedef struct { 631 http_auth_schemes_t scheme; 632 char *realm; 633 char *qop; 634 char *nonce; 635 char *opaque; 636 char *algo; 637 int stale; 638 int nc; /* Nonce count */ 639 } http_auth_challenge_t; 640 641 static void 642 init_http_auth_challenge(http_auth_challenge_t *b) 643 { 644 b->scheme = HTTPAS_UNKNOWN; 645 b->realm = b->qop = b->nonce = b->opaque = b->algo = NULL; 646 b->stale = b->nc = 0; 647 } 648 649 static void 650 clean_http_auth_challenge(http_auth_challenge_t *b) 651 { 652 if (b->realm) 653 free(b->realm); 654 if (b->qop) 655 free(b->qop); 656 if (b->nonce) 657 free(b->nonce); 658 if (b->opaque) 659 free(b->opaque); 660 if (b->algo) 661 free(b->algo); 662 init_http_auth_challenge(b); 663 } 664 665 /* Data holder for an array of challenges offered in an http response. */ 666 #define MAX_CHALLENGES 10 667 typedef struct { 668 http_auth_challenge_t *challenges[MAX_CHALLENGES]; 669 int count; /* Number of parsed challenges in the array */ 670 int valid; /* We did parse an authenticate header */ 671 } http_auth_challenges_t; 672 673 static void 674 init_http_auth_challenges(http_auth_challenges_t *cs) 675 { 676 int i; 677 for (i = 0; i < MAX_CHALLENGES; i++) 678 cs->challenges[i] = NULL; 679 cs->count = cs->valid = 0; 680 } 681 682 static void 683 clean_http_auth_challenges(http_auth_challenges_t *cs) 684 { 685 int i; 686 /* We rely on non-zero pointers being allocated, not on the count */ 687 for (i = 0; i < MAX_CHALLENGES; i++) { 688 if (cs->challenges[i] != NULL) { 689 clean_http_auth_challenge(cs->challenges[i]); 690 free(cs->challenges[i]); 691 } 692 } 693 init_http_auth_challenges(cs); 694 } 695 696 /* 697 * Enumeration for lexical elements. Separators will be returned as their own 698 * ascii value 699 */ 700 typedef enum {HTTPHL_WORD=256, HTTPHL_STRING=257, HTTPHL_END=258, 701 HTTPHL_ERROR = 259} http_header_lex_t; 702 703 /* 704 * Determine what kind of token comes next and return possible value 705 * in buf, which is supposed to have been allocated big enough by 706 * caller. Advance input pointer and return element type. 707 */ 708 static int 709 http_header_lex(const char **cpp, char *buf) 710 { 711 size_t l; 712 /* Eat initial whitespace */ 713 *cpp += strspn(*cpp, " \t"); 714 if (**cpp == 0) 715 return (HTTPHL_END); 716 717 /* Separator ? */ 718 if (**cpp == ',' || **cpp == '=') 719 return (*((*cpp)++)); 720 721 /* String ? */ 722 if (**cpp == '"') { 723 *cpp = http_parse_headerstring(++*cpp, buf); 724 if (*cpp == NULL) 725 return (HTTPHL_ERROR); 726 return (HTTPHL_STRING); 727 } 728 729 /* Read other token, until separator or whitespace */ 730 l = strcspn(*cpp, " \t,="); 731 memcpy(buf, *cpp, l); 732 buf[l] = 0; 733 *cpp += l; 734 return (HTTPHL_WORD); 735 } 736 737 /* 738 * Read challenges from http xxx-authenticate header and accumulate them 739 * in the challenges list structure. 740 * 741 * Headers with multiple challenges are specified by rfc2617, but 742 * servers (ie: squid) often send them in separate headers instead, 743 * which in turn is forbidden by the http spec (multiple headers with 744 * the same name are only allowed for pure comma-separated lists, see 745 * rfc2616 sec 4.2). 746 * 747 * We support both approaches anyway 748 */ 749 static int 750 http_parse_authenticate(const char *cp, http_auth_challenges_t *cs) 751 { 752 int ret = -1; 753 http_header_lex_t lex; 754 char *key = malloc(strlen(cp) + 1); 755 char *value = malloc(strlen(cp) + 1); 756 char *buf = malloc(strlen(cp) + 1); 757 758 if (key == NULL || value == NULL || buf == NULL) { 759 fetch_syserr(); 760 goto out; 761 } 762 763 /* In any case we've seen the header and we set the valid bit */ 764 cs->valid = 1; 765 766 /* Need word first */ 767 lex = http_header_lex(&cp, key); 768 if (lex != HTTPHL_WORD) 769 goto out; 770 771 /* Loop on challenges */ 772 for (; cs->count < MAX_CHALLENGES; cs->count++) { 773 cs->challenges[cs->count] = 774 malloc(sizeof(http_auth_challenge_t)); 775 if (cs->challenges[cs->count] == NULL) { 776 fetch_syserr(); 777 goto out; 778 } 779 init_http_auth_challenge(cs->challenges[cs->count]); 780 if (!strcasecmp(key, "basic")) { 781 cs->challenges[cs->count]->scheme = HTTPAS_BASIC; 782 } else if (!strcasecmp(key, "digest")) { 783 cs->challenges[cs->count]->scheme = HTTPAS_DIGEST; 784 } else { 785 cs->challenges[cs->count]->scheme = HTTPAS_UNKNOWN; 786 /* 787 * Continue parsing as basic or digest may 788 * follow, and the syntax is the same for 789 * all. We'll just ignore this one when 790 * looking at the list 791 */ 792 } 793 794 /* Loop on attributes */ 795 for (;;) { 796 /* Key */ 797 lex = http_header_lex(&cp, key); 798 if (lex != HTTPHL_WORD) 799 goto out; 800 801 /* Equal sign */ 802 lex = http_header_lex(&cp, buf); 803 if (lex != '=') 804 goto out; 805 806 /* Value */ 807 lex = http_header_lex(&cp, value); 808 if (lex != HTTPHL_WORD && lex != HTTPHL_STRING) 809 goto out; 810 811 if (!strcasecmp(key, "realm")) 812 cs->challenges[cs->count]->realm = 813 strdup(value); 814 else if (!strcasecmp(key, "qop")) 815 cs->challenges[cs->count]->qop = 816 strdup(value); 817 else if (!strcasecmp(key, "nonce")) 818 cs->challenges[cs->count]->nonce = 819 strdup(value); 820 else if (!strcasecmp(key, "opaque")) 821 cs->challenges[cs->count]->opaque = 822 strdup(value); 823 else if (!strcasecmp(key, "algorithm")) 824 cs->challenges[cs->count]->algo = 825 strdup(value); 826 else if (!strcasecmp(key, "stale")) 827 cs->challenges[cs->count]->stale = 828 strcasecmp(value, "no"); 829 /* Else ignore unknown attributes */ 830 831 /* Comma or Next challenge or End */ 832 lex = http_header_lex(&cp, key); 833 /* 834 * If we get a word here, this is the beginning of the 835 * next challenge. Break the attributes loop 836 */ 837 if (lex == HTTPHL_WORD) 838 break; 839 840 if (lex == HTTPHL_END) { 841 /* End while looking for ',' is normal exit */ 842 cs->count++; 843 ret = 0; 844 goto out; 845 } 846 /* Anything else is an error */ 847 if (lex != ',') 848 goto out; 849 850 } /* End attributes loop */ 851 } /* End challenge loop */ 852 853 /* 854 * Challenges max count exceeded. This really can't happen 855 * with normal data, something's fishy -> error 856 */ 857 858 out: 859 if (key) 860 free(key); 861 if (value) 862 free(value); 863 if (buf) 864 free(buf); 865 return (ret); 866 } 867 868 869 /* 870 * Parse a last-modified header 871 */ 872 static int 873 http_parse_mtime(const char *p, time_t *mtime) 874 { 875 char locale[64], *r; 876 struct tm tm; 877 878 strncpy(locale, setlocale(LC_TIME, NULL), sizeof(locale)); 879 setlocale(LC_TIME, "C"); 880 r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm); 881 /* 882 * Some proxies use UTC in response, but it should still be 883 * parsed. RFC2616 states GMT and UTC are exactly equal for HTTP. 884 */ 885 if (r == NULL) 886 r = strptime(p, "%a, %d %b %Y %H:%M:%S UTC", &tm); 887 /* XXX should add support for date-2 and date-3 */ 888 setlocale(LC_TIME, locale); 889 if (r == NULL) 890 return (-1); 891 DEBUG(fprintf(stderr, "last modified: [%04d-%02d-%02d " 892 "%02d:%02d:%02d]\n", 893 tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday, 894 tm.tm_hour, tm.tm_min, tm.tm_sec)); 895 *mtime = timegm(&tm); 896 return (0); 897 } 898 899 /* 900 * Parse a content-length header 901 */ 902 static int 903 http_parse_length(const char *p, off_t *length) 904 { 905 off_t len; 906 907 for (len = 0; *p && isdigit((unsigned char)*p); ++p) 908 len = len * 10 + (*p - '0'); 909 if (*p) 910 return (-1); 911 DEBUG(fprintf(stderr, "content length: [%lld]\n", 912 (long long)len)); 913 *length = len; 914 return (0); 915 } 916 917 /* 918 * Parse a content-range header 919 */ 920 static int 921 http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size) 922 { 923 off_t first, last, len; 924 925 if (strncasecmp(p, "bytes ", 6) != 0) 926 return (-1); 927 p += 6; 928 if (*p == '*') { 929 first = last = -1; 930 ++p; 931 } else { 932 for (first = 0; *p && isdigit((unsigned char)*p); ++p) 933 first = first * 10 + *p - '0'; 934 if (*p != '-') 935 return (-1); 936 for (last = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 937 last = last * 10 + *p - '0'; 938 } 939 if (first > last || *p != '/') 940 return (-1); 941 for (len = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 942 len = len * 10 + *p - '0'; 943 if (*p || len < last - first + 1) 944 return (-1); 945 if (first == -1) { 946 DEBUG(fprintf(stderr, "content range: [*/%lld]\n", 947 (long long)len)); 948 *length = 0; 949 } else { 950 DEBUG(fprintf(stderr, "content range: [%lld-%lld/%lld]\n", 951 (long long)first, (long long)last, (long long)len)); 952 *length = last - first + 1; 953 } 954 *offset = first; 955 *size = len; 956 return (0); 957 } 958 959 960 /***************************************************************************** 961 * Helper functions for authorization 962 */ 963 964 /* 965 * Base64 encoding 966 */ 967 static char * 968 http_base64(const char *src) 969 { 970 static const char base64[] = 971 "ABCDEFGHIJKLMNOPQRSTUVWXYZ" 972 "abcdefghijklmnopqrstuvwxyz" 973 "0123456789+/"; 974 char *str, *dst; 975 size_t l; 976 int t, r; 977 978 l = strlen(src); 979 if ((str = malloc(((l + 2) / 3) * 4 + 1)) == NULL) 980 return (NULL); 981 dst = str; 982 r = 0; 983 984 while (l >= 3) { 985 t = (src[0] << 16) | (src[1] << 8) | src[2]; 986 dst[0] = base64[(t >> 18) & 0x3f]; 987 dst[1] = base64[(t >> 12) & 0x3f]; 988 dst[2] = base64[(t >> 6) & 0x3f]; 989 dst[3] = base64[(t >> 0) & 0x3f]; 990 src += 3; l -= 3; 991 dst += 4; r += 4; 992 } 993 994 switch (l) { 995 case 2: 996 t = (src[0] << 16) | (src[1] << 8); 997 dst[0] = base64[(t >> 18) & 0x3f]; 998 dst[1] = base64[(t >> 12) & 0x3f]; 999 dst[2] = base64[(t >> 6) & 0x3f]; 1000 dst[3] = '='; 1001 dst += 4; 1002 r += 4; 1003 break; 1004 case 1: 1005 t = src[0] << 16; 1006 dst[0] = base64[(t >> 18) & 0x3f]; 1007 dst[1] = base64[(t >> 12) & 0x3f]; 1008 dst[2] = dst[3] = '='; 1009 dst += 4; 1010 r += 4; 1011 break; 1012 case 0: 1013 break; 1014 } 1015 1016 *dst = 0; 1017 return (str); 1018 } 1019 1020 1021 /* 1022 * Extract authorization parameters from environment value. 1023 * The value is like scheme:realm:user:pass 1024 */ 1025 typedef struct { 1026 char *scheme; 1027 char *realm; 1028 char *user; 1029 char *password; 1030 } http_auth_params_t; 1031 1032 static void 1033 init_http_auth_params(http_auth_params_t *s) 1034 { 1035 s->scheme = s->realm = s->user = s->password = NULL; 1036 } 1037 1038 static void 1039 clean_http_auth_params(http_auth_params_t *s) 1040 { 1041 if (s->scheme) 1042 free(s->scheme); 1043 if (s->realm) 1044 free(s->realm); 1045 if (s->user) 1046 free(s->user); 1047 if (s->password) 1048 free(s->password); 1049 init_http_auth_params(s); 1050 } 1051 1052 static int 1053 http_authfromenv(const char *p, http_auth_params_t *parms) 1054 { 1055 int ret = -1; 1056 char *v, *ve; 1057 char *str = strdup(p); 1058 1059 if (str == NULL) { 1060 fetch_syserr(); 1061 return (-1); 1062 } 1063 v = str; 1064 1065 if ((ve = strchr(v, ':')) == NULL) 1066 goto out; 1067 1068 *ve = 0; 1069 if ((parms->scheme = strdup(v)) == NULL) { 1070 fetch_syserr(); 1071 goto out; 1072 } 1073 v = ve + 1; 1074 1075 if ((ve = strchr(v, ':')) == NULL) 1076 goto out; 1077 1078 *ve = 0; 1079 if ((parms->realm = strdup(v)) == NULL) { 1080 fetch_syserr(); 1081 goto out; 1082 } 1083 v = ve + 1; 1084 1085 if ((ve = strchr(v, ':')) == NULL) 1086 goto out; 1087 1088 *ve = 0; 1089 if ((parms->user = strdup(v)) == NULL) { 1090 fetch_syserr(); 1091 goto out; 1092 } 1093 v = ve + 1; 1094 1095 1096 if ((parms->password = strdup(v)) == NULL) { 1097 fetch_syserr(); 1098 goto out; 1099 } 1100 ret = 0; 1101 out: 1102 if (ret == -1) 1103 clean_http_auth_params(parms); 1104 if (str) 1105 free(str); 1106 return (ret); 1107 } 1108 1109 1110 /* 1111 * Digest response: the code to compute the digest is taken from the 1112 * sample implementation in RFC2616 1113 */ 1114 #define IN const 1115 #define OUT 1116 1117 #define HASHLEN 16 1118 typedef char HASH[HASHLEN]; 1119 #define HASHHEXLEN 32 1120 typedef char HASHHEX[HASHHEXLEN+1]; 1121 1122 static const char *hexchars = "0123456789abcdef"; 1123 static void 1124 CvtHex(IN HASH Bin, OUT HASHHEX Hex) 1125 { 1126 unsigned short i; 1127 unsigned char j; 1128 1129 for (i = 0; i < HASHLEN; i++) { 1130 j = (Bin[i] >> 4) & 0xf; 1131 Hex[i*2] = hexchars[j]; 1132 j = Bin[i] & 0xf; 1133 Hex[i*2+1] = hexchars[j]; 1134 } 1135 Hex[HASHHEXLEN] = '\0'; 1136 }; 1137 1138 /* calculate H(A1) as per spec */ 1139 static void 1140 DigestCalcHA1( 1141 IN char * pszAlg, 1142 IN char * pszUserName, 1143 IN char * pszRealm, 1144 IN char * pszPassword, 1145 IN char * pszNonce, 1146 IN char * pszCNonce, 1147 OUT HASHHEX SessionKey 1148 ) 1149 { 1150 MD5_CTX Md5Ctx; 1151 HASH HA1; 1152 1153 MD5Init(&Md5Ctx); 1154 MD5Update(&Md5Ctx, pszUserName, strlen(pszUserName)); 1155 MD5Update(&Md5Ctx, ":", 1); 1156 MD5Update(&Md5Ctx, pszRealm, strlen(pszRealm)); 1157 MD5Update(&Md5Ctx, ":", 1); 1158 MD5Update(&Md5Ctx, pszPassword, strlen(pszPassword)); 1159 MD5Final(HA1, &Md5Ctx); 1160 if (strcasecmp(pszAlg, "md5-sess") == 0) { 1161 1162 MD5Init(&Md5Ctx); 1163 MD5Update(&Md5Ctx, HA1, HASHLEN); 1164 MD5Update(&Md5Ctx, ":", 1); 1165 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce)); 1166 MD5Update(&Md5Ctx, ":", 1); 1167 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce)); 1168 MD5Final(HA1, &Md5Ctx); 1169 } 1170 CvtHex(HA1, SessionKey); 1171 } 1172 1173 /* calculate request-digest/response-digest as per HTTP Digest spec */ 1174 static void 1175 DigestCalcResponse( 1176 IN HASHHEX HA1, /* H(A1) */ 1177 IN char * pszNonce, /* nonce from server */ 1178 IN char * pszNonceCount, /* 8 hex digits */ 1179 IN char * pszCNonce, /* client nonce */ 1180 IN char * pszQop, /* qop-value: "", "auth", "auth-int" */ 1181 IN char * pszMethod, /* method from the request */ 1182 IN char * pszDigestUri, /* requested URL */ 1183 IN HASHHEX HEntity, /* H(entity body) if qop="auth-int" */ 1184 OUT HASHHEX Response /* request-digest or response-digest */ 1185 ) 1186 { 1187 /* DEBUG(fprintf(stderr, 1188 "Calc: HA1[%s] Nonce[%s] qop[%s] method[%s] URI[%s]\n", 1189 HA1, pszNonce, pszQop, pszMethod, pszDigestUri));*/ 1190 MD5_CTX Md5Ctx; 1191 HASH HA2; 1192 HASH RespHash; 1193 HASHHEX HA2Hex; 1194 1195 // calculate H(A2) 1196 MD5Init(&Md5Ctx); 1197 MD5Update(&Md5Ctx, pszMethod, strlen(pszMethod)); 1198 MD5Update(&Md5Ctx, ":", 1); 1199 MD5Update(&Md5Ctx, pszDigestUri, strlen(pszDigestUri)); 1200 if (strcasecmp(pszQop, "auth-int") == 0) { 1201 MD5Update(&Md5Ctx, ":", 1); 1202 MD5Update(&Md5Ctx, HEntity, HASHHEXLEN); 1203 } 1204 MD5Final(HA2, &Md5Ctx); 1205 CvtHex(HA2, HA2Hex); 1206 1207 // calculate response 1208 MD5Init(&Md5Ctx); 1209 MD5Update(&Md5Ctx, HA1, HASHHEXLEN); 1210 MD5Update(&Md5Ctx, ":", 1); 1211 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce)); 1212 MD5Update(&Md5Ctx, ":", 1); 1213 if (*pszQop) { 1214 MD5Update(&Md5Ctx, pszNonceCount, strlen(pszNonceCount)); 1215 MD5Update(&Md5Ctx, ":", 1); 1216 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce)); 1217 MD5Update(&Md5Ctx, ":", 1); 1218 MD5Update(&Md5Ctx, pszQop, strlen(pszQop)); 1219 MD5Update(&Md5Ctx, ":", 1); 1220 } 1221 MD5Update(&Md5Ctx, HA2Hex, HASHHEXLEN); 1222 MD5Final(RespHash, &Md5Ctx); 1223 CvtHex(RespHash, Response); 1224 } 1225 1226 /* 1227 * Generate/Send a Digest authorization header 1228 * This looks like: [Proxy-]Authorization: credentials 1229 * 1230 * credentials = "Digest" digest-response 1231 * digest-response = 1#( username | realm | nonce | digest-uri 1232 * | response | [ algorithm ] | [cnonce] | 1233 * [opaque] | [message-qop] | 1234 * [nonce-count] | [auth-param] ) 1235 * username = "username" "=" username-value 1236 * username-value = quoted-string 1237 * digest-uri = "uri" "=" digest-uri-value 1238 * digest-uri-value = request-uri ; As specified by HTTP/1.1 1239 * message-qop = "qop" "=" qop-value 1240 * cnonce = "cnonce" "=" cnonce-value 1241 * cnonce-value = nonce-value 1242 * nonce-count = "nc" "=" nc-value 1243 * nc-value = 8LHEX 1244 * response = "response" "=" request-digest 1245 * request-digest = <"> 32LHEX <"> 1246 */ 1247 static int 1248 http_digest_auth(conn_t *conn, const char *hdr, http_auth_challenge_t *c, 1249 http_auth_params_t *parms, struct url *url) 1250 { 1251 int r; 1252 char noncecount[10]; 1253 char cnonce[40]; 1254 char *options = NULL; 1255 1256 if (!c->realm || !c->nonce) { 1257 DEBUG(fprintf(stderr, "realm/nonce not set in challenge\n")); 1258 return(-1); 1259 } 1260 if (!c->algo) 1261 c->algo = strdup(""); 1262 1263 if (asprintf(&options, "%s%s%s%s", 1264 *c->algo? ",algorithm=" : "", c->algo, 1265 c->opaque? ",opaque=" : "", c->opaque?c->opaque:"")== -1) 1266 return (-1); 1267 1268 if (!c->qop) { 1269 c->qop = strdup(""); 1270 *noncecount = 0; 1271 *cnonce = 0; 1272 } else { 1273 c->nc++; 1274 sprintf(noncecount, "%08x", c->nc); 1275 /* We don't try very hard with the cnonce ... */ 1276 sprintf(cnonce, "%x%lx", getpid(), (unsigned long)time(0)); 1277 } 1278 1279 HASHHEX HA1; 1280 DigestCalcHA1(c->algo, parms->user, c->realm, 1281 parms->password, c->nonce, cnonce, HA1); 1282 DEBUG(fprintf(stderr, "HA1: [%s]\n", HA1)); 1283 HASHHEX digest; 1284 DigestCalcResponse(HA1, c->nonce, noncecount, cnonce, c->qop, 1285 "GET", url->doc, "", digest); 1286 1287 if (c->qop[0]) { 1288 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\"," 1289 "nonce=\"%s\",uri=\"%s\",response=\"%s\"," 1290 "qop=\"auth\", cnonce=\"%s\", nc=%s%s", 1291 hdr, parms->user, c->realm, 1292 c->nonce, url->doc, digest, 1293 cnonce, noncecount, options); 1294 } else { 1295 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\"," 1296 "nonce=\"%s\",uri=\"%s\",response=\"%s\"%s", 1297 hdr, parms->user, c->realm, 1298 c->nonce, url->doc, digest, options); 1299 } 1300 if (options) 1301 free(options); 1302 return (r); 1303 } 1304 1305 /* 1306 * Encode username and password 1307 */ 1308 static int 1309 http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd) 1310 { 1311 char *upw, *auth; 1312 int r; 1313 1314 DEBUG(fprintf(stderr, "basic: usr: [%s]\n", usr)); 1315 DEBUG(fprintf(stderr, "basic: pwd: [%s]\n", pwd)); 1316 if (asprintf(&upw, "%s:%s", usr, pwd) == -1) 1317 return (-1); 1318 auth = http_base64(upw); 1319 free(upw); 1320 if (auth == NULL) 1321 return (-1); 1322 r = http_cmd(conn, "%s: Basic %s", hdr, auth); 1323 free(auth); 1324 return (r); 1325 } 1326 1327 /* 1328 * Chose the challenge to answer and call the appropriate routine to 1329 * produce the header. 1330 */ 1331 static int 1332 http_authorize(conn_t *conn, const char *hdr, http_auth_challenges_t *cs, 1333 http_auth_params_t *parms, struct url *url) 1334 { 1335 http_auth_challenge_t *digest = NULL; 1336 int i; 1337 1338 /* If user or pass are null we're not happy */ 1339 if (!parms->user || !parms->password) { 1340 DEBUG(fprintf(stderr, "NULL usr or pass\n")); 1341 return (-1); 1342 } 1343 1344 /* Look for a Digest */ 1345 for (i = 0; i < cs->count; i++) { 1346 if (cs->challenges[i]->scheme == HTTPAS_DIGEST) 1347 digest = cs->challenges[i]; 1348 } 1349 1350 /* Error if "Digest" was specified and there is no Digest challenge */ 1351 if (!digest && (parms->scheme && 1352 !strcasecmp(parms->scheme, "digest"))) { 1353 DEBUG(fprintf(stderr, 1354 "Digest auth in env, not supported by peer\n")); 1355 return (-1); 1356 } 1357 /* 1358 * If "basic" was specified in the environment, or there is no Digest 1359 * challenge, do the basic thing. Don't need a challenge for this, 1360 * so no need to check basic!=NULL 1361 */ 1362 if (!digest || (parms->scheme && !strcasecmp(parms->scheme,"basic"))) 1363 return (http_basic_auth(conn,hdr,parms->user,parms->password)); 1364 1365 /* Else, prefer digest. We just checked that it's not NULL */ 1366 return (http_digest_auth(conn, hdr, digest, parms, url)); 1367 } 1368 1369 /***************************************************************************** 1370 * Helper functions for connecting to a server or proxy 1371 */ 1372 1373 /* 1374 * Connect to the correct HTTP server or proxy. 1375 */ 1376 static conn_t * 1377 http_connect(struct url *URL, struct url *purl, const char *flags) 1378 { 1379 struct url *curl; 1380 conn_t *conn; 1381 hdr_t h; 1382 http_headerbuf_t headerbuf; 1383 const char *p; 1384 int verbose; 1385 int af, val; 1386 int serrno; 1387 1388 #ifdef INET6 1389 af = AF_UNSPEC; 1390 #else 1391 af = AF_INET; 1392 #endif 1393 1394 verbose = CHECK_FLAG('v'); 1395 if (CHECK_FLAG('4')) 1396 af = AF_INET; 1397 #ifdef INET6 1398 else if (CHECK_FLAG('6')) 1399 af = AF_INET6; 1400 #endif 1401 1402 curl = (purl != NULL) ? purl : URL; 1403 1404 if ((conn = fetch_connect(curl->host, curl->port, af, verbose)) == NULL) 1405 /* fetch_connect() has already set an error code */ 1406 return (NULL); 1407 init_http_headerbuf(&headerbuf); 1408 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 && purl) { 1409 http_cmd(conn, "CONNECT %s:%d HTTP/1.1", 1410 URL->host, URL->port); 1411 http_cmd(conn, "Host: %s:%d", 1412 URL->host, URL->port); 1413 http_cmd(conn, ""); 1414 if (http_get_reply(conn) != HTTP_OK) { 1415 http_seterr(conn->err); 1416 goto ouch; 1417 } 1418 /* Read and discard the rest of the proxy response */ 1419 if (fetch_getln(conn) < 0) { 1420 fetch_syserr(); 1421 goto ouch; 1422 } 1423 do { 1424 switch ((h = http_next_header(conn, &headerbuf, &p))) { 1425 case hdr_syserror: 1426 fetch_syserr(); 1427 goto ouch; 1428 case hdr_error: 1429 http_seterr(HTTP_PROTOCOL_ERROR); 1430 goto ouch; 1431 default: 1432 /* ignore */ ; 1433 } 1434 } while (h < hdr_end); 1435 } 1436 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 && 1437 fetch_ssl(conn, URL, verbose) == -1) { 1438 /* grrr */ 1439 errno = EAUTH; 1440 fetch_syserr(); 1441 goto ouch; 1442 } 1443 1444 val = 1; 1445 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val)); 1446 1447 clean_http_headerbuf(&headerbuf); 1448 return (conn); 1449 ouch: 1450 serrno = errno; 1451 clean_http_headerbuf(&headerbuf); 1452 fetch_close(conn); 1453 errno = serrno; 1454 return (NULL); 1455 } 1456 1457 static struct url * 1458 http_get_proxy(struct url * url, const char *flags) 1459 { 1460 struct url *purl; 1461 char *p; 1462 1463 if (flags != NULL && strchr(flags, 'd') != NULL) 1464 return (NULL); 1465 if (fetch_no_proxy_match(url->host)) 1466 return (NULL); 1467 if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) && 1468 *p && (purl = fetchParseURL(p))) { 1469 if (!*purl->scheme) 1470 strcpy(purl->scheme, SCHEME_HTTP); 1471 if (!purl->port) 1472 purl->port = fetch_default_proxy_port(purl->scheme); 1473 if (strcasecmp(purl->scheme, SCHEME_HTTP) == 0) 1474 return (purl); 1475 fetchFreeURL(purl); 1476 } 1477 return (NULL); 1478 } 1479 1480 static void 1481 http_print_html(FILE *out, FILE *in) 1482 { 1483 size_t len; 1484 char *line, *p, *q; 1485 int comment, tag; 1486 1487 comment = tag = 0; 1488 while ((line = fgetln(in, &len)) != NULL) { 1489 while (len && isspace((unsigned char)line[len - 1])) 1490 --len; 1491 for (p = q = line; q < line + len; ++q) { 1492 if (comment && *q == '-') { 1493 if (q + 2 < line + len && 1494 strcmp(q, "-->") == 0) { 1495 tag = comment = 0; 1496 q += 2; 1497 } 1498 } else if (tag && !comment && *q == '>') { 1499 p = q + 1; 1500 tag = 0; 1501 } else if (!tag && *q == '<') { 1502 if (q > p) 1503 fwrite(p, q - p, 1, out); 1504 tag = 1; 1505 if (q + 3 < line + len && 1506 strcmp(q, "<!--") == 0) { 1507 comment = 1; 1508 q += 3; 1509 } 1510 } 1511 } 1512 if (!tag && q > p) 1513 fwrite(p, q - p, 1, out); 1514 fputc('\n', out); 1515 } 1516 } 1517 1518 1519 /***************************************************************************** 1520 * Core 1521 */ 1522 1523 FILE * 1524 http_request(struct url *URL, const char *op, struct url_stat *us, 1525 struct url *purl, const char *flags) 1526 { 1527 1528 return (http_request_body(URL, op, us, purl, flags, NULL, NULL)); 1529 } 1530 1531 /* 1532 * Send a request and process the reply 1533 * 1534 * XXX This function is way too long, the do..while loop should be split 1535 * XXX off into a separate function. 1536 */ 1537 FILE * 1538 http_request_body(struct url *URL, const char *op, struct url_stat *us, 1539 struct url *purl, const char *flags, const char *content_type, 1540 const char *body) 1541 { 1542 char timebuf[80]; 1543 char hbuf[MAXHOSTNAMELEN + 7], *host; 1544 conn_t *conn; 1545 struct url *url, *new; 1546 int chunked, direct, ims, noredirect, verbose; 1547 int e, i, n, val; 1548 off_t offset, clength, length, size; 1549 time_t mtime; 1550 const char *p; 1551 FILE *f; 1552 hdr_t h; 1553 struct tm *timestruct; 1554 http_headerbuf_t headerbuf; 1555 http_auth_challenges_t server_challenges; 1556 http_auth_challenges_t proxy_challenges; 1557 size_t body_len; 1558 1559 /* The following calls don't allocate anything */ 1560 init_http_headerbuf(&headerbuf); 1561 init_http_auth_challenges(&server_challenges); 1562 init_http_auth_challenges(&proxy_challenges); 1563 1564 direct = CHECK_FLAG('d'); 1565 noredirect = CHECK_FLAG('A'); 1566 verbose = CHECK_FLAG('v'); 1567 ims = CHECK_FLAG('i'); 1568 1569 if (direct && purl) { 1570 fetchFreeURL(purl); 1571 purl = NULL; 1572 } 1573 1574 /* try the provided URL first */ 1575 url = URL; 1576 1577 n = MAX_REDIRECT; 1578 i = 0; 1579 1580 e = HTTP_PROTOCOL_ERROR; 1581 do { 1582 new = NULL; 1583 chunked = 0; 1584 offset = 0; 1585 clength = -1; 1586 length = -1; 1587 size = -1; 1588 mtime = 0; 1589 1590 /* check port */ 1591 if (!url->port) 1592 url->port = fetch_default_port(url->scheme); 1593 1594 /* were we redirected to an FTP URL? */ 1595 if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) { 1596 if (strcmp(op, "GET") == 0) 1597 return (ftp_request(url, "RETR", us, purl, flags)); 1598 else if (strcmp(op, "HEAD") == 0) 1599 return (ftp_request(url, "STAT", us, purl, flags)); 1600 } 1601 1602 /* connect to server or proxy */ 1603 if ((conn = http_connect(url, purl, flags)) == NULL) 1604 goto ouch; 1605 1606 host = url->host; 1607 #ifdef INET6 1608 if (strchr(url->host, ':')) { 1609 snprintf(hbuf, sizeof(hbuf), "[%s]", url->host); 1610 host = hbuf; 1611 } 1612 #endif 1613 if (url->port != fetch_default_port(url->scheme)) { 1614 if (host != hbuf) { 1615 strcpy(hbuf, host); 1616 host = hbuf; 1617 } 1618 snprintf(hbuf + strlen(hbuf), 1619 sizeof(hbuf) - strlen(hbuf), ":%d", url->port); 1620 } 1621 1622 /* send request */ 1623 if (verbose) 1624 fetch_info("requesting %s://%s%s", 1625 url->scheme, host, url->doc); 1626 if (purl && strcasecmp(URL->scheme, SCHEME_HTTPS) != 0) { 1627 http_cmd(conn, "%s %s://%s%s HTTP/1.1", 1628 op, url->scheme, host, url->doc); 1629 } else { 1630 http_cmd(conn, "%s %s HTTP/1.1", 1631 op, url->doc); 1632 } 1633 1634 if (ims && url->ims_time) { 1635 timestruct = gmtime((time_t *)&url->ims_time); 1636 (void)strftime(timebuf, 80, "%a, %d %b %Y %T GMT", 1637 timestruct); 1638 if (verbose) 1639 fetch_info("If-Modified-Since: %s", timebuf); 1640 http_cmd(conn, "If-Modified-Since: %s", timebuf); 1641 } 1642 /* virtual host */ 1643 http_cmd(conn, "Host: %s", host); 1644 1645 /* 1646 * Proxy authorization: we only send auth after we received 1647 * a 407 error. We do not first try basic anyway (changed 1648 * when support was added for digest-auth) 1649 */ 1650 if (purl && proxy_challenges.valid) { 1651 http_auth_params_t aparams; 1652 init_http_auth_params(&aparams); 1653 if (*purl->user || *purl->pwd) { 1654 aparams.user = strdup(purl->user); 1655 aparams.password = strdup(purl->pwd); 1656 } else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL && 1657 *p != '\0') { 1658 if (http_authfromenv(p, &aparams) < 0) { 1659 http_seterr(HTTP_NEED_PROXY_AUTH); 1660 goto ouch; 1661 } 1662 } else if (fetch_netrc_auth(purl) == 0) { 1663 aparams.user = strdup(purl->user); 1664 aparams.password = strdup(purl->pwd); 1665 } 1666 http_authorize(conn, "Proxy-Authorization", 1667 &proxy_challenges, &aparams, url); 1668 clean_http_auth_params(&aparams); 1669 } 1670 1671 /* 1672 * Server authorization: we never send "a priori" 1673 * Basic auth, which used to be done if user/pass were 1674 * set in the url. This would be weird because we'd send the 1675 * password in the clear even if Digest is finally to be 1676 * used (it would have made more sense for the 1677 * pre-digest version to do this when Basic was specified 1678 * in the environment) 1679 */ 1680 if (server_challenges.valid) { 1681 http_auth_params_t aparams; 1682 init_http_auth_params(&aparams); 1683 if (*url->user || *url->pwd) { 1684 aparams.user = strdup(url->user); 1685 aparams.password = strdup(url->pwd); 1686 } else if ((p = getenv("HTTP_AUTH")) != NULL && 1687 *p != '\0') { 1688 if (http_authfromenv(p, &aparams) < 0) { 1689 http_seterr(HTTP_NEED_AUTH); 1690 goto ouch; 1691 } 1692 } else if (fetch_netrc_auth(url) == 0) { 1693 aparams.user = strdup(url->user); 1694 aparams.password = strdup(url->pwd); 1695 } else if (fetchAuthMethod && 1696 fetchAuthMethod(url) == 0) { 1697 aparams.user = strdup(url->user); 1698 aparams.password = strdup(url->pwd); 1699 } else { 1700 http_seterr(HTTP_NEED_AUTH); 1701 goto ouch; 1702 } 1703 http_authorize(conn, "Authorization", 1704 &server_challenges, &aparams, url); 1705 clean_http_auth_params(&aparams); 1706 } 1707 1708 /* other headers */ 1709 if ((p = getenv("HTTP_ACCEPT")) != NULL) { 1710 if (*p != '\0') 1711 http_cmd(conn, "Accept: %s", p); 1712 } else { 1713 http_cmd(conn, "Accept: */*"); 1714 } 1715 if ((p = getenv("HTTP_REFERER")) != NULL && *p != '\0') { 1716 if (strcasecmp(p, "auto") == 0) 1717 http_cmd(conn, "Referer: %s://%s%s", 1718 url->scheme, host, url->doc); 1719 else 1720 http_cmd(conn, "Referer: %s", p); 1721 } 1722 if ((p = getenv("HTTP_USER_AGENT")) != NULL) { 1723 /* no User-Agent if defined but empty */ 1724 if (*p != '\0') 1725 http_cmd(conn, "User-Agent: %s", p); 1726 } else { 1727 /* default User-Agent */ 1728 http_cmd(conn, "User-Agent: %s " _LIBFETCH_VER, 1729 getprogname()); 1730 } 1731 if (url->offset > 0) 1732 http_cmd(conn, "Range: bytes=%lld-", (long long)url->offset); 1733 http_cmd(conn, "Connection: close"); 1734 1735 if (body) { 1736 body_len = strlen(body); 1737 http_cmd(conn, "Content-Length: %zu", body_len); 1738 if (content_type != NULL) 1739 http_cmd(conn, "Content-Type: %s", content_type); 1740 } 1741 1742 http_cmd(conn, ""); 1743 1744 if (body) 1745 fetch_write(conn, body, body_len); 1746 1747 /* 1748 * Force the queued request to be dispatched. Normally, one 1749 * would do this with shutdown(2) but squid proxies can be 1750 * configured to disallow such half-closed connections. To 1751 * be compatible with such configurations, fiddle with socket 1752 * options to force the pending data to be written. 1753 */ 1754 val = 0; 1755 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, 1756 sizeof(val)); 1757 val = 1; 1758 setsockopt(conn->sd, IPPROTO_TCP, TCP_NODELAY, &val, 1759 sizeof(val)); 1760 1761 /* get reply */ 1762 switch (http_get_reply(conn)) { 1763 case HTTP_OK: 1764 case HTTP_PARTIAL: 1765 case HTTP_NOT_MODIFIED: 1766 /* fine */ 1767 break; 1768 case HTTP_MOVED_PERM: 1769 case HTTP_MOVED_TEMP: 1770 case HTTP_SEE_OTHER: 1771 case HTTP_USE_PROXY: 1772 /* 1773 * Not so fine, but we still have to read the 1774 * headers to get the new location. 1775 */ 1776 break; 1777 case HTTP_NEED_AUTH: 1778 if (server_challenges.valid) { 1779 /* 1780 * We already sent out authorization code, 1781 * so there's nothing more we can do. 1782 */ 1783 http_seterr(conn->err); 1784 goto ouch; 1785 } 1786 /* try again, but send the password this time */ 1787 if (verbose) 1788 fetch_info("server requires authorization"); 1789 break; 1790 case HTTP_NEED_PROXY_AUTH: 1791 if (proxy_challenges.valid) { 1792 /* 1793 * We already sent our proxy 1794 * authorization code, so there's 1795 * nothing more we can do. */ 1796 http_seterr(conn->err); 1797 goto ouch; 1798 } 1799 /* try again, but send the password this time */ 1800 if (verbose) 1801 fetch_info("proxy requires authorization"); 1802 break; 1803 case HTTP_BAD_RANGE: 1804 /* 1805 * This can happen if we ask for 0 bytes because 1806 * we already have the whole file. Consider this 1807 * a success for now, and check sizes later. 1808 */ 1809 break; 1810 case HTTP_PROTOCOL_ERROR: 1811 /* fall through */ 1812 case -1: 1813 fetch_syserr(); 1814 goto ouch; 1815 default: 1816 http_seterr(conn->err); 1817 if (!verbose) 1818 goto ouch; 1819 /* fall through so we can get the full error message */ 1820 } 1821 1822 /* get headers. http_next_header expects one line readahead */ 1823 if (fetch_getln(conn) == -1) { 1824 fetch_syserr(); 1825 goto ouch; 1826 } 1827 do { 1828 switch ((h = http_next_header(conn, &headerbuf, &p))) { 1829 case hdr_syserror: 1830 fetch_syserr(); 1831 goto ouch; 1832 case hdr_error: 1833 http_seterr(HTTP_PROTOCOL_ERROR); 1834 goto ouch; 1835 case hdr_content_length: 1836 http_parse_length(p, &clength); 1837 break; 1838 case hdr_content_range: 1839 http_parse_range(p, &offset, &length, &size); 1840 break; 1841 case hdr_last_modified: 1842 http_parse_mtime(p, &mtime); 1843 break; 1844 case hdr_location: 1845 if (!HTTP_REDIRECT(conn->err)) 1846 break; 1847 /* 1848 * if the A flag is set, we don't follow 1849 * temporary redirects. 1850 */ 1851 if (noredirect && 1852 conn->err != HTTP_MOVED_PERM && 1853 conn->err != HTTP_PERM_REDIRECT && 1854 conn->err != HTTP_USE_PROXY) { 1855 n = 1; 1856 break; 1857 } 1858 if (new) 1859 free(new); 1860 if (verbose) 1861 fetch_info("%d redirect to %s", conn->err, p); 1862 if (*p == '/') 1863 /* absolute path */ 1864 new = fetchMakeURL(url->scheme, url->host, url->port, p, 1865 url->user, url->pwd); 1866 else 1867 new = fetchParseURL(p); 1868 if (new == NULL) { 1869 /* XXX should set an error code */ 1870 DEBUG(fprintf(stderr, "failed to parse new URL\n")); 1871 goto ouch; 1872 } 1873 1874 /* Only copy credentials if the host matches */ 1875 if (!strcmp(new->host, url->host) && !*new->user && !*new->pwd) { 1876 strcpy(new->user, url->user); 1877 strcpy(new->pwd, url->pwd); 1878 } 1879 new->offset = url->offset; 1880 new->length = url->length; 1881 break; 1882 case hdr_transfer_encoding: 1883 /* XXX weak test*/ 1884 chunked = (strcasecmp(p, "chunked") == 0); 1885 break; 1886 case hdr_www_authenticate: 1887 if (conn->err != HTTP_NEED_AUTH) 1888 break; 1889 if (http_parse_authenticate(p, &server_challenges) == 0) 1890 ++n; 1891 break; 1892 case hdr_proxy_authenticate: 1893 if (conn->err != HTTP_NEED_PROXY_AUTH) 1894 break; 1895 if (http_parse_authenticate(p, &proxy_challenges) == 0) 1896 ++n; 1897 break; 1898 case hdr_end: 1899 /* fall through */ 1900 case hdr_unknown: 1901 /* ignore */ 1902 break; 1903 } 1904 } while (h > hdr_end); 1905 1906 /* we need to provide authentication */ 1907 if (conn->err == HTTP_NEED_AUTH || 1908 conn->err == HTTP_NEED_PROXY_AUTH) { 1909 e = conn->err; 1910 if ((conn->err == HTTP_NEED_AUTH && 1911 !server_challenges.valid) || 1912 (conn->err == HTTP_NEED_PROXY_AUTH && 1913 !proxy_challenges.valid)) { 1914 /* 401/7 but no www/proxy-authenticate ?? */ 1915 DEBUG(fprintf(stderr, "401/7 and no auth header\n")); 1916 goto ouch; 1917 } 1918 fetch_close(conn); 1919 conn = NULL; 1920 continue; 1921 } 1922 1923 /* requested range not satisfiable */ 1924 if (conn->err == HTTP_BAD_RANGE) { 1925 if (url->offset == size && url->length == 0) { 1926 /* asked for 0 bytes; fake it */ 1927 offset = url->offset; 1928 clength = -1; 1929 conn->err = HTTP_OK; 1930 break; 1931 } else { 1932 http_seterr(conn->err); 1933 goto ouch; 1934 } 1935 } 1936 1937 /* we have a hit or an error */ 1938 if (conn->err == HTTP_OK 1939 || conn->err == HTTP_NOT_MODIFIED 1940 || conn->err == HTTP_PARTIAL 1941 || HTTP_ERROR(conn->err)) 1942 break; 1943 1944 /* all other cases: we got a redirect */ 1945 e = conn->err; 1946 clean_http_auth_challenges(&server_challenges); 1947 fetch_close(conn); 1948 conn = NULL; 1949 if (!new) { 1950 DEBUG(fprintf(stderr, "redirect with no new location\n")); 1951 break; 1952 } 1953 if (url != URL) 1954 fetchFreeURL(url); 1955 url = new; 1956 } while (++i < n); 1957 1958 /* we failed, or ran out of retries */ 1959 if (conn == NULL) { 1960 http_seterr(e); 1961 goto ouch; 1962 } 1963 1964 DEBUG(fprintf(stderr, "offset %lld, length %lld," 1965 " size %lld, clength %lld\n", 1966 (long long)offset, (long long)length, 1967 (long long)size, (long long)clength)); 1968 1969 if (conn->err == HTTP_NOT_MODIFIED) { 1970 http_seterr(HTTP_NOT_MODIFIED); 1971 return (NULL); 1972 } 1973 1974 /* check for inconsistencies */ 1975 if (clength != -1 && length != -1 && clength != length) { 1976 http_seterr(HTTP_PROTOCOL_ERROR); 1977 goto ouch; 1978 } 1979 if (clength == -1) 1980 clength = length; 1981 if (clength != -1) 1982 length = offset + clength; 1983 if (length != -1 && size != -1 && length != size) { 1984 http_seterr(HTTP_PROTOCOL_ERROR); 1985 goto ouch; 1986 } 1987 if (size == -1) 1988 size = length; 1989 1990 /* fill in stats */ 1991 if (us) { 1992 us->size = size; 1993 us->atime = us->mtime = mtime; 1994 } 1995 1996 /* too far? */ 1997 if (URL->offset > 0 && offset > URL->offset) { 1998 http_seterr(HTTP_PROTOCOL_ERROR); 1999 goto ouch; 2000 } 2001 2002 /* report back real offset and size */ 2003 URL->offset = offset; 2004 URL->length = clength; 2005 2006 /* wrap it up in a FILE */ 2007 if ((f = http_funopen(conn, chunked)) == NULL) { 2008 fetch_syserr(); 2009 goto ouch; 2010 } 2011 2012 if (url != URL) 2013 fetchFreeURL(url); 2014 if (purl) 2015 fetchFreeURL(purl); 2016 2017 if (HTTP_ERROR(conn->err)) { 2018 http_print_html(stderr, f); 2019 fclose(f); 2020 f = NULL; 2021 } 2022 clean_http_headerbuf(&headerbuf); 2023 clean_http_auth_challenges(&server_challenges); 2024 clean_http_auth_challenges(&proxy_challenges); 2025 return (f); 2026 2027 ouch: 2028 if (url != URL) 2029 fetchFreeURL(url); 2030 if (purl) 2031 fetchFreeURL(purl); 2032 if (conn != NULL) 2033 fetch_close(conn); 2034 clean_http_headerbuf(&headerbuf); 2035 clean_http_auth_challenges(&server_challenges); 2036 clean_http_auth_challenges(&proxy_challenges); 2037 return (NULL); 2038 } 2039 2040 2041 /***************************************************************************** 2042 * Entry points 2043 */ 2044 2045 /* 2046 * Retrieve and stat a file by HTTP 2047 */ 2048 FILE * 2049 fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags) 2050 { 2051 return (http_request(URL, "GET", us, http_get_proxy(URL, flags), flags)); 2052 } 2053 2054 /* 2055 * Retrieve a file by HTTP 2056 */ 2057 FILE * 2058 fetchGetHTTP(struct url *URL, const char *flags) 2059 { 2060 return (fetchXGetHTTP(URL, NULL, flags)); 2061 } 2062 2063 /* 2064 * Store a file by HTTP 2065 */ 2066 FILE * 2067 fetchPutHTTP(struct url *URL __unused, const char *flags __unused) 2068 { 2069 warnx("fetchPutHTTP(): not implemented"); 2070 return (NULL); 2071 } 2072 2073 /* 2074 * Get an HTTP document's metadata 2075 */ 2076 int 2077 fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags) 2078 { 2079 FILE *f; 2080 2081 f = http_request(URL, "HEAD", us, http_get_proxy(URL, flags), flags); 2082 if (f == NULL) 2083 return (-1); 2084 fclose(f); 2085 return (0); 2086 } 2087 2088 /* 2089 * List a directory 2090 */ 2091 struct url_ent * 2092 fetchListHTTP(struct url *url __unused, const char *flags __unused) 2093 { 2094 warnx("fetchListHTTP(): not implemented"); 2095 return (NULL); 2096 } 2097 2098 FILE * 2099 fetchReqHTTP(struct url *URL, const char *method, const char *flags, 2100 const char *content_type, const char *body) 2101 { 2102 2103 return (http_request_body(URL, method, NULL, http_get_proxy(URL, flags), 2104 flags, content_type, body)); 2105 } 2106