1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 2000-2014 Dag-Erling Smørgrav 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer 12 * in this position and unchanged. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. The name of the author may not be used to endorse or promote products 17 * derived from this software without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 20 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 21 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 22 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 23 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 24 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 25 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 26 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 27 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 28 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 29 */ 30 31 #include <sys/cdefs.h> 32 __FBSDID("$FreeBSD$"); 33 34 /* 35 * The following copyright applies to the base64 code: 36 * 37 *- 38 * Copyright 1997 Massachusetts Institute of Technology 39 * 40 * Permission to use, copy, modify, and distribute this software and 41 * its documentation for any purpose and without fee is hereby 42 * granted, provided that both the above copyright notice and this 43 * permission notice appear in all copies, that both the above 44 * copyright notice and this permission notice appear in all 45 * supporting documentation, and that the name of M.I.T. not be used 46 * in advertising or publicity pertaining to distribution of the 47 * software without specific, written prior permission. M.I.T. makes 48 * no representations about the suitability of this software for any 49 * purpose. It is provided "as is" without express or implied 50 * warranty. 51 * 52 * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''. M.I.T. DISCLAIMS 53 * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE, 54 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 55 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT 56 * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 57 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 58 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF 59 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND 60 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 61 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 62 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 63 * SUCH DAMAGE. 64 */ 65 66 #include <sys/param.h> 67 #include <sys/socket.h> 68 #include <sys/time.h> 69 70 #include <ctype.h> 71 #include <err.h> 72 #include <errno.h> 73 #include <locale.h> 74 #include <netdb.h> 75 #include <stdarg.h> 76 #include <stdbool.h> 77 #include <stdio.h> 78 #include <stdlib.h> 79 #include <string.h> 80 #include <time.h> 81 #include <unistd.h> 82 83 #ifdef WITH_SSL 84 #include <openssl/md5.h> 85 #define MD5Init(c) MD5_Init(c) 86 #define MD5Update(c, data, len) MD5_Update(c, data, len) 87 #define MD5Final(md, c) MD5_Final(md, c) 88 #else 89 #include <md5.h> 90 #endif 91 92 #include <netinet/in.h> 93 #include <netinet/tcp.h> 94 95 #include "fetch.h" 96 #include "common.h" 97 #include "httperr.h" 98 99 /* Maximum number of redirects to follow */ 100 #define MAX_REDIRECT 20 101 102 /* Symbolic names for reply codes we care about */ 103 #define HTTP_OK 200 104 #define HTTP_PARTIAL 206 105 #define HTTP_MOVED_PERM 301 106 #define HTTP_MOVED_TEMP 302 107 #define HTTP_SEE_OTHER 303 108 #define HTTP_NOT_MODIFIED 304 109 #define HTTP_USE_PROXY 305 110 #define HTTP_TEMP_REDIRECT 307 111 #define HTTP_PERM_REDIRECT 308 112 #define HTTP_NEED_AUTH 401 113 #define HTTP_NEED_PROXY_AUTH 407 114 #define HTTP_BAD_RANGE 416 115 #define HTTP_PROTOCOL_ERROR 999 116 117 #define HTTP_REDIRECT(xyz) ((xyz) == HTTP_MOVED_PERM \ 118 || (xyz) == HTTP_MOVED_TEMP \ 119 || (xyz) == HTTP_TEMP_REDIRECT \ 120 || (xyz) == HTTP_PERM_REDIRECT \ 121 || (xyz) == HTTP_USE_PROXY \ 122 || (xyz) == HTTP_SEE_OTHER) 123 124 #define HTTP_ERROR(xyz) ((xyz) >= 400 && (xyz) <= 599) 125 126 127 /***************************************************************************** 128 * I/O functions for decoding chunked streams 129 */ 130 131 struct httpio 132 { 133 conn_t *conn; /* connection */ 134 int chunked; /* chunked mode */ 135 char *buf; /* chunk buffer */ 136 size_t bufsize; /* size of chunk buffer */ 137 size_t buflen; /* amount of data currently in buffer */ 138 size_t bufpos; /* current read offset in buffer */ 139 int eof; /* end-of-file flag */ 140 int error; /* error flag */ 141 size_t chunksize; /* remaining size of current chunk */ 142 #ifndef NDEBUG 143 size_t total; 144 #endif 145 }; 146 147 /* 148 * Get next chunk header 149 */ 150 static int 151 http_new_chunk(struct httpio *io) 152 { 153 char *p; 154 155 if (fetch_getln(io->conn) == -1) 156 return (-1); 157 158 if (io->conn->buflen < 2 || !isxdigit((unsigned char)*io->conn->buf)) 159 return (-1); 160 161 for (p = io->conn->buf; *p && !isspace((unsigned char)*p); ++p) { 162 if (*p == ';') 163 break; 164 if (!isxdigit((unsigned char)*p)) 165 return (-1); 166 if (isdigit((unsigned char)*p)) { 167 io->chunksize = io->chunksize * 16 + 168 *p - '0'; 169 } else { 170 io->chunksize = io->chunksize * 16 + 171 10 + tolower((unsigned char)*p) - 'a'; 172 } 173 } 174 175 #ifndef NDEBUG 176 if (fetchDebug) { 177 io->total += io->chunksize; 178 if (io->chunksize == 0) 179 fprintf(stderr, "%s(): end of last chunk\n", __func__); 180 else 181 fprintf(stderr, "%s(): new chunk: %lu (%lu)\n", 182 __func__, (unsigned long)io->chunksize, 183 (unsigned long)io->total); 184 } 185 #endif 186 187 return (io->chunksize); 188 } 189 190 /* 191 * Grow the input buffer to at least len bytes 192 */ 193 static inline int 194 http_growbuf(struct httpio *io, size_t len) 195 { 196 char *tmp; 197 198 if (io->bufsize >= len) 199 return (0); 200 201 if ((tmp = realloc(io->buf, len)) == NULL) 202 return (-1); 203 io->buf = tmp; 204 io->bufsize = len; 205 return (0); 206 } 207 208 /* 209 * Fill the input buffer, do chunk decoding on the fly 210 */ 211 static ssize_t 212 http_fillbuf(struct httpio *io, size_t len) 213 { 214 ssize_t nbytes; 215 char ch; 216 217 if (io->error) 218 return (-1); 219 if (io->eof) 220 return (0); 221 222 /* not chunked: just fetch the requested amount */ 223 if (io->chunked == 0) { 224 if (http_growbuf(io, len) == -1) 225 return (-1); 226 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) { 227 io->error = errno; 228 return (-1); 229 } 230 io->buflen = nbytes; 231 io->bufpos = 0; 232 return (io->buflen); 233 } 234 235 /* chunked, but we ran out: get the next chunk header */ 236 if (io->chunksize == 0) { 237 switch (http_new_chunk(io)) { 238 case -1: 239 io->error = EPROTO; 240 return (-1); 241 case 0: 242 io->eof = 1; 243 return (0); 244 } 245 } 246 247 /* fetch the requested amount, but no more than the current chunk */ 248 if (len > io->chunksize) 249 len = io->chunksize; 250 if (http_growbuf(io, len) == -1) 251 return (-1); 252 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) { 253 io->error = errno; 254 return (-1); 255 } 256 io->bufpos = 0; 257 io->buflen = nbytes; 258 io->chunksize -= nbytes; 259 260 if (io->chunksize == 0) { 261 if (fetch_read(io->conn, &ch, 1) != 1 || ch != '\r' || 262 fetch_read(io->conn, &ch, 1) != 1 || ch != '\n') 263 return (-1); 264 } 265 266 return (io->buflen); 267 } 268 269 /* 270 * Read function 271 */ 272 static int 273 http_readfn(void *v, char *buf, int len) 274 { 275 struct httpio *io = (struct httpio *)v; 276 int rlen; 277 278 if (io->error) 279 return (-1); 280 if (io->eof) 281 return (0); 282 283 /* empty buffer */ 284 if (!io->buf || io->bufpos == io->buflen) { 285 if ((rlen = http_fillbuf(io, len)) < 0) { 286 if ((errno = io->error) == EINTR) 287 io->error = 0; 288 return (-1); 289 } else if (rlen == 0) { 290 return (0); 291 } 292 } 293 294 rlen = io->buflen - io->bufpos; 295 if (len < rlen) 296 rlen = len; 297 memcpy(buf, io->buf + io->bufpos, rlen); 298 io->bufpos += rlen; 299 return (rlen); 300 } 301 302 /* 303 * Write function 304 */ 305 static int 306 http_writefn(void *v, const char *buf, int len) 307 { 308 struct httpio *io = (struct httpio *)v; 309 310 return (fetch_write(io->conn, buf, len)); 311 } 312 313 /* 314 * Close function 315 */ 316 static int 317 http_closefn(void *v) 318 { 319 struct httpio *io = (struct httpio *)v; 320 int r; 321 322 r = fetch_close(io->conn); 323 if (io->buf) 324 free(io->buf); 325 free(io); 326 return (r); 327 } 328 329 /* 330 * Wrap a file descriptor up 331 */ 332 static FILE * 333 http_funopen(conn_t *conn, int chunked) 334 { 335 struct httpio *io; 336 FILE *f; 337 338 if ((io = calloc(1, sizeof(*io))) == NULL) { 339 fetch_syserr(); 340 return (NULL); 341 } 342 io->conn = conn; 343 io->chunked = chunked; 344 f = funopen(io, http_readfn, http_writefn, NULL, http_closefn); 345 if (f == NULL) { 346 fetch_syserr(); 347 free(io); 348 return (NULL); 349 } 350 return (f); 351 } 352 353 354 /***************************************************************************** 355 * Helper functions for talking to the server and parsing its replies 356 */ 357 358 /* Header types */ 359 typedef enum { 360 hdr_syserror = -2, 361 hdr_error = -1, 362 hdr_end = 0, 363 hdr_unknown = 1, 364 hdr_content_length, 365 hdr_content_range, 366 hdr_last_modified, 367 hdr_location, 368 hdr_transfer_encoding, 369 hdr_www_authenticate, 370 hdr_proxy_authenticate, 371 } hdr_t; 372 373 /* Names of interesting headers */ 374 static struct { 375 hdr_t num; 376 const char *name; 377 } hdr_names[] = { 378 { hdr_content_length, "Content-Length" }, 379 { hdr_content_range, "Content-Range" }, 380 { hdr_last_modified, "Last-Modified" }, 381 { hdr_location, "Location" }, 382 { hdr_transfer_encoding, "Transfer-Encoding" }, 383 { hdr_www_authenticate, "WWW-Authenticate" }, 384 { hdr_proxy_authenticate, "Proxy-Authenticate" }, 385 { hdr_unknown, NULL }, 386 }; 387 388 /* 389 * Send a formatted line; optionally echo to terminal 390 */ 391 static int 392 http_cmd(conn_t *conn, const char *fmt, ...) 393 { 394 va_list ap; 395 size_t len; 396 char *msg; 397 int r; 398 399 va_start(ap, fmt); 400 len = vasprintf(&msg, fmt, ap); 401 va_end(ap); 402 403 if (msg == NULL) { 404 errno = ENOMEM; 405 fetch_syserr(); 406 return (-1); 407 } 408 409 r = fetch_putln(conn, msg, len); 410 free(msg); 411 412 if (r == -1) { 413 fetch_syserr(); 414 return (-1); 415 } 416 417 return (0); 418 } 419 420 /* 421 * Get and parse status line 422 */ 423 static int 424 http_get_reply(conn_t *conn) 425 { 426 char *p; 427 428 if (fetch_getln(conn) == -1) 429 return (-1); 430 /* 431 * A valid status line looks like "HTTP/m.n xyz reason" where m 432 * and n are the major and minor protocol version numbers and xyz 433 * is the reply code. 434 * Unfortunately, there are servers out there (NCSA 1.5.1, to name 435 * just one) that do not send a version number, so we can't rely 436 * on finding one, but if we do, insist on it being 1.0 or 1.1. 437 * We don't care about the reason phrase. 438 */ 439 if (strncmp(conn->buf, "HTTP", 4) != 0) 440 return (HTTP_PROTOCOL_ERROR); 441 p = conn->buf + 4; 442 if (*p == '/') { 443 if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1')) 444 return (HTTP_PROTOCOL_ERROR); 445 p += 4; 446 } 447 if (*p != ' ' || 448 !isdigit((unsigned char)p[1]) || 449 !isdigit((unsigned char)p[2]) || 450 !isdigit((unsigned char)p[3])) 451 return (HTTP_PROTOCOL_ERROR); 452 453 conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0'); 454 return (conn->err); 455 } 456 457 /* 458 * Check a header; if the type matches the given string, return a pointer 459 * to the beginning of the value. 460 */ 461 static const char * 462 http_match(const char *str, const char *hdr) 463 { 464 while (*str && *hdr && 465 tolower((unsigned char)*str++) == tolower((unsigned char)*hdr++)) 466 /* nothing */; 467 if (*str || *hdr != ':') 468 return (NULL); 469 while (*hdr && isspace((unsigned char)*++hdr)) 470 /* nothing */; 471 return (hdr); 472 } 473 474 475 /* 476 * Get the next header and return the appropriate symbolic code. We 477 * need to read one line ahead for checking for a continuation line 478 * belonging to the current header (continuation lines start with 479 * white space). 480 * 481 * We get called with a fresh line already in the conn buffer, either 482 * from the previous http_next_header() invocation, or, the first 483 * time, from a fetch_getln() performed by our caller. 484 * 485 * This stops when we encounter an empty line (we dont read beyond the header 486 * area). 487 * 488 * Note that the "headerbuf" is just a place to return the result. Its 489 * contents are not used for the next call. This means that no cleanup 490 * is needed when ie doing another connection, just call the cleanup when 491 * fully done to deallocate memory. 492 */ 493 494 /* Limit the max number of continuation lines to some reasonable value */ 495 #define HTTP_MAX_CONT_LINES 10 496 497 /* Place into which to build a header from one or several lines */ 498 typedef struct { 499 char *buf; /* buffer */ 500 size_t bufsize; /* buffer size */ 501 size_t buflen; /* length of buffer contents */ 502 } http_headerbuf_t; 503 504 static void 505 init_http_headerbuf(http_headerbuf_t *buf) 506 { 507 buf->buf = NULL; 508 buf->bufsize = 0; 509 buf->buflen = 0; 510 } 511 512 static void 513 clean_http_headerbuf(http_headerbuf_t *buf) 514 { 515 if (buf->buf) 516 free(buf->buf); 517 init_http_headerbuf(buf); 518 } 519 520 /* Remove whitespace at the end of the buffer */ 521 static void 522 http_conn_trimright(conn_t *conn) 523 { 524 while (conn->buflen && 525 isspace((unsigned char)conn->buf[conn->buflen - 1])) 526 conn->buflen--; 527 conn->buf[conn->buflen] = '\0'; 528 } 529 530 static hdr_t 531 http_next_header(conn_t *conn, http_headerbuf_t *hbuf, const char **p) 532 { 533 unsigned int i, len; 534 535 /* 536 * Have to do the stripping here because of the first line. So 537 * it's done twice for the subsequent lines. No big deal 538 */ 539 http_conn_trimright(conn); 540 if (conn->buflen == 0) 541 return (hdr_end); 542 543 /* Copy the line to the headerbuf */ 544 if (hbuf->bufsize < conn->buflen + 1) { 545 if ((hbuf->buf = realloc(hbuf->buf, conn->buflen + 1)) == NULL) 546 return (hdr_syserror); 547 hbuf->bufsize = conn->buflen + 1; 548 } 549 strcpy(hbuf->buf, conn->buf); 550 hbuf->buflen = conn->buflen; 551 552 /* 553 * Fetch possible continuation lines. Stop at 1st non-continuation 554 * and leave it in the conn buffer 555 */ 556 for (i = 0; i < HTTP_MAX_CONT_LINES; i++) { 557 if (fetch_getln(conn) == -1) 558 return (hdr_syserror); 559 560 /* 561 * Note: we carry on the idea from the previous version 562 * that a pure whitespace line is equivalent to an empty 563 * one (so it's not continuation and will be handled when 564 * we are called next) 565 */ 566 http_conn_trimright(conn); 567 if (conn->buf[0] != ' ' && conn->buf[0] != "\t"[0]) 568 break; 569 570 /* Got a continuation line. Concatenate to previous */ 571 len = hbuf->buflen + conn->buflen; 572 if (hbuf->bufsize < len + 1) { 573 len *= 2; 574 if ((hbuf->buf = realloc(hbuf->buf, len + 1)) == NULL) 575 return (hdr_syserror); 576 hbuf->bufsize = len + 1; 577 } 578 strcpy(hbuf->buf + hbuf->buflen, conn->buf); 579 hbuf->buflen += conn->buflen; 580 } 581 582 /* 583 * We could check for malformed headers but we don't really care. 584 * A valid header starts with a token immediately followed by a 585 * colon; a token is any sequence of non-control, non-whitespace 586 * characters except "()<>@,;:\\\"{}". 587 */ 588 for (i = 0; hdr_names[i].num != hdr_unknown; i++) 589 if ((*p = http_match(hdr_names[i].name, hbuf->buf)) != NULL) 590 return (hdr_names[i].num); 591 592 return (hdr_unknown); 593 } 594 595 /************************** 596 * [Proxy-]Authenticate header parsing 597 */ 598 599 /* 600 * Read doublequote-delimited string into output buffer obuf (allocated 601 * by caller, whose responsibility it is to ensure that it's big enough) 602 * cp points to the first char after the initial '"' 603 * Handles \ quoting 604 * Returns pointer to the first char after the terminating double quote, or 605 * NULL for error. 606 */ 607 static const char * 608 http_parse_headerstring(const char *cp, char *obuf) 609 { 610 for (;;) { 611 switch (*cp) { 612 case 0: /* Unterminated string */ 613 *obuf = 0; 614 return (NULL); 615 case '"': /* Ending quote */ 616 *obuf = 0; 617 return (++cp); 618 case '\\': 619 if (*++cp == 0) { 620 *obuf = 0; 621 return (NULL); 622 } 623 /* FALLTHROUGH */ 624 default: 625 *obuf++ = *cp++; 626 } 627 } 628 } 629 630 /* Http auth challenge schemes */ 631 typedef enum {HTTPAS_UNKNOWN, HTTPAS_BASIC,HTTPAS_DIGEST} http_auth_schemes_t; 632 633 /* Data holder for a Basic or Digest challenge. */ 634 typedef struct { 635 http_auth_schemes_t scheme; 636 char *realm; 637 char *qop; 638 char *nonce; 639 char *opaque; 640 char *algo; 641 int stale; 642 int nc; /* Nonce count */ 643 } http_auth_challenge_t; 644 645 static void 646 init_http_auth_challenge(http_auth_challenge_t *b) 647 { 648 b->scheme = HTTPAS_UNKNOWN; 649 b->realm = b->qop = b->nonce = b->opaque = b->algo = NULL; 650 b->stale = b->nc = 0; 651 } 652 653 static void 654 clean_http_auth_challenge(http_auth_challenge_t *b) 655 { 656 if (b->realm) 657 free(b->realm); 658 if (b->qop) 659 free(b->qop); 660 if (b->nonce) 661 free(b->nonce); 662 if (b->opaque) 663 free(b->opaque); 664 if (b->algo) 665 free(b->algo); 666 init_http_auth_challenge(b); 667 } 668 669 /* Data holder for an array of challenges offered in an http response. */ 670 #define MAX_CHALLENGES 10 671 typedef struct { 672 http_auth_challenge_t *challenges[MAX_CHALLENGES]; 673 int count; /* Number of parsed challenges in the array */ 674 int valid; /* We did parse an authenticate header */ 675 } http_auth_challenges_t; 676 677 static void 678 init_http_auth_challenges(http_auth_challenges_t *cs) 679 { 680 int i; 681 for (i = 0; i < MAX_CHALLENGES; i++) 682 cs->challenges[i] = NULL; 683 cs->count = cs->valid = 0; 684 } 685 686 static void 687 clean_http_auth_challenges(http_auth_challenges_t *cs) 688 { 689 int i; 690 /* We rely on non-zero pointers being allocated, not on the count */ 691 for (i = 0; i < MAX_CHALLENGES; i++) { 692 if (cs->challenges[i] != NULL) { 693 clean_http_auth_challenge(cs->challenges[i]); 694 free(cs->challenges[i]); 695 } 696 } 697 init_http_auth_challenges(cs); 698 } 699 700 /* 701 * Enumeration for lexical elements. Separators will be returned as their own 702 * ascii value 703 */ 704 typedef enum {HTTPHL_WORD=256, HTTPHL_STRING=257, HTTPHL_END=258, 705 HTTPHL_ERROR = 259} http_header_lex_t; 706 707 /* 708 * Determine what kind of token comes next and return possible value 709 * in buf, which is supposed to have been allocated big enough by 710 * caller. Advance input pointer and return element type. 711 */ 712 static int 713 http_header_lex(const char **cpp, char *buf) 714 { 715 size_t l; 716 /* Eat initial whitespace */ 717 *cpp += strspn(*cpp, " \t"); 718 if (**cpp == 0) 719 return (HTTPHL_END); 720 721 /* Separator ? */ 722 if (**cpp == ',' || **cpp == '=') 723 return (*((*cpp)++)); 724 725 /* String ? */ 726 if (**cpp == '"') { 727 *cpp = http_parse_headerstring(++*cpp, buf); 728 if (*cpp == NULL) 729 return (HTTPHL_ERROR); 730 return (HTTPHL_STRING); 731 } 732 733 /* Read other token, until separator or whitespace */ 734 l = strcspn(*cpp, " \t,="); 735 memcpy(buf, *cpp, l); 736 buf[l] = 0; 737 *cpp += l; 738 return (HTTPHL_WORD); 739 } 740 741 /* 742 * Read challenges from http xxx-authenticate header and accumulate them 743 * in the challenges list structure. 744 * 745 * Headers with multiple challenges are specified by rfc2617, but 746 * servers (ie: squid) often send them in separate headers instead, 747 * which in turn is forbidden by the http spec (multiple headers with 748 * the same name are only allowed for pure comma-separated lists, see 749 * rfc2616 sec 4.2). 750 * 751 * We support both approaches anyway 752 */ 753 static int 754 http_parse_authenticate(const char *cp, http_auth_challenges_t *cs) 755 { 756 int ret = -1; 757 http_header_lex_t lex; 758 char *key = malloc(strlen(cp) + 1); 759 char *value = malloc(strlen(cp) + 1); 760 char *buf = malloc(strlen(cp) + 1); 761 762 if (key == NULL || value == NULL || buf == NULL) { 763 fetch_syserr(); 764 goto out; 765 } 766 767 /* In any case we've seen the header and we set the valid bit */ 768 cs->valid = 1; 769 770 /* Need word first */ 771 lex = http_header_lex(&cp, key); 772 if (lex != HTTPHL_WORD) 773 goto out; 774 775 /* Loop on challenges */ 776 for (; cs->count < MAX_CHALLENGES; cs->count++) { 777 cs->challenges[cs->count] = 778 malloc(sizeof(http_auth_challenge_t)); 779 if (cs->challenges[cs->count] == NULL) { 780 fetch_syserr(); 781 goto out; 782 } 783 init_http_auth_challenge(cs->challenges[cs->count]); 784 if (strcasecmp(key, "basic") == 0) { 785 cs->challenges[cs->count]->scheme = HTTPAS_BASIC; 786 } else if (strcasecmp(key, "digest") == 0) { 787 cs->challenges[cs->count]->scheme = HTTPAS_DIGEST; 788 } else { 789 cs->challenges[cs->count]->scheme = HTTPAS_UNKNOWN; 790 /* 791 * Continue parsing as basic or digest may 792 * follow, and the syntax is the same for 793 * all. We'll just ignore this one when 794 * looking at the list 795 */ 796 } 797 798 /* Loop on attributes */ 799 for (;;) { 800 /* Key */ 801 lex = http_header_lex(&cp, key); 802 if (lex != HTTPHL_WORD) 803 goto out; 804 805 /* Equal sign */ 806 lex = http_header_lex(&cp, buf); 807 if (lex != '=') 808 goto out; 809 810 /* Value */ 811 lex = http_header_lex(&cp, value); 812 if (lex != HTTPHL_WORD && lex != HTTPHL_STRING) 813 goto out; 814 815 if (strcasecmp(key, "realm") == 0) { 816 cs->challenges[cs->count]->realm = 817 strdup(value); 818 } else if (strcasecmp(key, "qop") == 0) { 819 cs->challenges[cs->count]->qop = 820 strdup(value); 821 } else if (strcasecmp(key, "nonce") == 0) { 822 cs->challenges[cs->count]->nonce = 823 strdup(value); 824 } else if (strcasecmp(key, "opaque") == 0) { 825 cs->challenges[cs->count]->opaque = 826 strdup(value); 827 } else if (strcasecmp(key, "algorithm") == 0) { 828 cs->challenges[cs->count]->algo = 829 strdup(value); 830 } else if (strcasecmp(key, "stale") == 0) { 831 cs->challenges[cs->count]->stale = 832 strcasecmp(value, "no"); 833 } else { 834 /* ignore unknown attributes */ 835 } 836 837 /* Comma or Next challenge or End */ 838 lex = http_header_lex(&cp, key); 839 /* 840 * If we get a word here, this is the beginning of the 841 * next challenge. Break the attributes loop 842 */ 843 if (lex == HTTPHL_WORD) 844 break; 845 846 if (lex == HTTPHL_END) { 847 /* End while looking for ',' is normal exit */ 848 cs->count++; 849 ret = 0; 850 goto out; 851 } 852 /* Anything else is an error */ 853 if (lex != ',') 854 goto out; 855 856 } /* End attributes loop */ 857 } /* End challenge loop */ 858 859 /* 860 * Challenges max count exceeded. This really can't happen 861 * with normal data, something's fishy -> error 862 */ 863 864 out: 865 if (key) 866 free(key); 867 if (value) 868 free(value); 869 if (buf) 870 free(buf); 871 return (ret); 872 } 873 874 875 /* 876 * Parse a last-modified header 877 */ 878 static int 879 http_parse_mtime(const char *p, time_t *mtime) 880 { 881 char locale[64], *r; 882 struct tm tm; 883 884 strlcpy(locale, setlocale(LC_TIME, NULL), sizeof(locale)); 885 setlocale(LC_TIME, "C"); 886 r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm); 887 /* 888 * Some proxies use UTC in response, but it should still be 889 * parsed. RFC2616 states GMT and UTC are exactly equal for HTTP. 890 */ 891 if (r == NULL) 892 r = strptime(p, "%a, %d %b %Y %H:%M:%S UTC", &tm); 893 /* XXX should add support for date-2 and date-3 */ 894 setlocale(LC_TIME, locale); 895 if (r == NULL) 896 return (-1); 897 DEBUGF("last modified: [%04d-%02d-%02d %02d:%02d:%02d]\n", 898 tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday, 899 tm.tm_hour, tm.tm_min, tm.tm_sec); 900 *mtime = timegm(&tm); 901 return (0); 902 } 903 904 /* 905 * Parse a content-length header 906 */ 907 static int 908 http_parse_length(const char *p, off_t *length) 909 { 910 off_t len; 911 912 for (len = 0; *p && isdigit((unsigned char)*p); ++p) 913 len = len * 10 + (*p - '0'); 914 if (*p) 915 return (-1); 916 DEBUGF("content length: [%lld]\n", (long long)len); 917 *length = len; 918 return (0); 919 } 920 921 /* 922 * Parse a content-range header 923 */ 924 static int 925 http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size) 926 { 927 off_t first, last, len; 928 929 if (strncasecmp(p, "bytes ", 6) != 0) 930 return (-1); 931 p += 6; 932 if (*p == '*') { 933 first = last = -1; 934 ++p; 935 } else { 936 for (first = 0; *p && isdigit((unsigned char)*p); ++p) 937 first = first * 10 + *p - '0'; 938 if (*p != '-') 939 return (-1); 940 for (last = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 941 last = last * 10 + *p - '0'; 942 } 943 if (first > last || *p != '/') 944 return (-1); 945 for (len = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 946 len = len * 10 + *p - '0'; 947 if (*p || len < last - first + 1) 948 return (-1); 949 if (first == -1) { 950 DEBUGF("content range: [*/%lld]\n", (long long)len); 951 *length = 0; 952 } else { 953 DEBUGF("content range: [%lld-%lld/%lld]\n", 954 (long long)first, (long long)last, (long long)len); 955 *length = last - first + 1; 956 } 957 *offset = first; 958 *size = len; 959 return (0); 960 } 961 962 963 /***************************************************************************** 964 * Helper functions for authorization 965 */ 966 967 /* 968 * Base64 encoding 969 */ 970 static char * 971 http_base64(const char *src) 972 { 973 static const char base64[] = 974 "ABCDEFGHIJKLMNOPQRSTUVWXYZ" 975 "abcdefghijklmnopqrstuvwxyz" 976 "0123456789+/"; 977 char *str, *dst; 978 size_t l; 979 int t; 980 981 l = strlen(src); 982 if ((str = malloc(((l + 2) / 3) * 4 + 1)) == NULL) 983 return (NULL); 984 dst = str; 985 986 while (l >= 3) { 987 t = (src[0] << 16) | (src[1] << 8) | src[2]; 988 dst[0] = base64[(t >> 18) & 0x3f]; 989 dst[1] = base64[(t >> 12) & 0x3f]; 990 dst[2] = base64[(t >> 6) & 0x3f]; 991 dst[3] = base64[(t >> 0) & 0x3f]; 992 src += 3; l -= 3; 993 dst += 4; 994 } 995 996 switch (l) { 997 case 2: 998 t = (src[0] << 16) | (src[1] << 8); 999 dst[0] = base64[(t >> 18) & 0x3f]; 1000 dst[1] = base64[(t >> 12) & 0x3f]; 1001 dst[2] = base64[(t >> 6) & 0x3f]; 1002 dst[3] = '='; 1003 dst += 4; 1004 break; 1005 case 1: 1006 t = src[0] << 16; 1007 dst[0] = base64[(t >> 18) & 0x3f]; 1008 dst[1] = base64[(t >> 12) & 0x3f]; 1009 dst[2] = dst[3] = '='; 1010 dst += 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 #if 0 1188 DEBUGF("Calc: HA1[%s] Nonce[%s] qop[%s] method[%s] URI[%s]\n", 1189 HA1, pszNonce, pszQop, pszMethod, pszDigestUri); 1190 #endif 1191 MD5_CTX Md5Ctx; 1192 HASH HA2; 1193 HASH RespHash; 1194 HASHHEX HA2Hex; 1195 1196 // calculate H(A2) 1197 MD5Init(&Md5Ctx); 1198 MD5Update(&Md5Ctx, pszMethod, strlen(pszMethod)); 1199 MD5Update(&Md5Ctx, ":", 1); 1200 MD5Update(&Md5Ctx, pszDigestUri, strlen(pszDigestUri)); 1201 if (strcasecmp(pszQop, "auth-int") == 0) { 1202 MD5Update(&Md5Ctx, ":", 1); 1203 MD5Update(&Md5Ctx, HEntity, HASHHEXLEN); 1204 } 1205 MD5Final(HA2, &Md5Ctx); 1206 CvtHex(HA2, HA2Hex); 1207 1208 // calculate response 1209 MD5Init(&Md5Ctx); 1210 MD5Update(&Md5Ctx, HA1, HASHHEXLEN); 1211 MD5Update(&Md5Ctx, ":", 1); 1212 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce)); 1213 MD5Update(&Md5Ctx, ":", 1); 1214 if (*pszQop) { 1215 MD5Update(&Md5Ctx, pszNonceCount, strlen(pszNonceCount)); 1216 MD5Update(&Md5Ctx, ":", 1); 1217 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce)); 1218 MD5Update(&Md5Ctx, ":", 1); 1219 MD5Update(&Md5Ctx, pszQop, strlen(pszQop)); 1220 MD5Update(&Md5Ctx, ":", 1); 1221 } 1222 MD5Update(&Md5Ctx, HA2Hex, HASHHEXLEN); 1223 MD5Final(RespHash, &Md5Ctx); 1224 CvtHex(RespHash, Response); 1225 } 1226 1227 /* 1228 * Generate/Send a Digest authorization header 1229 * This looks like: [Proxy-]Authorization: credentials 1230 * 1231 * credentials = "Digest" digest-response 1232 * digest-response = 1#( username | realm | nonce | digest-uri 1233 * | response | [ algorithm ] | [cnonce] | 1234 * [opaque] | [message-qop] | 1235 * [nonce-count] | [auth-param] ) 1236 * username = "username" "=" username-value 1237 * username-value = quoted-string 1238 * digest-uri = "uri" "=" digest-uri-value 1239 * digest-uri-value = request-uri ; As specified by HTTP/1.1 1240 * message-qop = "qop" "=" qop-value 1241 * cnonce = "cnonce" "=" cnonce-value 1242 * cnonce-value = nonce-value 1243 * nonce-count = "nc" "=" nc-value 1244 * nc-value = 8LHEX 1245 * response = "response" "=" request-digest 1246 * request-digest = <"> 32LHEX <"> 1247 */ 1248 static int 1249 http_digest_auth(conn_t *conn, const char *hdr, http_auth_challenge_t *c, 1250 http_auth_params_t *parms, struct url *url) 1251 { 1252 int r; 1253 char noncecount[10]; 1254 char cnonce[40]; 1255 char *options = NULL; 1256 1257 if (!c->realm || !c->nonce) { 1258 DEBUGF("realm/nonce not set in challenge\n"); 1259 return(-1); 1260 } 1261 if (!c->algo) 1262 c->algo = strdup(""); 1263 1264 if (asprintf(&options, "%s%s%s%s", 1265 *c->algo? ",algorithm=" : "", c->algo, 1266 c->opaque? ",opaque=" : "", c->opaque?c->opaque:"") < 0) 1267 return (-1); 1268 1269 if (!c->qop) { 1270 c->qop = strdup(""); 1271 *noncecount = 0; 1272 *cnonce = 0; 1273 } else { 1274 c->nc++; 1275 sprintf(noncecount, "%08x", c->nc); 1276 /* We don't try very hard with the cnonce ... */ 1277 sprintf(cnonce, "%x%lx", getpid(), (unsigned long)time(0)); 1278 } 1279 1280 HASHHEX HA1; 1281 DigestCalcHA1(c->algo, parms->user, c->realm, 1282 parms->password, c->nonce, cnonce, HA1); 1283 DEBUGF("HA1: [%s]\n", HA1); 1284 HASHHEX digest, null; 1285 memset(null, 0, sizeof(null)); 1286 DigestCalcResponse(HA1, c->nonce, noncecount, cnonce, c->qop, 1287 "GET", url->doc, null, digest); 1288 1289 if (c->qop[0]) { 1290 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\"," 1291 "nonce=\"%s\",uri=\"%s\",response=\"%s\"," 1292 "qop=\"auth\", cnonce=\"%s\", nc=%s%s", 1293 hdr, parms->user, c->realm, 1294 c->nonce, url->doc, digest, 1295 cnonce, noncecount, options); 1296 } else { 1297 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\"," 1298 "nonce=\"%s\",uri=\"%s\",response=\"%s\"%s", 1299 hdr, parms->user, c->realm, 1300 c->nonce, url->doc, digest, options); 1301 } 1302 if (options) 1303 free(options); 1304 return (r); 1305 } 1306 1307 /* 1308 * Encode username and password 1309 */ 1310 static int 1311 http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd) 1312 { 1313 char *upw, *auth; 1314 int r; 1315 1316 DEBUGF("basic: usr: [%s]\n", usr); 1317 DEBUGF("basic: pwd: [%s]\n", pwd); 1318 if (asprintf(&upw, "%s:%s", usr, pwd) == -1) 1319 return (-1); 1320 auth = http_base64(upw); 1321 free(upw); 1322 if (auth == NULL) 1323 return (-1); 1324 r = http_cmd(conn, "%s: Basic %s", hdr, auth); 1325 free(auth); 1326 return (r); 1327 } 1328 1329 /* 1330 * Chose the challenge to answer and call the appropriate routine to 1331 * produce the header. 1332 */ 1333 static int 1334 http_authorize(conn_t *conn, const char *hdr, http_auth_challenges_t *cs, 1335 http_auth_params_t *parms, struct url *url) 1336 { 1337 http_auth_challenge_t *digest = NULL; 1338 int i; 1339 1340 /* If user or pass are null we're not happy */ 1341 if (!parms->user || !parms->password) { 1342 DEBUGF("NULL usr or pass\n"); 1343 return (-1); 1344 } 1345 1346 /* Look for a Digest */ 1347 for (i = 0; i < cs->count; i++) { 1348 if (cs->challenges[i]->scheme == HTTPAS_DIGEST) 1349 digest = cs->challenges[i]; 1350 } 1351 1352 /* Error if "Digest" was specified and there is no Digest challenge */ 1353 if (!digest && 1354 (parms->scheme && strcasecmp(parms->scheme, "digest") == 0)) { 1355 DEBUGF("Digest auth in env, not supported by peer\n"); 1356 return (-1); 1357 } 1358 /* 1359 * If "basic" was specified in the environment, or there is no Digest 1360 * challenge, do the basic thing. Don't need a challenge for this, 1361 * so no need to check basic!=NULL 1362 */ 1363 if (!digest || 1364 (parms->scheme && strcasecmp(parms->scheme, "basic") == 0)) 1365 return (http_basic_auth(conn,hdr,parms->user,parms->password)); 1366 1367 /* Else, prefer digest. We just checked that it's not NULL */ 1368 return (http_digest_auth(conn, hdr, digest, parms, url)); 1369 } 1370 1371 /***************************************************************************** 1372 * Helper functions for connecting to a server or proxy 1373 */ 1374 1375 /* 1376 * Connect to the correct HTTP server or proxy. 1377 */ 1378 static conn_t * 1379 http_connect(struct url *URL, struct url *purl, const char *flags) 1380 { 1381 struct url *curl; 1382 conn_t *conn; 1383 hdr_t h; 1384 http_headerbuf_t headerbuf; 1385 const char *p; 1386 int verbose; 1387 int af, val; 1388 int serrno; 1389 bool isproxyauth = false; 1390 http_auth_challenges_t proxy_challenges; 1391 1392 #ifdef INET6 1393 af = AF_UNSPEC; 1394 #else 1395 af = AF_INET; 1396 #endif 1397 1398 verbose = CHECK_FLAG('v'); 1399 if (CHECK_FLAG('4')) 1400 af = AF_INET; 1401 #ifdef INET6 1402 else if (CHECK_FLAG('6')) 1403 af = AF_INET6; 1404 #endif 1405 1406 curl = (purl != NULL) ? purl : URL; 1407 1408 retry: 1409 if ((conn = fetch_connect(curl->host, curl->port, af, verbose)) == NULL) 1410 /* fetch_connect() has already set an error code */ 1411 return (NULL); 1412 init_http_headerbuf(&headerbuf); 1413 if (strcmp(URL->scheme, SCHEME_HTTPS) == 0 && purl) { 1414 init_http_auth_challenges(&proxy_challenges); 1415 http_cmd(conn, "CONNECT %s:%d HTTP/1.1", URL->host, URL->port); 1416 http_cmd(conn, "Host: %s:%d", URL->host, URL->port); 1417 if (isproxyauth) { 1418 http_auth_params_t aparams; 1419 init_http_auth_params(&aparams); 1420 if (*purl->user || *purl->pwd) { 1421 aparams.user = strdup(purl->user); 1422 aparams.password = strdup(purl->pwd); 1423 } else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL && 1424 *p != '\0') { 1425 if (http_authfromenv(p, &aparams) < 0) { 1426 http_seterr(HTTP_NEED_PROXY_AUTH); 1427 fetch_syserr(); 1428 goto ouch; 1429 } 1430 } else if (fetch_netrc_auth(purl) == 0) { 1431 aparams.user = strdup(purl->user); 1432 aparams.password = strdup(purl->pwd); 1433 } else { 1434 /* 1435 * No auth information found in system - exiting 1436 * with warning. 1437 */ 1438 warnx("Missing username and/or password set"); 1439 fetch_syserr(); 1440 goto ouch; 1441 } 1442 http_authorize(conn, "Proxy-Authorization", 1443 &proxy_challenges, &aparams, purl); 1444 clean_http_auth_params(&aparams); 1445 } 1446 http_cmd(conn, ""); 1447 /* Get reply from CONNECT Tunnel attempt */ 1448 int httpreply = http_get_reply(conn); 1449 if (httpreply != HTTP_OK) { 1450 http_seterr(httpreply); 1451 /* If the error is a 407/HTTP_NEED_PROXY_AUTH */ 1452 if (httpreply == HTTP_NEED_PROXY_AUTH && 1453 ! isproxyauth) { 1454 /* Try again with authentication. */ 1455 clean_http_headerbuf(&headerbuf); 1456 fetch_close(conn); 1457 isproxyauth = true; 1458 goto retry; 1459 } 1460 goto ouch; 1461 } 1462 /* Read and discard the rest of the proxy response */ 1463 if (fetch_getln(conn) < 0) { 1464 fetch_syserr(); 1465 goto ouch; 1466 } 1467 do { 1468 switch ((h = http_next_header(conn, &headerbuf, &p))) { 1469 case hdr_syserror: 1470 fetch_syserr(); 1471 goto ouch; 1472 case hdr_error: 1473 http_seterr(HTTP_PROTOCOL_ERROR); 1474 goto ouch; 1475 default: 1476 /* ignore */ ; 1477 } 1478 } while (h > hdr_end); 1479 } 1480 if (strcmp(URL->scheme, SCHEME_HTTPS) == 0 && 1481 fetch_ssl(conn, URL, verbose) == -1) { 1482 /* grrr */ 1483 errno = EAUTH; 1484 fetch_syserr(); 1485 goto ouch; 1486 } 1487 1488 val = 1; 1489 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val)); 1490 1491 clean_http_headerbuf(&headerbuf); 1492 return (conn); 1493 ouch: 1494 serrno = errno; 1495 clean_http_headerbuf(&headerbuf); 1496 fetch_close(conn); 1497 errno = serrno; 1498 return (NULL); 1499 } 1500 1501 static struct url * 1502 http_get_proxy(struct url * url, const char *flags) 1503 { 1504 struct url *purl; 1505 char *p; 1506 1507 if (flags != NULL && strchr(flags, 'd') != NULL) 1508 return (NULL); 1509 if (fetch_no_proxy_match(url->host)) 1510 return (NULL); 1511 if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) && 1512 *p && (purl = fetchParseURL(p))) { 1513 if (!*purl->scheme) 1514 strcpy(purl->scheme, SCHEME_HTTP); 1515 if (!purl->port) 1516 purl->port = fetch_default_proxy_port(purl->scheme); 1517 if (strcmp(purl->scheme, SCHEME_HTTP) == 0) 1518 return (purl); 1519 fetchFreeURL(purl); 1520 } 1521 return (NULL); 1522 } 1523 1524 static void 1525 http_print_html(FILE *out, FILE *in) 1526 { 1527 ssize_t len = 0; 1528 size_t cap; 1529 char *line = NULL, *p, *q; 1530 int comment, tag; 1531 1532 comment = tag = 0; 1533 while ((len = getline(&line, &cap, in)) >= 0) { 1534 while (len && isspace((unsigned char)line[len - 1])) 1535 --len; 1536 for (p = q = line; q < line + len; ++q) { 1537 if (comment && *q == '-') { 1538 if (q + 2 < line + len && 1539 strcmp(q, "-->") == 0) { 1540 tag = comment = 0; 1541 q += 2; 1542 } 1543 } else if (tag && !comment && *q == '>') { 1544 p = q + 1; 1545 tag = 0; 1546 } else if (!tag && *q == '<') { 1547 if (q > p) 1548 fwrite(p, q - p, 1, out); 1549 tag = 1; 1550 if (q + 3 < line + len && 1551 strcmp(q, "<!--") == 0) { 1552 comment = 1; 1553 q += 3; 1554 } 1555 } 1556 } 1557 if (!tag && q > p) 1558 fwrite(p, q - p, 1, out); 1559 fputc('\n', out); 1560 } 1561 1562 free(line); 1563 } 1564 1565 1566 /***************************************************************************** 1567 * Core 1568 */ 1569 1570 FILE * 1571 http_request(struct url *URL, const char *op, struct url_stat *us, 1572 struct url *purl, const char *flags) 1573 { 1574 1575 return (http_request_body(URL, op, us, purl, flags, NULL, NULL)); 1576 } 1577 1578 /* 1579 * Send a request and process the reply 1580 * 1581 * XXX This function is way too long, the do..while loop should be split 1582 * XXX off into a separate function. 1583 */ 1584 FILE * 1585 http_request_body(struct url *URL, const char *op, struct url_stat *us, 1586 struct url *purl, const char *flags, const char *content_type, 1587 const char *body) 1588 { 1589 char timebuf[80]; 1590 char hbuf[MAXHOSTNAMELEN + 7], *host; 1591 conn_t *conn; 1592 struct url *url, *new; 1593 int chunked, direct, ims, noredirect, verbose; 1594 int e, i, n, val; 1595 off_t offset, clength, length, size; 1596 time_t mtime; 1597 const char *p; 1598 FILE *f; 1599 hdr_t h; 1600 struct tm *timestruct; 1601 http_headerbuf_t headerbuf; 1602 http_auth_challenges_t server_challenges; 1603 http_auth_challenges_t proxy_challenges; 1604 size_t body_len; 1605 1606 /* The following calls don't allocate anything */ 1607 init_http_headerbuf(&headerbuf); 1608 init_http_auth_challenges(&server_challenges); 1609 init_http_auth_challenges(&proxy_challenges); 1610 1611 direct = CHECK_FLAG('d'); 1612 noredirect = CHECK_FLAG('A'); 1613 verbose = CHECK_FLAG('v'); 1614 ims = CHECK_FLAG('i'); 1615 1616 if (direct && purl) { 1617 fetchFreeURL(purl); 1618 purl = NULL; 1619 } 1620 1621 /* try the provided URL first */ 1622 url = URL; 1623 1624 n = MAX_REDIRECT; 1625 i = 0; 1626 1627 e = HTTP_PROTOCOL_ERROR; 1628 do { 1629 new = NULL; 1630 chunked = 0; 1631 offset = 0; 1632 clength = -1; 1633 length = -1; 1634 size = -1; 1635 mtime = 0; 1636 1637 /* check port */ 1638 if (!url->port) 1639 url->port = fetch_default_port(url->scheme); 1640 1641 /* were we redirected to an FTP URL? */ 1642 if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) { 1643 if (strcmp(op, "GET") == 0) 1644 return (ftp_request(url, "RETR", us, purl, flags)); 1645 else if (strcmp(op, "HEAD") == 0) 1646 return (ftp_request(url, "STAT", us, purl, flags)); 1647 } 1648 1649 /* connect to server or proxy */ 1650 if ((conn = http_connect(url, purl, flags)) == NULL) 1651 goto ouch; 1652 1653 /* append port number only if necessary */ 1654 host = url->host; 1655 if (url->port != fetch_default_port(url->scheme)) { 1656 snprintf(hbuf, sizeof(hbuf), "%s:%d", host, url->port); 1657 host = hbuf; 1658 } 1659 1660 /* send request */ 1661 if (verbose) 1662 fetch_info("requesting %s://%s%s", 1663 url->scheme, host, url->doc); 1664 if (purl && strcmp(url->scheme, SCHEME_HTTPS) != 0) { 1665 http_cmd(conn, "%s %s://%s%s HTTP/1.1", 1666 op, url->scheme, host, url->doc); 1667 } else { 1668 http_cmd(conn, "%s %s HTTP/1.1", 1669 op, url->doc); 1670 } 1671 1672 if (ims && url->ims_time) { 1673 timestruct = gmtime((time_t *)&url->ims_time); 1674 (void)strftime(timebuf, 80, "%a, %d %b %Y %T GMT", 1675 timestruct); 1676 if (verbose) 1677 fetch_info("If-Modified-Since: %s", timebuf); 1678 http_cmd(conn, "If-Modified-Since: %s", timebuf); 1679 } 1680 /* virtual host */ 1681 http_cmd(conn, "Host: %s", host); 1682 1683 /* 1684 * Proxy authorization: we only send auth after we received 1685 * a 407 error. We do not first try basic anyway (changed 1686 * when support was added for digest-auth) 1687 */ 1688 if (purl && proxy_challenges.valid) { 1689 http_auth_params_t aparams; 1690 init_http_auth_params(&aparams); 1691 if (*purl->user || *purl->pwd) { 1692 aparams.user = strdup(purl->user); 1693 aparams.password = strdup(purl->pwd); 1694 } else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL && 1695 *p != '\0') { 1696 if (http_authfromenv(p, &aparams) < 0) { 1697 http_seterr(HTTP_NEED_PROXY_AUTH); 1698 goto ouch; 1699 } 1700 } else if (fetch_netrc_auth(purl) == 0) { 1701 aparams.user = strdup(purl->user); 1702 aparams.password = strdup(purl->pwd); 1703 } 1704 http_authorize(conn, "Proxy-Authorization", 1705 &proxy_challenges, &aparams, url); 1706 clean_http_auth_params(&aparams); 1707 } 1708 1709 /* 1710 * Server authorization: we never send "a priori" 1711 * Basic auth, which used to be done if user/pass were 1712 * set in the url. This would be weird because we'd send the 1713 * password in the clear even if Digest is finally to be 1714 * used (it would have made more sense for the 1715 * pre-digest version to do this when Basic was specified 1716 * in the environment) 1717 */ 1718 if (server_challenges.valid) { 1719 http_auth_params_t aparams; 1720 init_http_auth_params(&aparams); 1721 if (*url->user || *url->pwd) { 1722 aparams.user = strdup(url->user); 1723 aparams.password = strdup(url->pwd); 1724 } else if ((p = getenv("HTTP_AUTH")) != NULL && 1725 *p != '\0') { 1726 if (http_authfromenv(p, &aparams) < 0) { 1727 http_seterr(HTTP_NEED_AUTH); 1728 goto ouch; 1729 } 1730 } else if (fetch_netrc_auth(url) == 0) { 1731 aparams.user = strdup(url->user); 1732 aparams.password = strdup(url->pwd); 1733 } else if (fetchAuthMethod && 1734 fetchAuthMethod(url) == 0) { 1735 aparams.user = strdup(url->user); 1736 aparams.password = strdup(url->pwd); 1737 } else { 1738 http_seterr(HTTP_NEED_AUTH); 1739 goto ouch; 1740 } 1741 http_authorize(conn, "Authorization", 1742 &server_challenges, &aparams, url); 1743 clean_http_auth_params(&aparams); 1744 } 1745 1746 /* other headers */ 1747 if ((p = getenv("HTTP_ACCEPT")) != NULL) { 1748 if (*p != '\0') 1749 http_cmd(conn, "Accept: %s", p); 1750 } else { 1751 http_cmd(conn, "Accept: */*"); 1752 } 1753 if ((p = getenv("HTTP_REFERER")) != NULL && *p != '\0') { 1754 if (strcasecmp(p, "auto") == 0) 1755 http_cmd(conn, "Referer: %s://%s%s", 1756 url->scheme, host, url->doc); 1757 else 1758 http_cmd(conn, "Referer: %s", p); 1759 } 1760 if ((p = getenv("HTTP_USER_AGENT")) != NULL) { 1761 /* no User-Agent if defined but empty */ 1762 if (*p != '\0') 1763 http_cmd(conn, "User-Agent: %s", p); 1764 } else { 1765 /* default User-Agent */ 1766 http_cmd(conn, "User-Agent: %s " _LIBFETCH_VER, 1767 getprogname()); 1768 } 1769 if (url->offset > 0) 1770 http_cmd(conn, "Range: bytes=%lld-", (long long)url->offset); 1771 http_cmd(conn, "Connection: close"); 1772 1773 if (body) { 1774 body_len = strlen(body); 1775 http_cmd(conn, "Content-Length: %zu", body_len); 1776 if (content_type != NULL) 1777 http_cmd(conn, "Content-Type: %s", content_type); 1778 } 1779 1780 http_cmd(conn, ""); 1781 1782 if (body) 1783 fetch_write(conn, body, body_len); 1784 1785 /* 1786 * Force the queued request to be dispatched. Normally, one 1787 * would do this with shutdown(2) but squid proxies can be 1788 * configured to disallow such half-closed connections. To 1789 * be compatible with such configurations, fiddle with socket 1790 * options to force the pending data to be written. 1791 */ 1792 val = 0; 1793 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, 1794 sizeof(val)); 1795 val = 1; 1796 setsockopt(conn->sd, IPPROTO_TCP, TCP_NODELAY, &val, 1797 sizeof(val)); 1798 1799 /* get reply */ 1800 switch (http_get_reply(conn)) { 1801 case HTTP_OK: 1802 case HTTP_PARTIAL: 1803 case HTTP_NOT_MODIFIED: 1804 /* fine */ 1805 break; 1806 case HTTP_MOVED_PERM: 1807 case HTTP_MOVED_TEMP: 1808 case HTTP_TEMP_REDIRECT: 1809 case HTTP_PERM_REDIRECT: 1810 case HTTP_SEE_OTHER: 1811 case HTTP_USE_PROXY: 1812 /* 1813 * Not so fine, but we still have to read the 1814 * headers to get the new location. 1815 */ 1816 break; 1817 case HTTP_NEED_AUTH: 1818 if (server_challenges.valid) { 1819 /* 1820 * We already sent out authorization code, 1821 * so there's nothing more we can do. 1822 */ 1823 http_seterr(conn->err); 1824 goto ouch; 1825 } 1826 /* try again, but send the password this time */ 1827 if (verbose) 1828 fetch_info("server requires authorization"); 1829 break; 1830 case HTTP_NEED_PROXY_AUTH: 1831 if (proxy_challenges.valid) { 1832 /* 1833 * We already sent our proxy 1834 * authorization code, so there's 1835 * nothing more we can do. */ 1836 http_seterr(conn->err); 1837 goto ouch; 1838 } 1839 /* try again, but send the password this time */ 1840 if (verbose) 1841 fetch_info("proxy requires authorization"); 1842 break; 1843 case HTTP_BAD_RANGE: 1844 /* 1845 * This can happen if we ask for 0 bytes because 1846 * we already have the whole file. Consider this 1847 * a success for now, and check sizes later. 1848 */ 1849 break; 1850 case HTTP_PROTOCOL_ERROR: 1851 /* fall through */ 1852 case -1: 1853 fetch_syserr(); 1854 goto ouch; 1855 default: 1856 http_seterr(conn->err); 1857 if (!verbose) 1858 goto ouch; 1859 /* fall through so we can get the full error message */ 1860 } 1861 1862 /* get headers. http_next_header expects one line readahead */ 1863 if (fetch_getln(conn) == -1) { 1864 fetch_syserr(); 1865 goto ouch; 1866 } 1867 do { 1868 switch ((h = http_next_header(conn, &headerbuf, &p))) { 1869 case hdr_syserror: 1870 fetch_syserr(); 1871 goto ouch; 1872 case hdr_error: 1873 http_seterr(HTTP_PROTOCOL_ERROR); 1874 goto ouch; 1875 case hdr_content_length: 1876 http_parse_length(p, &clength); 1877 break; 1878 case hdr_content_range: 1879 http_parse_range(p, &offset, &length, &size); 1880 break; 1881 case hdr_last_modified: 1882 http_parse_mtime(p, &mtime); 1883 break; 1884 case hdr_location: 1885 if (!HTTP_REDIRECT(conn->err)) 1886 break; 1887 /* 1888 * if the A flag is set, we don't follow 1889 * temporary redirects. 1890 */ 1891 if (noredirect && 1892 conn->err != HTTP_MOVED_PERM && 1893 conn->err != HTTP_PERM_REDIRECT && 1894 conn->err != HTTP_USE_PROXY) { 1895 n = 1; 1896 break; 1897 } 1898 if (new) 1899 free(new); 1900 if (verbose) 1901 fetch_info("%d redirect to %s", 1902 conn->err, p); 1903 if (*p == '/') 1904 /* absolute path */ 1905 new = fetchMakeURL(url->scheme, url->host, 1906 url->port, p, url->user, url->pwd); 1907 else 1908 new = fetchParseURL(p); 1909 if (new == NULL) { 1910 /* XXX should set an error code */ 1911 DEBUGF("failed to parse new URL\n"); 1912 goto ouch; 1913 } 1914 1915 /* Only copy credentials if the host matches */ 1916 if (strcmp(new->host, url->host) == 0 && 1917 !*new->user && !*new->pwd) { 1918 strcpy(new->user, url->user); 1919 strcpy(new->pwd, url->pwd); 1920 } 1921 new->offset = url->offset; 1922 new->length = url->length; 1923 new->ims_time = url->ims_time; 1924 break; 1925 case hdr_transfer_encoding: 1926 /* XXX weak test*/ 1927 chunked = (strcasecmp(p, "chunked") == 0); 1928 break; 1929 case hdr_www_authenticate: 1930 if (conn->err != HTTP_NEED_AUTH) 1931 break; 1932 if (http_parse_authenticate(p, &server_challenges) == 0) 1933 ++n; 1934 break; 1935 case hdr_proxy_authenticate: 1936 if (conn->err != HTTP_NEED_PROXY_AUTH) 1937 break; 1938 if (http_parse_authenticate(p, &proxy_challenges) == 0) 1939 ++n; 1940 break; 1941 case hdr_end: 1942 /* fall through */ 1943 case hdr_unknown: 1944 /* ignore */ 1945 break; 1946 } 1947 } while (h > hdr_end); 1948 1949 /* we need to provide authentication */ 1950 if (conn->err == HTTP_NEED_AUTH || 1951 conn->err == HTTP_NEED_PROXY_AUTH) { 1952 e = conn->err; 1953 if ((conn->err == HTTP_NEED_AUTH && 1954 !server_challenges.valid) || 1955 (conn->err == HTTP_NEED_PROXY_AUTH && 1956 !proxy_challenges.valid)) { 1957 /* 401/7 but no www/proxy-authenticate ?? */ 1958 DEBUGF("%03d without auth header\n", conn->err); 1959 goto ouch; 1960 } 1961 fetch_close(conn); 1962 conn = NULL; 1963 continue; 1964 } 1965 1966 /* requested range not satisfiable */ 1967 if (conn->err == HTTP_BAD_RANGE) { 1968 if (url->offset > 0 && url->length == 0) { 1969 /* asked for 0 bytes; fake it */ 1970 offset = url->offset; 1971 clength = -1; 1972 conn->err = HTTP_OK; 1973 break; 1974 } else { 1975 http_seterr(conn->err); 1976 goto ouch; 1977 } 1978 } 1979 1980 /* we have a hit or an error */ 1981 if (conn->err == HTTP_OK 1982 || conn->err == HTTP_NOT_MODIFIED 1983 || conn->err == HTTP_PARTIAL 1984 || HTTP_ERROR(conn->err)) 1985 break; 1986 1987 /* all other cases: we got a redirect */ 1988 e = conn->err; 1989 clean_http_auth_challenges(&server_challenges); 1990 fetch_close(conn); 1991 conn = NULL; 1992 if (!new) { 1993 DEBUGF("redirect with no new location\n"); 1994 break; 1995 } 1996 if (url != URL) 1997 fetchFreeURL(url); 1998 url = new; 1999 } while (++i < n); 2000 2001 /* we failed, or ran out of retries */ 2002 if (conn == NULL) { 2003 http_seterr(e); 2004 goto ouch; 2005 } 2006 2007 DEBUGF("offset %lld, length %lld, size %lld, clength %lld\n", 2008 (long long)offset, (long long)length, 2009 (long long)size, (long long)clength); 2010 2011 if (conn->err == HTTP_NOT_MODIFIED) { 2012 http_seterr(HTTP_NOT_MODIFIED); 2013 return (NULL); 2014 } 2015 2016 /* check for inconsistencies */ 2017 if (clength != -1 && length != -1 && clength != length) { 2018 http_seterr(HTTP_PROTOCOL_ERROR); 2019 goto ouch; 2020 } 2021 if (clength == -1) 2022 clength = length; 2023 if (clength != -1) 2024 length = offset + clength; 2025 if (length != -1 && size != -1 && length != size) { 2026 http_seterr(HTTP_PROTOCOL_ERROR); 2027 goto ouch; 2028 } 2029 if (size == -1) 2030 size = length; 2031 2032 /* fill in stats */ 2033 if (us) { 2034 us->size = size; 2035 us->atime = us->mtime = mtime; 2036 } 2037 2038 /* too far? */ 2039 if (URL->offset > 0 && offset > URL->offset) { 2040 http_seterr(HTTP_PROTOCOL_ERROR); 2041 goto ouch; 2042 } 2043 2044 /* report back real offset and size */ 2045 URL->offset = offset; 2046 URL->length = clength; 2047 2048 /* wrap it up in a FILE */ 2049 if ((f = http_funopen(conn, chunked)) == NULL) { 2050 fetch_syserr(); 2051 goto ouch; 2052 } 2053 2054 if (url != URL) 2055 fetchFreeURL(url); 2056 if (purl) 2057 fetchFreeURL(purl); 2058 2059 if (HTTP_ERROR(conn->err)) { 2060 http_print_html(stderr, f); 2061 fclose(f); 2062 f = NULL; 2063 } 2064 clean_http_headerbuf(&headerbuf); 2065 clean_http_auth_challenges(&server_challenges); 2066 clean_http_auth_challenges(&proxy_challenges); 2067 return (f); 2068 2069 ouch: 2070 if (url != URL) 2071 fetchFreeURL(url); 2072 if (purl) 2073 fetchFreeURL(purl); 2074 if (conn != NULL) 2075 fetch_close(conn); 2076 clean_http_headerbuf(&headerbuf); 2077 clean_http_auth_challenges(&server_challenges); 2078 clean_http_auth_challenges(&proxy_challenges); 2079 return (NULL); 2080 } 2081 2082 2083 /***************************************************************************** 2084 * Entry points 2085 */ 2086 2087 /* 2088 * Retrieve and stat a file by HTTP 2089 */ 2090 FILE * 2091 fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags) 2092 { 2093 return (http_request(URL, "GET", us, http_get_proxy(URL, flags), flags)); 2094 } 2095 2096 /* 2097 * Retrieve a file by HTTP 2098 */ 2099 FILE * 2100 fetchGetHTTP(struct url *URL, const char *flags) 2101 { 2102 return (fetchXGetHTTP(URL, NULL, flags)); 2103 } 2104 2105 /* 2106 * Store a file by HTTP 2107 */ 2108 FILE * 2109 fetchPutHTTP(struct url *URL __unused, const char *flags __unused) 2110 { 2111 warnx("fetchPutHTTP(): not implemented"); 2112 return (NULL); 2113 } 2114 2115 /* 2116 * Get an HTTP document's metadata 2117 */ 2118 int 2119 fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags) 2120 { 2121 FILE *f; 2122 2123 f = http_request(URL, "HEAD", us, http_get_proxy(URL, flags), flags); 2124 if (f == NULL) 2125 return (-1); 2126 fclose(f); 2127 return (0); 2128 } 2129 2130 /* 2131 * List a directory 2132 */ 2133 struct url_ent * 2134 fetchListHTTP(struct url *url __unused, const char *flags __unused) 2135 { 2136 warnx("fetchListHTTP(): not implemented"); 2137 return (NULL); 2138 } 2139 2140 /* 2141 * Arbitrary HTTP verb and content requests 2142 */ 2143 FILE * 2144 fetchReqHTTP(struct url *URL, const char *method, const char *flags, 2145 const char *content_type, const char *body) 2146 { 2147 2148 return (http_request_body(URL, method, NULL, http_get_proxy(URL, flags), 2149 flags, content_type, body)); 2150 } 2151