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_PERM_REDIRECT \ 118 || (xyz) == HTTP_USE_PROXY \ 119 || (xyz) == HTTP_SEE_OTHER) 120 121 #define HTTP_ERROR(xyz) ((xyz) > 400 && (xyz) < 599) 122 123 124 /***************************************************************************** 125 * I/O functions for decoding chunked streams 126 */ 127 128 struct httpio 129 { 130 conn_t *conn; /* connection */ 131 int chunked; /* chunked mode */ 132 char *buf; /* chunk buffer */ 133 size_t bufsize; /* size of chunk buffer */ 134 size_t buflen; /* amount of data currently in buffer */ 135 size_t bufpos; /* current read offset in buffer */ 136 int eof; /* end-of-file flag */ 137 int error; /* error flag */ 138 size_t chunksize; /* remaining size of current chunk */ 139 #ifndef NDEBUG 140 size_t total; 141 #endif 142 }; 143 144 /* 145 * Get next chunk header 146 */ 147 static int 148 http_new_chunk(struct httpio *io) 149 { 150 char *p; 151 152 if (fetch_getln(io->conn) == -1) 153 return (-1); 154 155 if (io->conn->buflen < 2 || !isxdigit((unsigned char)*io->conn->buf)) 156 return (-1); 157 158 for (p = io->conn->buf; *p && !isspace((unsigned char)*p); ++p) { 159 if (*p == ';') 160 break; 161 if (!isxdigit((unsigned char)*p)) 162 return (-1); 163 if (isdigit((unsigned char)*p)) { 164 io->chunksize = io->chunksize * 16 + 165 *p - '0'; 166 } else { 167 io->chunksize = io->chunksize * 16 + 168 10 + tolower((unsigned char)*p) - 'a'; 169 } 170 } 171 172 #ifndef NDEBUG 173 if (fetchDebug) { 174 io->total += io->chunksize; 175 if (io->chunksize == 0) 176 fprintf(stderr, "%s(): end of last chunk\n", __func__); 177 else 178 fprintf(stderr, "%s(): new chunk: %lu (%lu)\n", 179 __func__, (unsigned long)io->chunksize, 180 (unsigned long)io->total); 181 } 182 #endif 183 184 return (io->chunksize); 185 } 186 187 /* 188 * Grow the input buffer to at least len bytes 189 */ 190 static inline int 191 http_growbuf(struct httpio *io, size_t len) 192 { 193 char *tmp; 194 195 if (io->bufsize >= len) 196 return (0); 197 198 if ((tmp = realloc(io->buf, len)) == NULL) 199 return (-1); 200 io->buf = tmp; 201 io->bufsize = len; 202 return (0); 203 } 204 205 /* 206 * Fill the input buffer, do chunk decoding on the fly 207 */ 208 static ssize_t 209 http_fillbuf(struct httpio *io, size_t len) 210 { 211 ssize_t nbytes; 212 char ch; 213 214 if (io->error) 215 return (-1); 216 if (io->eof) 217 return (0); 218 219 /* not chunked: just fetch the requested amount */ 220 if (io->chunked == 0) { 221 if (http_growbuf(io, len) == -1) 222 return (-1); 223 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) { 224 io->error = errno; 225 return (-1); 226 } 227 io->buflen = nbytes; 228 io->bufpos = 0; 229 return (io->buflen); 230 } 231 232 /* chunked, but we ran out: get the next chunk header */ 233 if (io->chunksize == 0) { 234 switch (http_new_chunk(io)) { 235 case -1: 236 io->error = EPROTO; 237 return (-1); 238 case 0: 239 io->eof = 1; 240 return (0); 241 } 242 } 243 244 /* fetch the requested amount, but no more than the current chunk */ 245 if (len > io->chunksize) 246 len = io->chunksize; 247 if (http_growbuf(io, len) == -1) 248 return (-1); 249 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) { 250 io->error = errno; 251 return (-1); 252 } 253 io->bufpos = 0; 254 io->buflen = nbytes; 255 io->chunksize -= nbytes; 256 257 if (io->chunksize == 0) { 258 if (fetch_read(io->conn, &ch, 1) != 1 || ch != '\r' || 259 fetch_read(io->conn, &ch, 1) != 1 || ch != '\n') 260 return (-1); 261 } 262 263 return (io->buflen); 264 } 265 266 /* 267 * Read function 268 */ 269 static int 270 http_readfn(void *v, char *buf, int len) 271 { 272 struct httpio *io = (struct httpio *)v; 273 int rlen; 274 275 if (io->error) 276 return (-1); 277 if (io->eof) 278 return (0); 279 280 /* empty buffer */ 281 if (!io->buf || io->bufpos == io->buflen) { 282 if ((rlen = http_fillbuf(io, len)) < 0) { 283 if ((errno = io->error) == EINTR) 284 io->error = 0; 285 return (-1); 286 } else if (rlen == 0) { 287 return (0); 288 } 289 } 290 291 rlen = io->buflen - io->bufpos; 292 if (len < rlen) 293 rlen = len; 294 memcpy(buf, io->buf + io->bufpos, rlen); 295 io->bufpos += rlen; 296 return (rlen); 297 } 298 299 /* 300 * Write function 301 */ 302 static int 303 http_writefn(void *v, const char *buf, int len) 304 { 305 struct httpio *io = (struct httpio *)v; 306 307 return (fetch_write(io->conn, buf, len)); 308 } 309 310 /* 311 * Close function 312 */ 313 static int 314 http_closefn(void *v) 315 { 316 struct httpio *io = (struct httpio *)v; 317 int r; 318 319 r = fetch_close(io->conn); 320 if (io->buf) 321 free(io->buf); 322 free(io); 323 return (r); 324 } 325 326 /* 327 * Wrap a file descriptor up 328 */ 329 static FILE * 330 http_funopen(conn_t *conn, int chunked) 331 { 332 struct httpio *io; 333 FILE *f; 334 335 if ((io = calloc(1, sizeof(*io))) == NULL) { 336 fetch_syserr(); 337 return (NULL); 338 } 339 io->conn = conn; 340 io->chunked = chunked; 341 f = funopen(io, http_readfn, http_writefn, NULL, http_closefn); 342 if (f == NULL) { 343 fetch_syserr(); 344 free(io); 345 return (NULL); 346 } 347 return (f); 348 } 349 350 351 /***************************************************************************** 352 * Helper functions for talking to the server and parsing its replies 353 */ 354 355 /* Header types */ 356 typedef enum { 357 hdr_syserror = -2, 358 hdr_error = -1, 359 hdr_end = 0, 360 hdr_unknown = 1, 361 hdr_content_length, 362 hdr_content_range, 363 hdr_last_modified, 364 hdr_location, 365 hdr_transfer_encoding, 366 hdr_www_authenticate, 367 hdr_proxy_authenticate, 368 } hdr_t; 369 370 /* Names of interesting headers */ 371 static struct { 372 hdr_t num; 373 const char *name; 374 } hdr_names[] = { 375 { hdr_content_length, "Content-Length" }, 376 { hdr_content_range, "Content-Range" }, 377 { hdr_last_modified, "Last-Modified" }, 378 { hdr_location, "Location" }, 379 { hdr_transfer_encoding, "Transfer-Encoding" }, 380 { hdr_www_authenticate, "WWW-Authenticate" }, 381 { hdr_proxy_authenticate, "Proxy-Authenticate" }, 382 { hdr_unknown, NULL }, 383 }; 384 385 /* 386 * Send a formatted line; optionally echo to terminal 387 */ 388 static int 389 http_cmd(conn_t *conn, const char *fmt, ...) 390 { 391 va_list ap; 392 size_t len; 393 char *msg; 394 int r; 395 396 va_start(ap, fmt); 397 len = vasprintf(&msg, fmt, ap); 398 va_end(ap); 399 400 if (msg == NULL) { 401 errno = ENOMEM; 402 fetch_syserr(); 403 return (-1); 404 } 405 406 r = fetch_putln(conn, msg, len); 407 free(msg); 408 409 if (r == -1) { 410 fetch_syserr(); 411 return (-1); 412 } 413 414 return (0); 415 } 416 417 /* 418 * Get and parse status line 419 */ 420 static int 421 http_get_reply(conn_t *conn) 422 { 423 char *p; 424 425 if (fetch_getln(conn) == -1) 426 return (-1); 427 /* 428 * A valid status line looks like "HTTP/m.n xyz reason" where m 429 * and n are the major and minor protocol version numbers and xyz 430 * is the reply code. 431 * Unfortunately, there are servers out there (NCSA 1.5.1, to name 432 * just one) that do not send a version number, so we can't rely 433 * on finding one, but if we do, insist on it being 1.0 or 1.1. 434 * We don't care about the reason phrase. 435 */ 436 if (strncmp(conn->buf, "HTTP", 4) != 0) 437 return (HTTP_PROTOCOL_ERROR); 438 p = conn->buf + 4; 439 if (*p == '/') { 440 if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1')) 441 return (HTTP_PROTOCOL_ERROR); 442 p += 4; 443 } 444 if (*p != ' ' || 445 !isdigit((unsigned char)p[1]) || 446 !isdigit((unsigned char)p[2]) || 447 !isdigit((unsigned char)p[3])) 448 return (HTTP_PROTOCOL_ERROR); 449 450 conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0'); 451 return (conn->err); 452 } 453 454 /* 455 * Check a header; if the type matches the given string, return a pointer 456 * to the beginning of the value. 457 */ 458 static const char * 459 http_match(const char *str, const char *hdr) 460 { 461 while (*str && *hdr && 462 tolower((unsigned char)*str++) == tolower((unsigned char)*hdr++)) 463 /* nothing */; 464 if (*str || *hdr != ':') 465 return (NULL); 466 while (*hdr && isspace((unsigned char)*++hdr)) 467 /* nothing */; 468 return (hdr); 469 } 470 471 472 /* 473 * Get the next header and return the appropriate symbolic code. We 474 * need to read one line ahead for checking for a continuation line 475 * belonging to the current header (continuation lines start with 476 * white space). 477 * 478 * We get called with a fresh line already in the conn buffer, either 479 * from the previous http_next_header() invocation, or, the first 480 * time, from a fetch_getln() performed by our caller. 481 * 482 * This stops when we encounter an empty line (we dont read beyond the header 483 * area). 484 * 485 * Note that the "headerbuf" is just a place to return the result. Its 486 * contents are not used for the next call. This means that no cleanup 487 * is needed when ie doing another connection, just call the cleanup when 488 * fully done to deallocate memory. 489 */ 490 491 /* Limit the max number of continuation lines to some reasonable value */ 492 #define HTTP_MAX_CONT_LINES 10 493 494 /* Place into which to build a header from one or several lines */ 495 typedef struct { 496 char *buf; /* buffer */ 497 size_t bufsize; /* buffer size */ 498 size_t buflen; /* length of buffer contents */ 499 } http_headerbuf_t; 500 501 static void 502 init_http_headerbuf(http_headerbuf_t *buf) 503 { 504 buf->buf = NULL; 505 buf->bufsize = 0; 506 buf->buflen = 0; 507 } 508 509 static void 510 clean_http_headerbuf(http_headerbuf_t *buf) 511 { 512 if (buf->buf) 513 free(buf->buf); 514 init_http_headerbuf(buf); 515 } 516 517 /* Remove whitespace at the end of the buffer */ 518 static void 519 http_conn_trimright(conn_t *conn) 520 { 521 while (conn->buflen && 522 isspace((unsigned char)conn->buf[conn->buflen - 1])) 523 conn->buflen--; 524 conn->buf[conn->buflen] = '\0'; 525 } 526 527 static hdr_t 528 http_next_header(conn_t *conn, http_headerbuf_t *hbuf, const char **p) 529 { 530 unsigned int i, len; 531 532 /* 533 * Have to do the stripping here because of the first line. So 534 * it's done twice for the subsequent lines. No big deal 535 */ 536 http_conn_trimright(conn); 537 if (conn->buflen == 0) 538 return (hdr_end); 539 540 /* Copy the line to the headerbuf */ 541 if (hbuf->bufsize < conn->buflen + 1) { 542 if ((hbuf->buf = realloc(hbuf->buf, conn->buflen + 1)) == NULL) 543 return (hdr_syserror); 544 hbuf->bufsize = conn->buflen + 1; 545 } 546 strcpy(hbuf->buf, conn->buf); 547 hbuf->buflen = conn->buflen; 548 549 /* 550 * Fetch possible continuation lines. Stop at 1st non-continuation 551 * and leave it in the conn buffer 552 */ 553 for (i = 0; i < HTTP_MAX_CONT_LINES; i++) { 554 if (fetch_getln(conn) == -1) 555 return (hdr_syserror); 556 557 /* 558 * Note: we carry on the idea from the previous version 559 * that a pure whitespace line is equivalent to an empty 560 * one (so it's not continuation and will be handled when 561 * we are called next) 562 */ 563 http_conn_trimright(conn); 564 if (conn->buf[0] != ' ' && conn->buf[0] != "\t"[0]) 565 break; 566 567 /* Got a continuation line. Concatenate to previous */ 568 len = hbuf->buflen + conn->buflen; 569 if (hbuf->bufsize < len + 1) { 570 len *= 2; 571 if ((hbuf->buf = realloc(hbuf->buf, len + 1)) == NULL) 572 return (hdr_syserror); 573 hbuf->bufsize = len + 1; 574 } 575 strcpy(hbuf->buf + hbuf->buflen, conn->buf); 576 hbuf->buflen += conn->buflen; 577 } 578 579 /* 580 * We could check for malformed headers but we don't really care. 581 * A valid header starts with a token immediately followed by a 582 * colon; a token is any sequence of non-control, non-whitespace 583 * characters except "()<>@,;:\\\"{}". 584 */ 585 for (i = 0; hdr_names[i].num != hdr_unknown; i++) 586 if ((*p = http_match(hdr_names[i].name, hbuf->buf)) != NULL) 587 return (hdr_names[i].num); 588 589 return (hdr_unknown); 590 } 591 592 /************************** 593 * [Proxy-]Authenticate header parsing 594 */ 595 596 /* 597 * Read doublequote-delimited string into output buffer obuf (allocated 598 * by caller, whose responsibility it is to ensure that it's big enough) 599 * cp points to the first char after the initial '"' 600 * Handles \ quoting 601 * Returns pointer to the first char after the terminating double quote, or 602 * NULL for error. 603 */ 604 static const char * 605 http_parse_headerstring(const char *cp, char *obuf) 606 { 607 for (;;) { 608 switch (*cp) { 609 case 0: /* Unterminated string */ 610 *obuf = 0; 611 return (NULL); 612 case '"': /* Ending quote */ 613 *obuf = 0; 614 return (++cp); 615 case '\\': 616 if (*++cp == 0) { 617 *obuf = 0; 618 return (NULL); 619 } 620 /* FALLTHROUGH */ 621 default: 622 *obuf++ = *cp++; 623 } 624 } 625 } 626 627 /* Http auth challenge schemes */ 628 typedef enum {HTTPAS_UNKNOWN, HTTPAS_BASIC,HTTPAS_DIGEST} http_auth_schemes_t; 629 630 /* Data holder for a Basic or Digest challenge. */ 631 typedef struct { 632 http_auth_schemes_t scheme; 633 char *realm; 634 char *qop; 635 char *nonce; 636 char *opaque; 637 char *algo; 638 int stale; 639 int nc; /* Nonce count */ 640 } http_auth_challenge_t; 641 642 static void 643 init_http_auth_challenge(http_auth_challenge_t *b) 644 { 645 b->scheme = HTTPAS_UNKNOWN; 646 b->realm = b->qop = b->nonce = b->opaque = b->algo = NULL; 647 b->stale = b->nc = 0; 648 } 649 650 static void 651 clean_http_auth_challenge(http_auth_challenge_t *b) 652 { 653 if (b->realm) 654 free(b->realm); 655 if (b->qop) 656 free(b->qop); 657 if (b->nonce) 658 free(b->nonce); 659 if (b->opaque) 660 free(b->opaque); 661 if (b->algo) 662 free(b->algo); 663 init_http_auth_challenge(b); 664 } 665 666 /* Data holder for an array of challenges offered in an http response. */ 667 #define MAX_CHALLENGES 10 668 typedef struct { 669 http_auth_challenge_t *challenges[MAX_CHALLENGES]; 670 int count; /* Number of parsed challenges in the array */ 671 int valid; /* We did parse an authenticate header */ 672 } http_auth_challenges_t; 673 674 static void 675 init_http_auth_challenges(http_auth_challenges_t *cs) 676 { 677 int i; 678 for (i = 0; i < MAX_CHALLENGES; i++) 679 cs->challenges[i] = NULL; 680 cs->count = cs->valid = 0; 681 } 682 683 static void 684 clean_http_auth_challenges(http_auth_challenges_t *cs) 685 { 686 int i; 687 /* We rely on non-zero pointers being allocated, not on the count */ 688 for (i = 0; i < MAX_CHALLENGES; i++) { 689 if (cs->challenges[i] != NULL) { 690 clean_http_auth_challenge(cs->challenges[i]); 691 free(cs->challenges[i]); 692 } 693 } 694 init_http_auth_challenges(cs); 695 } 696 697 /* 698 * Enumeration for lexical elements. Separators will be returned as their own 699 * ascii value 700 */ 701 typedef enum {HTTPHL_WORD=256, HTTPHL_STRING=257, HTTPHL_END=258, 702 HTTPHL_ERROR = 259} http_header_lex_t; 703 704 /* 705 * Determine what kind of token comes next and return possible value 706 * in buf, which is supposed to have been allocated big enough by 707 * caller. Advance input pointer and return element type. 708 */ 709 static int 710 http_header_lex(const char **cpp, char *buf) 711 { 712 size_t l; 713 /* Eat initial whitespace */ 714 *cpp += strspn(*cpp, " \t"); 715 if (**cpp == 0) 716 return (HTTPHL_END); 717 718 /* Separator ? */ 719 if (**cpp == ',' || **cpp == '=') 720 return (*((*cpp)++)); 721 722 /* String ? */ 723 if (**cpp == '"') { 724 *cpp = http_parse_headerstring(++*cpp, buf); 725 if (*cpp == NULL) 726 return (HTTPHL_ERROR); 727 return (HTTPHL_STRING); 728 } 729 730 /* Read other token, until separator or whitespace */ 731 l = strcspn(*cpp, " \t,="); 732 memcpy(buf, *cpp, l); 733 buf[l] = 0; 734 *cpp += l; 735 return (HTTPHL_WORD); 736 } 737 738 /* 739 * Read challenges from http xxx-authenticate header and accumulate them 740 * in the challenges list structure. 741 * 742 * Headers with multiple challenges are specified by rfc2617, but 743 * servers (ie: squid) often send them in separate headers instead, 744 * which in turn is forbidden by the http spec (multiple headers with 745 * the same name are only allowed for pure comma-separated lists, see 746 * rfc2616 sec 4.2). 747 * 748 * We support both approaches anyway 749 */ 750 static int 751 http_parse_authenticate(const char *cp, http_auth_challenges_t *cs) 752 { 753 int ret = -1; 754 http_header_lex_t lex; 755 char *key = malloc(strlen(cp) + 1); 756 char *value = malloc(strlen(cp) + 1); 757 char *buf = malloc(strlen(cp) + 1); 758 759 if (key == NULL || value == NULL || buf == NULL) { 760 fetch_syserr(); 761 goto out; 762 } 763 764 /* In any case we've seen the header and we set the valid bit */ 765 cs->valid = 1; 766 767 /* Need word first */ 768 lex = http_header_lex(&cp, key); 769 if (lex != HTTPHL_WORD) 770 goto out; 771 772 /* Loop on challenges */ 773 for (; cs->count < MAX_CHALLENGES; cs->count++) { 774 cs->challenges[cs->count] = 775 malloc(sizeof(http_auth_challenge_t)); 776 if (cs->challenges[cs->count] == NULL) { 777 fetch_syserr(); 778 goto out; 779 } 780 init_http_auth_challenge(cs->challenges[cs->count]); 781 if (!strcasecmp(key, "basic")) { 782 cs->challenges[cs->count]->scheme = HTTPAS_BASIC; 783 } else if (!strcasecmp(key, "digest")) { 784 cs->challenges[cs->count]->scheme = HTTPAS_DIGEST; 785 } else { 786 cs->challenges[cs->count]->scheme = HTTPAS_UNKNOWN; 787 /* 788 * Continue parsing as basic or digest may 789 * follow, and the syntax is the same for 790 * all. We'll just ignore this one when 791 * looking at the list 792 */ 793 } 794 795 /* Loop on attributes */ 796 for (;;) { 797 /* Key */ 798 lex = http_header_lex(&cp, key); 799 if (lex != HTTPHL_WORD) 800 goto out; 801 802 /* Equal sign */ 803 lex = http_header_lex(&cp, buf); 804 if (lex != '=') 805 goto out; 806 807 /* Value */ 808 lex = http_header_lex(&cp, value); 809 if (lex != HTTPHL_WORD && lex != HTTPHL_STRING) 810 goto out; 811 812 if (!strcasecmp(key, "realm")) 813 cs->challenges[cs->count]->realm = 814 strdup(value); 815 else if (!strcasecmp(key, "qop")) 816 cs->challenges[cs->count]->qop = 817 strdup(value); 818 else if (!strcasecmp(key, "nonce")) 819 cs->challenges[cs->count]->nonce = 820 strdup(value); 821 else if (!strcasecmp(key, "opaque")) 822 cs->challenges[cs->count]->opaque = 823 strdup(value); 824 else if (!strcasecmp(key, "algorithm")) 825 cs->challenges[cs->count]->algo = 826 strdup(value); 827 else if (!strcasecmp(key, "stale")) 828 cs->challenges[cs->count]->stale = 829 strcasecmp(value, "no"); 830 /* Else ignore unknown attributes */ 831 832 /* Comma or Next challenge or End */ 833 lex = http_header_lex(&cp, key); 834 /* 835 * If we get a word here, this is the beginning of the 836 * next challenge. Break the attributes loop 837 */ 838 if (lex == HTTPHL_WORD) 839 break; 840 841 if (lex == HTTPHL_END) { 842 /* End while looking for ',' is normal exit */ 843 cs->count++; 844 ret = 0; 845 goto out; 846 } 847 /* Anything else is an error */ 848 if (lex != ',') 849 goto out; 850 851 } /* End attributes loop */ 852 } /* End challenge loop */ 853 854 /* 855 * Challenges max count exceeded. This really can't happen 856 * with normal data, something's fishy -> error 857 */ 858 859 out: 860 if (key) 861 free(key); 862 if (value) 863 free(value); 864 if (buf) 865 free(buf); 866 return (ret); 867 } 868 869 870 /* 871 * Parse a last-modified header 872 */ 873 static int 874 http_parse_mtime(const char *p, time_t *mtime) 875 { 876 char locale[64], *r; 877 struct tm tm; 878 879 strlcpy(locale, setlocale(LC_TIME, NULL), sizeof(locale)); 880 setlocale(LC_TIME, "C"); 881 r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm); 882 /* 883 * Some proxies use UTC in response, but it should still be 884 * parsed. RFC2616 states GMT and UTC are exactly equal for HTTP. 885 */ 886 if (r == NULL) 887 r = strptime(p, "%a, %d %b %Y %H:%M:%S UTC", &tm); 888 /* XXX should add support for date-2 and date-3 */ 889 setlocale(LC_TIME, locale); 890 if (r == NULL) 891 return (-1); 892 DEBUG(fprintf(stderr, "last modified: [%04d-%02d-%02d " 893 "%02d:%02d:%02d]\n", 894 tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday, 895 tm.tm_hour, tm.tm_min, tm.tm_sec)); 896 *mtime = timegm(&tm); 897 return (0); 898 } 899 900 /* 901 * Parse a content-length header 902 */ 903 static int 904 http_parse_length(const char *p, off_t *length) 905 { 906 off_t len; 907 908 for (len = 0; *p && isdigit((unsigned char)*p); ++p) 909 len = len * 10 + (*p - '0'); 910 if (*p) 911 return (-1); 912 DEBUG(fprintf(stderr, "content length: [%lld]\n", 913 (long long)len)); 914 *length = len; 915 return (0); 916 } 917 918 /* 919 * Parse a content-range header 920 */ 921 static int 922 http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size) 923 { 924 off_t first, last, len; 925 926 if (strncasecmp(p, "bytes ", 6) != 0) 927 return (-1); 928 p += 6; 929 if (*p == '*') { 930 first = last = -1; 931 ++p; 932 } else { 933 for (first = 0; *p && isdigit((unsigned char)*p); ++p) 934 first = first * 10 + *p - '0'; 935 if (*p != '-') 936 return (-1); 937 for (last = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 938 last = last * 10 + *p - '0'; 939 } 940 if (first > last || *p != '/') 941 return (-1); 942 for (len = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 943 len = len * 10 + *p - '0'; 944 if (*p || len < last - first + 1) 945 return (-1); 946 if (first == -1) { 947 DEBUG(fprintf(stderr, "content range: [*/%lld]\n", 948 (long long)len)); 949 *length = 0; 950 } else { 951 DEBUG(fprintf(stderr, "content range: [%lld-%lld/%lld]\n", 952 (long long)first, (long long)last, (long long)len)); 953 *length = last - first + 1; 954 } 955 *offset = first; 956 *size = len; 957 return (0); 958 } 959 960 961 /***************************************************************************** 962 * Helper functions for authorization 963 */ 964 965 /* 966 * Base64 encoding 967 */ 968 static char * 969 http_base64(const char *src) 970 { 971 static const char base64[] = 972 "ABCDEFGHIJKLMNOPQRSTUVWXYZ" 973 "abcdefghijklmnopqrstuvwxyz" 974 "0123456789+/"; 975 char *str, *dst; 976 size_t l; 977 int t, r; 978 979 l = strlen(src); 980 if ((str = malloc(((l + 2) / 3) * 4 + 1)) == NULL) 981 return (NULL); 982 dst = str; 983 r = 0; 984 985 while (l >= 3) { 986 t = (src[0] << 16) | (src[1] << 8) | src[2]; 987 dst[0] = base64[(t >> 18) & 0x3f]; 988 dst[1] = base64[(t >> 12) & 0x3f]; 989 dst[2] = base64[(t >> 6) & 0x3f]; 990 dst[3] = base64[(t >> 0) & 0x3f]; 991 src += 3; l -= 3; 992 dst += 4; r += 4; 993 } 994 995 switch (l) { 996 case 2: 997 t = (src[0] << 16) | (src[1] << 8); 998 dst[0] = base64[(t >> 18) & 0x3f]; 999 dst[1] = base64[(t >> 12) & 0x3f]; 1000 dst[2] = base64[(t >> 6) & 0x3f]; 1001 dst[3] = '='; 1002 dst += 4; 1003 r += 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 r += 4; 1012 break; 1013 case 0: 1014 break; 1015 } 1016 1017 *dst = 0; 1018 return (str); 1019 } 1020 1021 1022 /* 1023 * Extract authorization parameters from environment value. 1024 * The value is like scheme:realm:user:pass 1025 */ 1026 typedef struct { 1027 char *scheme; 1028 char *realm; 1029 char *user; 1030 char *password; 1031 } http_auth_params_t; 1032 1033 static void 1034 init_http_auth_params(http_auth_params_t *s) 1035 { 1036 s->scheme = s->realm = s->user = s->password = NULL; 1037 } 1038 1039 static void 1040 clean_http_auth_params(http_auth_params_t *s) 1041 { 1042 if (s->scheme) 1043 free(s->scheme); 1044 if (s->realm) 1045 free(s->realm); 1046 if (s->user) 1047 free(s->user); 1048 if (s->password) 1049 free(s->password); 1050 init_http_auth_params(s); 1051 } 1052 1053 static int 1054 http_authfromenv(const char *p, http_auth_params_t *parms) 1055 { 1056 int ret = -1; 1057 char *v, *ve; 1058 char *str = strdup(p); 1059 1060 if (str == NULL) { 1061 fetch_syserr(); 1062 return (-1); 1063 } 1064 v = str; 1065 1066 if ((ve = strchr(v, ':')) == NULL) 1067 goto out; 1068 1069 *ve = 0; 1070 if ((parms->scheme = strdup(v)) == NULL) { 1071 fetch_syserr(); 1072 goto out; 1073 } 1074 v = ve + 1; 1075 1076 if ((ve = strchr(v, ':')) == NULL) 1077 goto out; 1078 1079 *ve = 0; 1080 if ((parms->realm = strdup(v)) == NULL) { 1081 fetch_syserr(); 1082 goto out; 1083 } 1084 v = ve + 1; 1085 1086 if ((ve = strchr(v, ':')) == NULL) 1087 goto out; 1088 1089 *ve = 0; 1090 if ((parms->user = strdup(v)) == NULL) { 1091 fetch_syserr(); 1092 goto out; 1093 } 1094 v = ve + 1; 1095 1096 1097 if ((parms->password = strdup(v)) == NULL) { 1098 fetch_syserr(); 1099 goto out; 1100 } 1101 ret = 0; 1102 out: 1103 if (ret == -1) 1104 clean_http_auth_params(parms); 1105 if (str) 1106 free(str); 1107 return (ret); 1108 } 1109 1110 1111 /* 1112 * Digest response: the code to compute the digest is taken from the 1113 * sample implementation in RFC2616 1114 */ 1115 #define IN const 1116 #define OUT 1117 1118 #define HASHLEN 16 1119 typedef char HASH[HASHLEN]; 1120 #define HASHHEXLEN 32 1121 typedef char HASHHEX[HASHHEXLEN+1]; 1122 1123 static const char *hexchars = "0123456789abcdef"; 1124 static void 1125 CvtHex(IN HASH Bin, OUT HASHHEX Hex) 1126 { 1127 unsigned short i; 1128 unsigned char j; 1129 1130 for (i = 0; i < HASHLEN; i++) { 1131 j = (Bin[i] >> 4) & 0xf; 1132 Hex[i*2] = hexchars[j]; 1133 j = Bin[i] & 0xf; 1134 Hex[i*2+1] = hexchars[j]; 1135 } 1136 Hex[HASHHEXLEN] = '\0'; 1137 }; 1138 1139 /* calculate H(A1) as per spec */ 1140 static void 1141 DigestCalcHA1( 1142 IN char * pszAlg, 1143 IN char * pszUserName, 1144 IN char * pszRealm, 1145 IN char * pszPassword, 1146 IN char * pszNonce, 1147 IN char * pszCNonce, 1148 OUT HASHHEX SessionKey 1149 ) 1150 { 1151 MD5_CTX Md5Ctx; 1152 HASH HA1; 1153 1154 MD5Init(&Md5Ctx); 1155 MD5Update(&Md5Ctx, pszUserName, strlen(pszUserName)); 1156 MD5Update(&Md5Ctx, ":", 1); 1157 MD5Update(&Md5Ctx, pszRealm, strlen(pszRealm)); 1158 MD5Update(&Md5Ctx, ":", 1); 1159 MD5Update(&Md5Ctx, pszPassword, strlen(pszPassword)); 1160 MD5Final(HA1, &Md5Ctx); 1161 if (strcasecmp(pszAlg, "md5-sess") == 0) { 1162 1163 MD5Init(&Md5Ctx); 1164 MD5Update(&Md5Ctx, HA1, HASHLEN); 1165 MD5Update(&Md5Ctx, ":", 1); 1166 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce)); 1167 MD5Update(&Md5Ctx, ":", 1); 1168 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce)); 1169 MD5Final(HA1, &Md5Ctx); 1170 } 1171 CvtHex(HA1, SessionKey); 1172 } 1173 1174 /* calculate request-digest/response-digest as per HTTP Digest spec */ 1175 static void 1176 DigestCalcResponse( 1177 IN HASHHEX HA1, /* H(A1) */ 1178 IN char * pszNonce, /* nonce from server */ 1179 IN char * pszNonceCount, /* 8 hex digits */ 1180 IN char * pszCNonce, /* client nonce */ 1181 IN char * pszQop, /* qop-value: "", "auth", "auth-int" */ 1182 IN char * pszMethod, /* method from the request */ 1183 IN char * pszDigestUri, /* requested URL */ 1184 IN HASHHEX HEntity, /* H(entity body) if qop="auth-int" */ 1185 OUT HASHHEX Response /* request-digest or response-digest */ 1186 ) 1187 { 1188 /* DEBUG(fprintf(stderr, 1189 "Calc: HA1[%s] Nonce[%s] qop[%s] method[%s] URI[%s]\n", 1190 HA1, pszNonce, pszQop, pszMethod, pszDigestUri));*/ 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 DEBUG(fprintf(stderr, "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:"")== -1) 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 DEBUG(fprintf(stderr, "HA1: [%s]\n", HA1)); 1284 HASHHEX digest; 1285 DigestCalcResponse(HA1, c->nonce, noncecount, cnonce, c->qop, 1286 "GET", url->doc, "", digest); 1287 1288 if (c->qop[0]) { 1289 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\"," 1290 "nonce=\"%s\",uri=\"%s\",response=\"%s\"," 1291 "qop=\"auth\", cnonce=\"%s\", nc=%s%s", 1292 hdr, parms->user, c->realm, 1293 c->nonce, url->doc, digest, 1294 cnonce, noncecount, options); 1295 } else { 1296 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\"," 1297 "nonce=\"%s\",uri=\"%s\",response=\"%s\"%s", 1298 hdr, parms->user, c->realm, 1299 c->nonce, url->doc, digest, options); 1300 } 1301 if (options) 1302 free(options); 1303 return (r); 1304 } 1305 1306 /* 1307 * Encode username and password 1308 */ 1309 static int 1310 http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd) 1311 { 1312 char *upw, *auth; 1313 int r; 1314 1315 DEBUG(fprintf(stderr, "basic: usr: [%s]\n", usr)); 1316 DEBUG(fprintf(stderr, "basic: pwd: [%s]\n", pwd)); 1317 if (asprintf(&upw, "%s:%s", usr, pwd) == -1) 1318 return (-1); 1319 auth = http_base64(upw); 1320 free(upw); 1321 if (auth == NULL) 1322 return (-1); 1323 r = http_cmd(conn, "%s: Basic %s", hdr, auth); 1324 free(auth); 1325 return (r); 1326 } 1327 1328 /* 1329 * Chose the challenge to answer and call the appropriate routine to 1330 * produce the header. 1331 */ 1332 static int 1333 http_authorize(conn_t *conn, const char *hdr, http_auth_challenges_t *cs, 1334 http_auth_params_t *parms, struct url *url) 1335 { 1336 http_auth_challenge_t *digest = NULL; 1337 int i; 1338 1339 /* If user or pass are null we're not happy */ 1340 if (!parms->user || !parms->password) { 1341 DEBUG(fprintf(stderr, "NULL usr or pass\n")); 1342 return (-1); 1343 } 1344 1345 /* Look for a Digest */ 1346 for (i = 0; i < cs->count; i++) { 1347 if (cs->challenges[i]->scheme == HTTPAS_DIGEST) 1348 digest = cs->challenges[i]; 1349 } 1350 1351 /* Error if "Digest" was specified and there is no Digest challenge */ 1352 if (!digest && (parms->scheme && 1353 !strcasecmp(parms->scheme, "digest"))) { 1354 DEBUG(fprintf(stderr, 1355 "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 || (parms->scheme && !strcasecmp(parms->scheme,"basic"))) 1364 return (http_basic_auth(conn,hdr,parms->user,parms->password)); 1365 1366 /* Else, prefer digest. We just checked that it's not NULL */ 1367 return (http_digest_auth(conn, hdr, digest, parms, url)); 1368 } 1369 1370 /***************************************************************************** 1371 * Helper functions for connecting to a server or proxy 1372 */ 1373 1374 /* 1375 * Connect to the correct HTTP server or proxy. 1376 */ 1377 static conn_t * 1378 http_connect(struct url *URL, struct url *purl, const char *flags) 1379 { 1380 struct url *curl; 1381 conn_t *conn; 1382 hdr_t h; 1383 http_headerbuf_t headerbuf; 1384 const char *p; 1385 int verbose; 1386 int af, val; 1387 int serrno; 1388 1389 #ifdef INET6 1390 af = AF_UNSPEC; 1391 #else 1392 af = AF_INET; 1393 #endif 1394 1395 verbose = CHECK_FLAG('v'); 1396 if (CHECK_FLAG('4')) 1397 af = AF_INET; 1398 #ifdef INET6 1399 else if (CHECK_FLAG('6')) 1400 af = AF_INET6; 1401 #endif 1402 1403 curl = (purl != NULL) ? purl : URL; 1404 1405 if ((conn = fetch_connect(curl->host, curl->port, af, verbose)) == NULL) 1406 /* fetch_connect() has already set an error code */ 1407 return (NULL); 1408 init_http_headerbuf(&headerbuf); 1409 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 && purl) { 1410 http_cmd(conn, "CONNECT %s:%d HTTP/1.1", 1411 URL->host, URL->port); 1412 http_cmd(conn, "Host: %s:%d", 1413 URL->host, URL->port); 1414 http_cmd(conn, ""); 1415 if (http_get_reply(conn) != HTTP_OK) { 1416 http_seterr(conn->err); 1417 goto ouch; 1418 } 1419 /* Read and discard the rest of the proxy response */ 1420 if (fetch_getln(conn) < 0) { 1421 fetch_syserr(); 1422 goto ouch; 1423 } 1424 do { 1425 switch ((h = http_next_header(conn, &headerbuf, &p))) { 1426 case hdr_syserror: 1427 fetch_syserr(); 1428 goto ouch; 1429 case hdr_error: 1430 http_seterr(HTTP_PROTOCOL_ERROR); 1431 goto ouch; 1432 default: 1433 /* ignore */ ; 1434 } 1435 } while (h < hdr_end); 1436 } 1437 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 && 1438 fetch_ssl(conn, URL, verbose) == -1) { 1439 /* grrr */ 1440 errno = EAUTH; 1441 fetch_syserr(); 1442 goto ouch; 1443 } 1444 1445 val = 1; 1446 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val)); 1447 1448 clean_http_headerbuf(&headerbuf); 1449 return (conn); 1450 ouch: 1451 serrno = errno; 1452 clean_http_headerbuf(&headerbuf); 1453 fetch_close(conn); 1454 errno = serrno; 1455 return (NULL); 1456 } 1457 1458 static struct url * 1459 http_get_proxy(struct url * url, const char *flags) 1460 { 1461 struct url *purl; 1462 char *p; 1463 1464 if (flags != NULL && strchr(flags, 'd') != NULL) 1465 return (NULL); 1466 if (fetch_no_proxy_match(url->host)) 1467 return (NULL); 1468 if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) && 1469 *p && (purl = fetchParseURL(p))) { 1470 if (!*purl->scheme) 1471 strcpy(purl->scheme, SCHEME_HTTP); 1472 if (!purl->port) 1473 purl->port = fetch_default_proxy_port(purl->scheme); 1474 if (strcasecmp(purl->scheme, SCHEME_HTTP) == 0) 1475 return (purl); 1476 fetchFreeURL(purl); 1477 } 1478 return (NULL); 1479 } 1480 1481 static void 1482 http_print_html(FILE *out, FILE *in) 1483 { 1484 size_t len; 1485 char *line, *p, *q; 1486 int comment, tag; 1487 1488 comment = tag = 0; 1489 while ((line = fgetln(in, &len)) != NULL) { 1490 while (len && isspace((unsigned char)line[len - 1])) 1491 --len; 1492 for (p = q = line; q < line + len; ++q) { 1493 if (comment && *q == '-') { 1494 if (q + 2 < line + len && 1495 strcmp(q, "-->") == 0) { 1496 tag = comment = 0; 1497 q += 2; 1498 } 1499 } else if (tag && !comment && *q == '>') { 1500 p = q + 1; 1501 tag = 0; 1502 } else if (!tag && *q == '<') { 1503 if (q > p) 1504 fwrite(p, q - p, 1, out); 1505 tag = 1; 1506 if (q + 3 < line + len && 1507 strcmp(q, "<!--") == 0) { 1508 comment = 1; 1509 q += 3; 1510 } 1511 } 1512 } 1513 if (!tag && q > p) 1514 fwrite(p, q - p, 1, out); 1515 fputc('\n', out); 1516 } 1517 } 1518 1519 1520 /***************************************************************************** 1521 * Core 1522 */ 1523 1524 FILE * 1525 http_request(struct url *URL, const char *op, struct url_stat *us, 1526 struct url *purl, const char *flags) 1527 { 1528 1529 return (http_request_body(URL, op, us, purl, flags, NULL, NULL)); 1530 } 1531 1532 /* 1533 * Send a request and process the reply 1534 * 1535 * XXX This function is way too long, the do..while loop should be split 1536 * XXX off into a separate function. 1537 */ 1538 FILE * 1539 http_request_body(struct url *URL, const char *op, struct url_stat *us, 1540 struct url *purl, const char *flags, const char *content_type, 1541 const char *body) 1542 { 1543 char timebuf[80]; 1544 char hbuf[MAXHOSTNAMELEN + 7], *host; 1545 conn_t *conn; 1546 struct url *url, *new; 1547 int chunked, direct, ims, noredirect, verbose; 1548 int e, i, n, val; 1549 off_t offset, clength, length, size; 1550 time_t mtime; 1551 const char *p; 1552 FILE *f; 1553 hdr_t h; 1554 struct tm *timestruct; 1555 http_headerbuf_t headerbuf; 1556 http_auth_challenges_t server_challenges; 1557 http_auth_challenges_t proxy_challenges; 1558 size_t body_len; 1559 1560 /* The following calls don't allocate anything */ 1561 init_http_headerbuf(&headerbuf); 1562 init_http_auth_challenges(&server_challenges); 1563 init_http_auth_challenges(&proxy_challenges); 1564 1565 direct = CHECK_FLAG('d'); 1566 noredirect = CHECK_FLAG('A'); 1567 verbose = CHECK_FLAG('v'); 1568 ims = CHECK_FLAG('i'); 1569 1570 if (direct && purl) { 1571 fetchFreeURL(purl); 1572 purl = NULL; 1573 } 1574 1575 /* try the provided URL first */ 1576 url = URL; 1577 1578 n = MAX_REDIRECT; 1579 i = 0; 1580 1581 e = HTTP_PROTOCOL_ERROR; 1582 do { 1583 new = NULL; 1584 chunked = 0; 1585 offset = 0; 1586 clength = -1; 1587 length = -1; 1588 size = -1; 1589 mtime = 0; 1590 1591 /* check port */ 1592 if (!url->port) 1593 url->port = fetch_default_port(url->scheme); 1594 1595 /* were we redirected to an FTP URL? */ 1596 if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) { 1597 if (strcmp(op, "GET") == 0) 1598 return (ftp_request(url, "RETR", us, purl, flags)); 1599 else if (strcmp(op, "HEAD") == 0) 1600 return (ftp_request(url, "STAT", us, purl, flags)); 1601 } 1602 1603 /* connect to server or proxy */ 1604 if ((conn = http_connect(url, purl, flags)) == NULL) 1605 goto ouch; 1606 1607 host = url->host; 1608 #ifdef INET6 1609 if (strchr(url->host, ':')) { 1610 snprintf(hbuf, sizeof(hbuf), "[%s]", url->host); 1611 host = hbuf; 1612 } 1613 #endif 1614 if (url->port != fetch_default_port(url->scheme)) { 1615 if (host != hbuf) { 1616 strcpy(hbuf, host); 1617 host = hbuf; 1618 } 1619 snprintf(hbuf + strlen(hbuf), 1620 sizeof(hbuf) - strlen(hbuf), ":%d", url->port); 1621 } 1622 1623 /* send request */ 1624 if (verbose) 1625 fetch_info("requesting %s://%s%s", 1626 url->scheme, host, url->doc); 1627 if (purl && strcasecmp(URL->scheme, SCHEME_HTTPS) != 0) { 1628 http_cmd(conn, "%s %s://%s%s HTTP/1.1", 1629 op, url->scheme, host, url->doc); 1630 } else { 1631 http_cmd(conn, "%s %s HTTP/1.1", 1632 op, url->doc); 1633 } 1634 1635 if (ims && url->ims_time) { 1636 timestruct = gmtime((time_t *)&url->ims_time); 1637 (void)strftime(timebuf, 80, "%a, %d %b %Y %T GMT", 1638 timestruct); 1639 if (verbose) 1640 fetch_info("If-Modified-Since: %s", timebuf); 1641 http_cmd(conn, "If-Modified-Since: %s", timebuf); 1642 } 1643 /* virtual host */ 1644 http_cmd(conn, "Host: %s", host); 1645 1646 /* 1647 * Proxy authorization: we only send auth after we received 1648 * a 407 error. We do not first try basic anyway (changed 1649 * when support was added for digest-auth) 1650 */ 1651 if (purl && proxy_challenges.valid) { 1652 http_auth_params_t aparams; 1653 init_http_auth_params(&aparams); 1654 if (*purl->user || *purl->pwd) { 1655 aparams.user = strdup(purl->user); 1656 aparams.password = strdup(purl->pwd); 1657 } else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL && 1658 *p != '\0') { 1659 if (http_authfromenv(p, &aparams) < 0) { 1660 http_seterr(HTTP_NEED_PROXY_AUTH); 1661 goto ouch; 1662 } 1663 } else if (fetch_netrc_auth(purl) == 0) { 1664 aparams.user = strdup(purl->user); 1665 aparams.password = strdup(purl->pwd); 1666 } 1667 http_authorize(conn, "Proxy-Authorization", 1668 &proxy_challenges, &aparams, url); 1669 clean_http_auth_params(&aparams); 1670 } 1671 1672 /* 1673 * Server authorization: we never send "a priori" 1674 * Basic auth, which used to be done if user/pass were 1675 * set in the url. This would be weird because we'd send the 1676 * password in the clear even if Digest is finally to be 1677 * used (it would have made more sense for the 1678 * pre-digest version to do this when Basic was specified 1679 * in the environment) 1680 */ 1681 if (server_challenges.valid) { 1682 http_auth_params_t aparams; 1683 init_http_auth_params(&aparams); 1684 if (*url->user || *url->pwd) { 1685 aparams.user = strdup(url->user); 1686 aparams.password = strdup(url->pwd); 1687 } else if ((p = getenv("HTTP_AUTH")) != NULL && 1688 *p != '\0') { 1689 if (http_authfromenv(p, &aparams) < 0) { 1690 http_seterr(HTTP_NEED_AUTH); 1691 goto ouch; 1692 } 1693 } else if (fetch_netrc_auth(url) == 0) { 1694 aparams.user = strdup(url->user); 1695 aparams.password = strdup(url->pwd); 1696 } else if (fetchAuthMethod && 1697 fetchAuthMethod(url) == 0) { 1698 aparams.user = strdup(url->user); 1699 aparams.password = strdup(url->pwd); 1700 } else { 1701 http_seterr(HTTP_NEED_AUTH); 1702 goto ouch; 1703 } 1704 http_authorize(conn, "Authorization", 1705 &server_challenges, &aparams, url); 1706 clean_http_auth_params(&aparams); 1707 } 1708 1709 /* other headers */ 1710 if ((p = getenv("HTTP_ACCEPT")) != NULL) { 1711 if (*p != '\0') 1712 http_cmd(conn, "Accept: %s", p); 1713 } else { 1714 http_cmd(conn, "Accept: */*"); 1715 } 1716 if ((p = getenv("HTTP_REFERER")) != NULL && *p != '\0') { 1717 if (strcasecmp(p, "auto") == 0) 1718 http_cmd(conn, "Referer: %s://%s%s", 1719 url->scheme, host, url->doc); 1720 else 1721 http_cmd(conn, "Referer: %s", p); 1722 } 1723 if ((p = getenv("HTTP_USER_AGENT")) != NULL) { 1724 /* no User-Agent if defined but empty */ 1725 if (*p != '\0') 1726 http_cmd(conn, "User-Agent: %s", p); 1727 } else { 1728 /* default User-Agent */ 1729 http_cmd(conn, "User-Agent: %s " _LIBFETCH_VER, 1730 getprogname()); 1731 } 1732 if (url->offset > 0) 1733 http_cmd(conn, "Range: bytes=%lld-", (long long)url->offset); 1734 http_cmd(conn, "Connection: close"); 1735 1736 if (body) { 1737 body_len = strlen(body); 1738 http_cmd(conn, "Content-Length: %zu", body_len); 1739 if (content_type != NULL) 1740 http_cmd(conn, "Content-Type: %s", content_type); 1741 } 1742 1743 http_cmd(conn, ""); 1744 1745 if (body) 1746 fetch_write(conn, body, body_len); 1747 1748 /* 1749 * Force the queued request to be dispatched. Normally, one 1750 * would do this with shutdown(2) but squid proxies can be 1751 * configured to disallow such half-closed connections. To 1752 * be compatible with such configurations, fiddle with socket 1753 * options to force the pending data to be written. 1754 */ 1755 val = 0; 1756 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, 1757 sizeof(val)); 1758 val = 1; 1759 setsockopt(conn->sd, IPPROTO_TCP, TCP_NODELAY, &val, 1760 sizeof(val)); 1761 1762 /* get reply */ 1763 switch (http_get_reply(conn)) { 1764 case HTTP_OK: 1765 case HTTP_PARTIAL: 1766 case HTTP_NOT_MODIFIED: 1767 /* fine */ 1768 break; 1769 case HTTP_MOVED_PERM: 1770 case HTTP_MOVED_TEMP: 1771 case HTTP_TEMP_REDIRECT: 1772 case HTTP_PERM_REDIRECT: 1773 case HTTP_SEE_OTHER: 1774 case HTTP_USE_PROXY: 1775 /* 1776 * Not so fine, but we still have to read the 1777 * headers to get the new location. 1778 */ 1779 break; 1780 case HTTP_NEED_AUTH: 1781 if (server_challenges.valid) { 1782 /* 1783 * We already sent out authorization code, 1784 * so there's nothing more we can do. 1785 */ 1786 http_seterr(conn->err); 1787 goto ouch; 1788 } 1789 /* try again, but send the password this time */ 1790 if (verbose) 1791 fetch_info("server requires authorization"); 1792 break; 1793 case HTTP_NEED_PROXY_AUTH: 1794 if (proxy_challenges.valid) { 1795 /* 1796 * We already sent our proxy 1797 * authorization code, so there's 1798 * nothing more we can do. */ 1799 http_seterr(conn->err); 1800 goto ouch; 1801 } 1802 /* try again, but send the password this time */ 1803 if (verbose) 1804 fetch_info("proxy requires authorization"); 1805 break; 1806 case HTTP_BAD_RANGE: 1807 /* 1808 * This can happen if we ask for 0 bytes because 1809 * we already have the whole file. Consider this 1810 * a success for now, and check sizes later. 1811 */ 1812 break; 1813 case HTTP_PROTOCOL_ERROR: 1814 /* fall through */ 1815 case -1: 1816 fetch_syserr(); 1817 goto ouch; 1818 default: 1819 http_seterr(conn->err); 1820 if (!verbose) 1821 goto ouch; 1822 /* fall through so we can get the full error message */ 1823 } 1824 1825 /* get headers. http_next_header expects one line readahead */ 1826 if (fetch_getln(conn) == -1) { 1827 fetch_syserr(); 1828 goto ouch; 1829 } 1830 do { 1831 switch ((h = http_next_header(conn, &headerbuf, &p))) { 1832 case hdr_syserror: 1833 fetch_syserr(); 1834 goto ouch; 1835 case hdr_error: 1836 http_seterr(HTTP_PROTOCOL_ERROR); 1837 goto ouch; 1838 case hdr_content_length: 1839 http_parse_length(p, &clength); 1840 break; 1841 case hdr_content_range: 1842 http_parse_range(p, &offset, &length, &size); 1843 break; 1844 case hdr_last_modified: 1845 http_parse_mtime(p, &mtime); 1846 break; 1847 case hdr_location: 1848 if (!HTTP_REDIRECT(conn->err)) 1849 break; 1850 /* 1851 * if the A flag is set, we don't follow 1852 * temporary redirects. 1853 */ 1854 if (noredirect && 1855 conn->err != HTTP_MOVED_PERM && 1856 conn->err != HTTP_PERM_REDIRECT && 1857 conn->err != HTTP_USE_PROXY) { 1858 n = 1; 1859 break; 1860 } 1861 if (new) 1862 free(new); 1863 if (verbose) 1864 fetch_info("%d redirect to %s", conn->err, p); 1865 if (*p == '/') 1866 /* absolute path */ 1867 new = fetchMakeURL(url->scheme, url->host, url->port, p, 1868 url->user, url->pwd); 1869 else 1870 new = fetchParseURL(p); 1871 if (new == NULL) { 1872 /* XXX should set an error code */ 1873 DEBUG(fprintf(stderr, "failed to parse new URL\n")); 1874 goto ouch; 1875 } 1876 1877 /* Only copy credentials if the host matches */ 1878 if (!strcmp(new->host, url->host) && !*new->user && !*new->pwd) { 1879 strcpy(new->user, url->user); 1880 strcpy(new->pwd, url->pwd); 1881 } 1882 new->offset = url->offset; 1883 new->length = url->length; 1884 break; 1885 case hdr_transfer_encoding: 1886 /* XXX weak test*/ 1887 chunked = (strcasecmp(p, "chunked") == 0); 1888 break; 1889 case hdr_www_authenticate: 1890 if (conn->err != HTTP_NEED_AUTH) 1891 break; 1892 if (http_parse_authenticate(p, &server_challenges) == 0) 1893 ++n; 1894 break; 1895 case hdr_proxy_authenticate: 1896 if (conn->err != HTTP_NEED_PROXY_AUTH) 1897 break; 1898 if (http_parse_authenticate(p, &proxy_challenges) == 0) 1899 ++n; 1900 break; 1901 case hdr_end: 1902 /* fall through */ 1903 case hdr_unknown: 1904 /* ignore */ 1905 break; 1906 } 1907 } while (h > hdr_end); 1908 1909 /* we need to provide authentication */ 1910 if (conn->err == HTTP_NEED_AUTH || 1911 conn->err == HTTP_NEED_PROXY_AUTH) { 1912 e = conn->err; 1913 if ((conn->err == HTTP_NEED_AUTH && 1914 !server_challenges.valid) || 1915 (conn->err == HTTP_NEED_PROXY_AUTH && 1916 !proxy_challenges.valid)) { 1917 /* 401/7 but no www/proxy-authenticate ?? */ 1918 DEBUG(fprintf(stderr, "401/7 and no auth header\n")); 1919 goto ouch; 1920 } 1921 fetch_close(conn); 1922 conn = NULL; 1923 continue; 1924 } 1925 1926 /* requested range not satisfiable */ 1927 if (conn->err == HTTP_BAD_RANGE) { 1928 if (url->offset == size && url->length == 0) { 1929 /* asked for 0 bytes; fake it */ 1930 offset = url->offset; 1931 clength = -1; 1932 conn->err = HTTP_OK; 1933 break; 1934 } else { 1935 http_seterr(conn->err); 1936 goto ouch; 1937 } 1938 } 1939 1940 /* we have a hit or an error */ 1941 if (conn->err == HTTP_OK 1942 || conn->err == HTTP_NOT_MODIFIED 1943 || conn->err == HTTP_PARTIAL 1944 || HTTP_ERROR(conn->err)) 1945 break; 1946 1947 /* all other cases: we got a redirect */ 1948 e = conn->err; 1949 clean_http_auth_challenges(&server_challenges); 1950 fetch_close(conn); 1951 conn = NULL; 1952 if (!new) { 1953 DEBUG(fprintf(stderr, "redirect with no new location\n")); 1954 break; 1955 } 1956 if (url != URL) 1957 fetchFreeURL(url); 1958 url = new; 1959 } while (++i < n); 1960 1961 /* we failed, or ran out of retries */ 1962 if (conn == NULL) { 1963 http_seterr(e); 1964 goto ouch; 1965 } 1966 1967 DEBUG(fprintf(stderr, "offset %lld, length %lld," 1968 " size %lld, clength %lld\n", 1969 (long long)offset, (long long)length, 1970 (long long)size, (long long)clength)); 1971 1972 if (conn->err == HTTP_NOT_MODIFIED) { 1973 http_seterr(HTTP_NOT_MODIFIED); 1974 return (NULL); 1975 } 1976 1977 /* check for inconsistencies */ 1978 if (clength != -1 && length != -1 && clength != length) { 1979 http_seterr(HTTP_PROTOCOL_ERROR); 1980 goto ouch; 1981 } 1982 if (clength == -1) 1983 clength = length; 1984 if (clength != -1) 1985 length = offset + clength; 1986 if (length != -1 && size != -1 && length != size) { 1987 http_seterr(HTTP_PROTOCOL_ERROR); 1988 goto ouch; 1989 } 1990 if (size == -1) 1991 size = length; 1992 1993 /* fill in stats */ 1994 if (us) { 1995 us->size = size; 1996 us->atime = us->mtime = mtime; 1997 } 1998 1999 /* too far? */ 2000 if (URL->offset > 0 && offset > URL->offset) { 2001 http_seterr(HTTP_PROTOCOL_ERROR); 2002 goto ouch; 2003 } 2004 2005 /* report back real offset and size */ 2006 URL->offset = offset; 2007 URL->length = clength; 2008 2009 /* wrap it up in a FILE */ 2010 if ((f = http_funopen(conn, chunked)) == NULL) { 2011 fetch_syserr(); 2012 goto ouch; 2013 } 2014 2015 if (url != URL) 2016 fetchFreeURL(url); 2017 if (purl) 2018 fetchFreeURL(purl); 2019 2020 if (HTTP_ERROR(conn->err)) { 2021 http_print_html(stderr, f); 2022 fclose(f); 2023 f = NULL; 2024 } 2025 clean_http_headerbuf(&headerbuf); 2026 clean_http_auth_challenges(&server_challenges); 2027 clean_http_auth_challenges(&proxy_challenges); 2028 return (f); 2029 2030 ouch: 2031 if (url != URL) 2032 fetchFreeURL(url); 2033 if (purl) 2034 fetchFreeURL(purl); 2035 if (conn != NULL) 2036 fetch_close(conn); 2037 clean_http_headerbuf(&headerbuf); 2038 clean_http_auth_challenges(&server_challenges); 2039 clean_http_auth_challenges(&proxy_challenges); 2040 return (NULL); 2041 } 2042 2043 2044 /***************************************************************************** 2045 * Entry points 2046 */ 2047 2048 /* 2049 * Retrieve and stat a file by HTTP 2050 */ 2051 FILE * 2052 fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags) 2053 { 2054 return (http_request(URL, "GET", us, http_get_proxy(URL, flags), flags)); 2055 } 2056 2057 /* 2058 * Retrieve a file by HTTP 2059 */ 2060 FILE * 2061 fetchGetHTTP(struct url *URL, const char *flags) 2062 { 2063 return (fetchXGetHTTP(URL, NULL, flags)); 2064 } 2065 2066 /* 2067 * Store a file by HTTP 2068 */ 2069 FILE * 2070 fetchPutHTTP(struct url *URL __unused, const char *flags __unused) 2071 { 2072 warnx("fetchPutHTTP(): not implemented"); 2073 return (NULL); 2074 } 2075 2076 /* 2077 * Get an HTTP document's metadata 2078 */ 2079 int 2080 fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags) 2081 { 2082 FILE *f; 2083 2084 f = http_request(URL, "HEAD", us, http_get_proxy(URL, flags), flags); 2085 if (f == NULL) 2086 return (-1); 2087 fclose(f); 2088 return (0); 2089 } 2090 2091 /* 2092 * List a directory 2093 */ 2094 struct url_ent * 2095 fetchListHTTP(struct url *url __unused, const char *flags __unused) 2096 { 2097 warnx("fetchListHTTP(): not implemented"); 2098 return (NULL); 2099 } 2100 2101 FILE * 2102 fetchReqHTTP(struct url *URL, const char *method, const char *flags, 2103 const char *content_type, const char *body) 2104 { 2105 2106 return (http_request_body(URL, method, NULL, http_get_proxy(URL, flags), 2107 flags, content_type, body)); 2108 } 2109