1 /*- 2 * Copyright (c) 2000-2014 Dag-Erling Smørgrav 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer 10 * in this position and unchanged. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 3. The name of the author may not be used to endorse or promote products 15 * derived from this software without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 20 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 21 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 22 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 26 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29 #include <sys/cdefs.h> 30 __FBSDID("$FreeBSD$"); 31 32 /* 33 * The following copyright applies to the base64 code: 34 * 35 *- 36 * Copyright 1997 Massachusetts Institute of Technology 37 * 38 * Permission to use, copy, modify, and distribute this software and 39 * its documentation for any purpose and without fee is hereby 40 * granted, provided that both the above copyright notice and this 41 * permission notice appear in all copies, that both the above 42 * copyright notice and this permission notice appear in all 43 * supporting documentation, and that the name of M.I.T. not be used 44 * in advertising or publicity pertaining to distribution of the 45 * software without specific, written prior permission. M.I.T. makes 46 * no representations about the suitability of this software for any 47 * purpose. It is provided "as is" without express or implied 48 * warranty. 49 * 50 * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''. M.I.T. DISCLAIMS 51 * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE, 52 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 53 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT 54 * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 55 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 56 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF 57 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND 58 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 59 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 60 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 61 * SUCH DAMAGE. 62 */ 63 64 #include <sys/param.h> 65 #include <sys/socket.h> 66 #include <sys/time.h> 67 68 #include <ctype.h> 69 #include <err.h> 70 #include <errno.h> 71 #include <locale.h> 72 #include <netdb.h> 73 #include <stdarg.h> 74 #include <stdio.h> 75 #include <stdlib.h> 76 #include <string.h> 77 #include <time.h> 78 #include <unistd.h> 79 80 #ifdef WITH_SSL 81 #include <openssl/md5.h> 82 #define MD5Init(c) MD5_Init(c) 83 #define MD5Update(c, data, len) MD5_Update(c, data, len) 84 #define MD5Final(md, c) MD5_Final(md, c) 85 #else 86 #include <md5.h> 87 #endif 88 89 #include <netinet/in.h> 90 #include <netinet/tcp.h> 91 92 #include "fetch.h" 93 #include "common.h" 94 #include "httperr.h" 95 96 /* Maximum number of redirects to follow */ 97 #define MAX_REDIRECT 20 98 99 /* Symbolic names for reply codes we care about */ 100 #define HTTP_OK 200 101 #define HTTP_PARTIAL 206 102 #define HTTP_MOVED_PERM 301 103 #define HTTP_MOVED_TEMP 302 104 #define HTTP_SEE_OTHER 303 105 #define HTTP_NOT_MODIFIED 304 106 #define HTTP_USE_PROXY 305 107 #define HTTP_TEMP_REDIRECT 307 108 #define HTTP_PERM_REDIRECT 308 109 #define HTTP_NEED_AUTH 401 110 #define HTTP_NEED_PROXY_AUTH 407 111 #define HTTP_BAD_RANGE 416 112 #define HTTP_PROTOCOL_ERROR 999 113 114 #define HTTP_REDIRECT(xyz) ((xyz) == HTTP_MOVED_PERM \ 115 || (xyz) == HTTP_MOVED_TEMP \ 116 || (xyz) == HTTP_TEMP_REDIRECT \ 117 || (xyz) == HTTP_USE_PROXY \ 118 || (xyz) == HTTP_SEE_OTHER) 119 120 #define HTTP_ERROR(xyz) ((xyz) > 400 && (xyz) < 599) 121 122 123 /***************************************************************************** 124 * I/O functions for decoding chunked streams 125 */ 126 127 struct httpio 128 { 129 conn_t *conn; /* connection */ 130 int chunked; /* chunked mode */ 131 char *buf; /* chunk buffer */ 132 size_t bufsize; /* size of chunk buffer */ 133 ssize_t buflen; /* amount of data currently in buffer */ 134 int bufpos; /* current read offset in buffer */ 135 int eof; /* end-of-file flag */ 136 int error; /* error flag */ 137 size_t chunksize; /* remaining size of current chunk */ 138 #ifndef NDEBUG 139 size_t total; 140 #endif 141 }; 142 143 /* 144 * Get next chunk header 145 */ 146 static int 147 http_new_chunk(struct httpio *io) 148 { 149 char *p; 150 151 if (fetch_getln(io->conn) == -1) 152 return (-1); 153 154 if (io->conn->buflen < 2 || !isxdigit((unsigned char)*io->conn->buf)) 155 return (-1); 156 157 for (p = io->conn->buf; *p && !isspace((unsigned char)*p); ++p) { 158 if (*p == ';') 159 break; 160 if (!isxdigit((unsigned char)*p)) 161 return (-1); 162 if (isdigit((unsigned char)*p)) { 163 io->chunksize = io->chunksize * 16 + 164 *p - '0'; 165 } else { 166 io->chunksize = io->chunksize * 16 + 167 10 + tolower((unsigned char)*p) - 'a'; 168 } 169 } 170 171 #ifndef NDEBUG 172 if (fetchDebug) { 173 io->total += io->chunksize; 174 if (io->chunksize == 0) 175 fprintf(stderr, "%s(): end of last chunk\n", __func__); 176 else 177 fprintf(stderr, "%s(): new chunk: %lu (%lu)\n", 178 __func__, (unsigned long)io->chunksize, 179 (unsigned long)io->total); 180 } 181 #endif 182 183 return (io->chunksize); 184 } 185 186 /* 187 * Grow the input buffer to at least len bytes 188 */ 189 static inline int 190 http_growbuf(struct httpio *io, size_t len) 191 { 192 char *tmp; 193 194 if (io->bufsize >= len) 195 return (0); 196 197 if ((tmp = realloc(io->buf, len)) == NULL) 198 return (-1); 199 io->buf = tmp; 200 io->bufsize = len; 201 return (0); 202 } 203 204 /* 205 * Fill the input buffer, do chunk decoding on the fly 206 */ 207 static ssize_t 208 http_fillbuf(struct httpio *io, size_t len) 209 { 210 ssize_t nbytes; 211 char ch; 212 213 if (io->error) 214 return (-1); 215 if (io->eof) 216 return (0); 217 218 if (io->chunked == 0) { 219 if (http_growbuf(io, len) == -1) 220 return (-1); 221 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) { 222 io->error = errno; 223 return (-1); 224 } 225 io->buflen = nbytes; 226 io->bufpos = 0; 227 return (io->buflen); 228 } 229 230 if (io->chunksize == 0) { 231 switch (http_new_chunk(io)) { 232 case -1: 233 io->error = EPROTO; 234 return (-1); 235 case 0: 236 io->eof = 1; 237 return (0); 238 } 239 } 240 241 if (len > io->chunksize) 242 len = io->chunksize; 243 if (http_growbuf(io, len) == -1) 244 return (-1); 245 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) { 246 io->error = errno; 247 return (-1); 248 } 249 io->buflen = nbytes; 250 io->chunksize -= io->buflen; 251 252 if (io->chunksize == 0) { 253 if (fetch_read(io->conn, &ch, 1) != 1 || ch != '\r' || 254 fetch_read(io->conn, &ch, 1) != 1 || ch != '\n') 255 return (-1); 256 } 257 258 io->bufpos = 0; 259 260 return (io->buflen); 261 } 262 263 /* 264 * Read function 265 */ 266 static int 267 http_readfn(void *v, char *buf, int len) 268 { 269 struct httpio *io = (struct httpio *)v; 270 int rlen; 271 272 if (io->error) 273 return (-1); 274 if (io->eof) 275 return (0); 276 277 /* empty buffer */ 278 if (!io->buf || io->bufpos == io->buflen) { 279 if ((rlen = http_fillbuf(io, len)) < 0) { 280 if ((errno = io->error) == EINTR) 281 io->error = 0; 282 return (-1); 283 } else if (rlen == 0) { 284 return (0); 285 } 286 } 287 288 rlen = io->buflen - io->bufpos; 289 if (len < rlen) 290 rlen = len; 291 memcpy(buf, io->buf + io->bufpos, rlen); 292 io->bufpos += rlen; 293 return (rlen); 294 } 295 296 /* 297 * Write function 298 */ 299 static int 300 http_writefn(void *v, const char *buf, int len) 301 { 302 struct httpio *io = (struct httpio *)v; 303 304 return (fetch_write(io->conn, buf, len)); 305 } 306 307 /* 308 * Close function 309 */ 310 static int 311 http_closefn(void *v) 312 { 313 struct httpio *io = (struct httpio *)v; 314 int r; 315 316 r = fetch_close(io->conn); 317 if (io->buf) 318 free(io->buf); 319 free(io); 320 return (r); 321 } 322 323 /* 324 * Wrap a file descriptor up 325 */ 326 static FILE * 327 http_funopen(conn_t *conn, int chunked) 328 { 329 struct httpio *io; 330 FILE *f; 331 332 if ((io = calloc(1, sizeof(*io))) == NULL) { 333 fetch_syserr(); 334 return (NULL); 335 } 336 io->conn = conn; 337 io->chunked = chunked; 338 f = funopen(io, http_readfn, http_writefn, NULL, http_closefn); 339 if (f == NULL) { 340 fetch_syserr(); 341 free(io); 342 return (NULL); 343 } 344 return (f); 345 } 346 347 348 /***************************************************************************** 349 * Helper functions for talking to the server and parsing its replies 350 */ 351 352 /* Header types */ 353 typedef enum { 354 hdr_syserror = -2, 355 hdr_error = -1, 356 hdr_end = 0, 357 hdr_unknown = 1, 358 hdr_content_length, 359 hdr_content_range, 360 hdr_last_modified, 361 hdr_location, 362 hdr_transfer_encoding, 363 hdr_www_authenticate, 364 hdr_proxy_authenticate, 365 } hdr_t; 366 367 /* Names of interesting headers */ 368 static struct { 369 hdr_t num; 370 const char *name; 371 } hdr_names[] = { 372 { hdr_content_length, "Content-Length" }, 373 { hdr_content_range, "Content-Range" }, 374 { hdr_last_modified, "Last-Modified" }, 375 { hdr_location, "Location" }, 376 { hdr_transfer_encoding, "Transfer-Encoding" }, 377 { hdr_www_authenticate, "WWW-Authenticate" }, 378 { hdr_proxy_authenticate, "Proxy-Authenticate" }, 379 { hdr_unknown, NULL }, 380 }; 381 382 /* 383 * Send a formatted line; optionally echo to terminal 384 */ 385 static int 386 http_cmd(conn_t *conn, const char *fmt, ...) 387 { 388 va_list ap; 389 size_t len; 390 char *msg; 391 int r; 392 393 va_start(ap, fmt); 394 len = vasprintf(&msg, fmt, ap); 395 va_end(ap); 396 397 if (msg == NULL) { 398 errno = ENOMEM; 399 fetch_syserr(); 400 return (-1); 401 } 402 403 r = fetch_putln(conn, msg, len); 404 free(msg); 405 406 if (r == -1) { 407 fetch_syserr(); 408 return (-1); 409 } 410 411 return (0); 412 } 413 414 /* 415 * Get and parse status line 416 */ 417 static int 418 http_get_reply(conn_t *conn) 419 { 420 char *p; 421 422 if (fetch_getln(conn) == -1) 423 return (-1); 424 /* 425 * A valid status line looks like "HTTP/m.n xyz reason" where m 426 * and n are the major and minor protocol version numbers and xyz 427 * is the reply code. 428 * Unfortunately, there are servers out there (NCSA 1.5.1, to name 429 * just one) that do not send a version number, so we can't rely 430 * on finding one, but if we do, insist on it being 1.0 or 1.1. 431 * We don't care about the reason phrase. 432 */ 433 if (strncmp(conn->buf, "HTTP", 4) != 0) 434 return (HTTP_PROTOCOL_ERROR); 435 p = conn->buf + 4; 436 if (*p == '/') { 437 if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1')) 438 return (HTTP_PROTOCOL_ERROR); 439 p += 4; 440 } 441 if (*p != ' ' || 442 !isdigit((unsigned char)p[1]) || 443 !isdigit((unsigned char)p[2]) || 444 !isdigit((unsigned char)p[3])) 445 return (HTTP_PROTOCOL_ERROR); 446 447 conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0'); 448 return (conn->err); 449 } 450 451 /* 452 * Check a header; if the type matches the given string, return a pointer 453 * to the beginning of the value. 454 */ 455 static const char * 456 http_match(const char *str, const char *hdr) 457 { 458 while (*str && *hdr && 459 tolower((unsigned char)*str++) == tolower((unsigned char)*hdr++)) 460 /* nothing */; 461 if (*str || *hdr != ':') 462 return (NULL); 463 while (*hdr && isspace((unsigned char)*++hdr)) 464 /* nothing */; 465 return (hdr); 466 } 467 468 469 /* 470 * Get the next header and return the appropriate symbolic code. We 471 * need to read one line ahead for checking for a continuation line 472 * belonging to the current header (continuation lines start with 473 * white space). 474 * 475 * We get called with a fresh line already in the conn buffer, either 476 * from the previous http_next_header() invocation, or, the first 477 * time, from a fetch_getln() performed by our caller. 478 * 479 * This stops when we encounter an empty line (we dont read beyond the header 480 * area). 481 * 482 * Note that the "headerbuf" is just a place to return the result. Its 483 * contents are not used for the next call. This means that no cleanup 484 * is needed when ie doing another connection, just call the cleanup when 485 * fully done to deallocate memory. 486 */ 487 488 /* Limit the max number of continuation lines to some reasonable value */ 489 #define HTTP_MAX_CONT_LINES 10 490 491 /* Place into which to build a header from one or several lines */ 492 typedef struct { 493 char *buf; /* buffer */ 494 size_t bufsize; /* buffer size */ 495 size_t buflen; /* length of buffer contents */ 496 } http_headerbuf_t; 497 498 static void 499 init_http_headerbuf(http_headerbuf_t *buf) 500 { 501 buf->buf = NULL; 502 buf->bufsize = 0; 503 buf->buflen = 0; 504 } 505 506 static void 507 clean_http_headerbuf(http_headerbuf_t *buf) 508 { 509 if (buf->buf) 510 free(buf->buf); 511 init_http_headerbuf(buf); 512 } 513 514 /* Remove whitespace at the end of the buffer */ 515 static void 516 http_conn_trimright(conn_t *conn) 517 { 518 while (conn->buflen && 519 isspace((unsigned char)conn->buf[conn->buflen - 1])) 520 conn->buflen--; 521 conn->buf[conn->buflen] = '\0'; 522 } 523 524 static hdr_t 525 http_next_header(conn_t *conn, http_headerbuf_t *hbuf, const char **p) 526 { 527 unsigned int i, len; 528 529 /* 530 * Have to do the stripping here because of the first line. So 531 * it's done twice for the subsequent lines. No big deal 532 */ 533 http_conn_trimright(conn); 534 if (conn->buflen == 0) 535 return (hdr_end); 536 537 /* Copy the line to the headerbuf */ 538 if (hbuf->bufsize < conn->buflen + 1) { 539 if ((hbuf->buf = realloc(hbuf->buf, conn->buflen + 1)) == NULL) 540 return (hdr_syserror); 541 hbuf->bufsize = conn->buflen + 1; 542 } 543 strcpy(hbuf->buf, conn->buf); 544 hbuf->buflen = conn->buflen; 545 546 /* 547 * Fetch possible continuation lines. Stop at 1st non-continuation 548 * and leave it in the conn buffer 549 */ 550 for (i = 0; i < HTTP_MAX_CONT_LINES; i++) { 551 if (fetch_getln(conn) == -1) 552 return (hdr_syserror); 553 554 /* 555 * Note: we carry on the idea from the previous version 556 * that a pure whitespace line is equivalent to an empty 557 * one (so it's not continuation and will be handled when 558 * we are called next) 559 */ 560 http_conn_trimright(conn); 561 if (conn->buf[0] != ' ' && conn->buf[0] != "\t"[0]) 562 break; 563 564 /* Got a continuation line. Concatenate to previous */ 565 len = hbuf->buflen + conn->buflen; 566 if (hbuf->bufsize < len + 1) { 567 len *= 2; 568 if ((hbuf->buf = realloc(hbuf->buf, len + 1)) == NULL) 569 return (hdr_syserror); 570 hbuf->bufsize = len + 1; 571 } 572 strcpy(hbuf->buf + hbuf->buflen, conn->buf); 573 hbuf->buflen += conn->buflen; 574 } 575 576 /* 577 * We could check for malformed headers but we don't really care. 578 * A valid header starts with a token immediately followed by a 579 * colon; a token is any sequence of non-control, non-whitespace 580 * characters except "()<>@,;:\\\"{}". 581 */ 582 for (i = 0; hdr_names[i].num != hdr_unknown; i++) 583 if ((*p = http_match(hdr_names[i].name, hbuf->buf)) != NULL) 584 return (hdr_names[i].num); 585 586 return (hdr_unknown); 587 } 588 589 /************************** 590 * [Proxy-]Authenticate header parsing 591 */ 592 593 /* 594 * Read doublequote-delimited string into output buffer obuf (allocated 595 * by caller, whose responsibility it is to ensure that it's big enough) 596 * cp points to the first char after the initial '"' 597 * Handles \ quoting 598 * Returns pointer to the first char after the terminating double quote, or 599 * NULL for error. 600 */ 601 static const char * 602 http_parse_headerstring(const char *cp, char *obuf) 603 { 604 for (;;) { 605 switch (*cp) { 606 case 0: /* Unterminated string */ 607 *obuf = 0; 608 return (NULL); 609 case '"': /* Ending quote */ 610 *obuf = 0; 611 return (++cp); 612 case '\\': 613 if (*++cp == 0) { 614 *obuf = 0; 615 return (NULL); 616 } 617 /* FALLTHROUGH */ 618 default: 619 *obuf++ = *cp++; 620 } 621 } 622 } 623 624 /* Http auth challenge schemes */ 625 typedef enum {HTTPAS_UNKNOWN, HTTPAS_BASIC,HTTPAS_DIGEST} http_auth_schemes_t; 626 627 /* Data holder for a Basic or Digest challenge. */ 628 typedef struct { 629 http_auth_schemes_t scheme; 630 char *realm; 631 char *qop; 632 char *nonce; 633 char *opaque; 634 char *algo; 635 int stale; 636 int nc; /* Nonce count */ 637 } http_auth_challenge_t; 638 639 static void 640 init_http_auth_challenge(http_auth_challenge_t *b) 641 { 642 b->scheme = HTTPAS_UNKNOWN; 643 b->realm = b->qop = b->nonce = b->opaque = b->algo = NULL; 644 b->stale = b->nc = 0; 645 } 646 647 static void 648 clean_http_auth_challenge(http_auth_challenge_t *b) 649 { 650 if (b->realm) 651 free(b->realm); 652 if (b->qop) 653 free(b->qop); 654 if (b->nonce) 655 free(b->nonce); 656 if (b->opaque) 657 free(b->opaque); 658 if (b->algo) 659 free(b->algo); 660 init_http_auth_challenge(b); 661 } 662 663 /* Data holder for an array of challenges offered in an http response. */ 664 #define MAX_CHALLENGES 10 665 typedef struct { 666 http_auth_challenge_t *challenges[MAX_CHALLENGES]; 667 int count; /* Number of parsed challenges in the array */ 668 int valid; /* We did parse an authenticate header */ 669 } http_auth_challenges_t; 670 671 static void 672 init_http_auth_challenges(http_auth_challenges_t *cs) 673 { 674 int i; 675 for (i = 0; i < MAX_CHALLENGES; i++) 676 cs->challenges[i] = NULL; 677 cs->count = cs->valid = 0; 678 } 679 680 static void 681 clean_http_auth_challenges(http_auth_challenges_t *cs) 682 { 683 int i; 684 /* We rely on non-zero pointers being allocated, not on the count */ 685 for (i = 0; i < MAX_CHALLENGES; i++) { 686 if (cs->challenges[i] != NULL) { 687 clean_http_auth_challenge(cs->challenges[i]); 688 free(cs->challenges[i]); 689 } 690 } 691 init_http_auth_challenges(cs); 692 } 693 694 /* 695 * Enumeration for lexical elements. Separators will be returned as their own 696 * ascii value 697 */ 698 typedef enum {HTTPHL_WORD=256, HTTPHL_STRING=257, HTTPHL_END=258, 699 HTTPHL_ERROR = 259} http_header_lex_t; 700 701 /* 702 * Determine what kind of token comes next and return possible value 703 * in buf, which is supposed to have been allocated big enough by 704 * caller. Advance input pointer and return element type. 705 */ 706 static int 707 http_header_lex(const char **cpp, char *buf) 708 { 709 size_t l; 710 /* Eat initial whitespace */ 711 *cpp += strspn(*cpp, " \t"); 712 if (**cpp == 0) 713 return (HTTPHL_END); 714 715 /* Separator ? */ 716 if (**cpp == ',' || **cpp == '=') 717 return (*((*cpp)++)); 718 719 /* String ? */ 720 if (**cpp == '"') { 721 *cpp = http_parse_headerstring(++*cpp, buf); 722 if (*cpp == NULL) 723 return (HTTPHL_ERROR); 724 return (HTTPHL_STRING); 725 } 726 727 /* Read other token, until separator or whitespace */ 728 l = strcspn(*cpp, " \t,="); 729 memcpy(buf, *cpp, l); 730 buf[l] = 0; 731 *cpp += l; 732 return (HTTPHL_WORD); 733 } 734 735 /* 736 * Read challenges from http xxx-authenticate header and accumulate them 737 * in the challenges list structure. 738 * 739 * Headers with multiple challenges are specified by rfc2617, but 740 * servers (ie: squid) often send them in separate headers instead, 741 * which in turn is forbidden by the http spec (multiple headers with 742 * the same name are only allowed for pure comma-separated lists, see 743 * rfc2616 sec 4.2). 744 * 745 * We support both approaches anyway 746 */ 747 static int 748 http_parse_authenticate(const char *cp, http_auth_challenges_t *cs) 749 { 750 int ret = -1; 751 http_header_lex_t lex; 752 char *key = malloc(strlen(cp) + 1); 753 char *value = malloc(strlen(cp) + 1); 754 char *buf = malloc(strlen(cp) + 1); 755 756 if (key == NULL || value == NULL || buf == NULL) { 757 fetch_syserr(); 758 goto out; 759 } 760 761 /* In any case we've seen the header and we set the valid bit */ 762 cs->valid = 1; 763 764 /* Need word first */ 765 lex = http_header_lex(&cp, key); 766 if (lex != HTTPHL_WORD) 767 goto out; 768 769 /* Loop on challenges */ 770 for (; cs->count < MAX_CHALLENGES; cs->count++) { 771 cs->challenges[cs->count] = 772 malloc(sizeof(http_auth_challenge_t)); 773 if (cs->challenges[cs->count] == NULL) { 774 fetch_syserr(); 775 goto out; 776 } 777 init_http_auth_challenge(cs->challenges[cs->count]); 778 if (!strcasecmp(key, "basic")) { 779 cs->challenges[cs->count]->scheme = HTTPAS_BASIC; 780 } else if (!strcasecmp(key, "digest")) { 781 cs->challenges[cs->count]->scheme = HTTPAS_DIGEST; 782 } else { 783 cs->challenges[cs->count]->scheme = HTTPAS_UNKNOWN; 784 /* 785 * Continue parsing as basic or digest may 786 * follow, and the syntax is the same for 787 * all. We'll just ignore this one when 788 * looking at the list 789 */ 790 } 791 792 /* Loop on attributes */ 793 for (;;) { 794 /* Key */ 795 lex = http_header_lex(&cp, key); 796 if (lex != HTTPHL_WORD) 797 goto out; 798 799 /* Equal sign */ 800 lex = http_header_lex(&cp, buf); 801 if (lex != '=') 802 goto out; 803 804 /* Value */ 805 lex = http_header_lex(&cp, value); 806 if (lex != HTTPHL_WORD && lex != HTTPHL_STRING) 807 goto out; 808 809 if (!strcasecmp(key, "realm")) 810 cs->challenges[cs->count]->realm = 811 strdup(value); 812 else if (!strcasecmp(key, "qop")) 813 cs->challenges[cs->count]->qop = 814 strdup(value); 815 else if (!strcasecmp(key, "nonce")) 816 cs->challenges[cs->count]->nonce = 817 strdup(value); 818 else if (!strcasecmp(key, "opaque")) 819 cs->challenges[cs->count]->opaque = 820 strdup(value); 821 else if (!strcasecmp(key, "algorithm")) 822 cs->challenges[cs->count]->algo = 823 strdup(value); 824 else if (!strcasecmp(key, "stale")) 825 cs->challenges[cs->count]->stale = 826 strcasecmp(value, "no"); 827 /* Else ignore unknown attributes */ 828 829 /* Comma or Next challenge or End */ 830 lex = http_header_lex(&cp, key); 831 /* 832 * If we get a word here, this is the beginning of the 833 * next challenge. Break the attributes loop 834 */ 835 if (lex == HTTPHL_WORD) 836 break; 837 838 if (lex == HTTPHL_END) { 839 /* End while looking for ',' is normal exit */ 840 cs->count++; 841 ret = 0; 842 goto out; 843 } 844 /* Anything else is an error */ 845 if (lex != ',') 846 goto out; 847 848 } /* End attributes loop */ 849 } /* End challenge loop */ 850 851 /* 852 * Challenges max count exceeded. This really can't happen 853 * with normal data, something's fishy -> error 854 */ 855 856 out: 857 if (key) 858 free(key); 859 if (value) 860 free(value); 861 if (buf) 862 free(buf); 863 return (ret); 864 } 865 866 867 /* 868 * Parse a last-modified header 869 */ 870 static int 871 http_parse_mtime(const char *p, time_t *mtime) 872 { 873 char locale[64], *r; 874 struct tm tm; 875 876 strncpy(locale, setlocale(LC_TIME, NULL), sizeof(locale)); 877 setlocale(LC_TIME, "C"); 878 r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm); 879 /* 880 * Some proxies use UTC in response, but it should still be 881 * parsed. RFC2616 states GMT and UTC are exactly equal for HTTP. 882 */ 883 if (r == NULL) 884 r = strptime(p, "%a, %d %b %Y %H:%M:%S UTC", &tm); 885 /* XXX should add support for date-2 and date-3 */ 886 setlocale(LC_TIME, locale); 887 if (r == NULL) 888 return (-1); 889 DEBUG(fprintf(stderr, "last modified: [%04d-%02d-%02d " 890 "%02d:%02d:%02d]\n", 891 tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday, 892 tm.tm_hour, tm.tm_min, tm.tm_sec)); 893 *mtime = timegm(&tm); 894 return (0); 895 } 896 897 /* 898 * Parse a content-length header 899 */ 900 static int 901 http_parse_length(const char *p, off_t *length) 902 { 903 off_t len; 904 905 for (len = 0; *p && isdigit((unsigned char)*p); ++p) 906 len = len * 10 + (*p - '0'); 907 if (*p) 908 return (-1); 909 DEBUG(fprintf(stderr, "content length: [%lld]\n", 910 (long long)len)); 911 *length = len; 912 return (0); 913 } 914 915 /* 916 * Parse a content-range header 917 */ 918 static int 919 http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size) 920 { 921 off_t first, last, len; 922 923 if (strncasecmp(p, "bytes ", 6) != 0) 924 return (-1); 925 p += 6; 926 if (*p == '*') { 927 first = last = -1; 928 ++p; 929 } else { 930 for (first = 0; *p && isdigit((unsigned char)*p); ++p) 931 first = first * 10 + *p - '0'; 932 if (*p != '-') 933 return (-1); 934 for (last = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 935 last = last * 10 + *p - '0'; 936 } 937 if (first > last || *p != '/') 938 return (-1); 939 for (len = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 940 len = len * 10 + *p - '0'; 941 if (*p || len < last - first + 1) 942 return (-1); 943 if (first == -1) { 944 DEBUG(fprintf(stderr, "content range: [*/%lld]\n", 945 (long long)len)); 946 *length = 0; 947 } else { 948 DEBUG(fprintf(stderr, "content range: [%lld-%lld/%lld]\n", 949 (long long)first, (long long)last, (long long)len)); 950 *length = last - first + 1; 951 } 952 *offset = first; 953 *size = len; 954 return (0); 955 } 956 957 958 /***************************************************************************** 959 * Helper functions for authorization 960 */ 961 962 /* 963 * Base64 encoding 964 */ 965 static char * 966 http_base64(const char *src) 967 { 968 static const char base64[] = 969 "ABCDEFGHIJKLMNOPQRSTUVWXYZ" 970 "abcdefghijklmnopqrstuvwxyz" 971 "0123456789+/"; 972 char *str, *dst; 973 size_t l; 974 int t, r; 975 976 l = strlen(src); 977 if ((str = malloc(((l + 2) / 3) * 4 + 1)) == NULL) 978 return (NULL); 979 dst = str; 980 r = 0; 981 982 while (l >= 3) { 983 t = (src[0] << 16) | (src[1] << 8) | src[2]; 984 dst[0] = base64[(t >> 18) & 0x3f]; 985 dst[1] = base64[(t >> 12) & 0x3f]; 986 dst[2] = base64[(t >> 6) & 0x3f]; 987 dst[3] = base64[(t >> 0) & 0x3f]; 988 src += 3; l -= 3; 989 dst += 4; r += 4; 990 } 991 992 switch (l) { 993 case 2: 994 t = (src[0] << 16) | (src[1] << 8); 995 dst[0] = base64[(t >> 18) & 0x3f]; 996 dst[1] = base64[(t >> 12) & 0x3f]; 997 dst[2] = base64[(t >> 6) & 0x3f]; 998 dst[3] = '='; 999 dst += 4; 1000 r += 4; 1001 break; 1002 case 1: 1003 t = src[0] << 16; 1004 dst[0] = base64[(t >> 18) & 0x3f]; 1005 dst[1] = base64[(t >> 12) & 0x3f]; 1006 dst[2] = dst[3] = '='; 1007 dst += 4; 1008 r += 4; 1009 break; 1010 case 0: 1011 break; 1012 } 1013 1014 *dst = 0; 1015 return (str); 1016 } 1017 1018 1019 /* 1020 * Extract authorization parameters from environment value. 1021 * The value is like scheme:realm:user:pass 1022 */ 1023 typedef struct { 1024 char *scheme; 1025 char *realm; 1026 char *user; 1027 char *password; 1028 } http_auth_params_t; 1029 1030 static void 1031 init_http_auth_params(http_auth_params_t *s) 1032 { 1033 s->scheme = s->realm = s->user = s->password = NULL; 1034 } 1035 1036 static void 1037 clean_http_auth_params(http_auth_params_t *s) 1038 { 1039 if (s->scheme) 1040 free(s->scheme); 1041 if (s->realm) 1042 free(s->realm); 1043 if (s->user) 1044 free(s->user); 1045 if (s->password) 1046 free(s->password); 1047 init_http_auth_params(s); 1048 } 1049 1050 static int 1051 http_authfromenv(const char *p, http_auth_params_t *parms) 1052 { 1053 int ret = -1; 1054 char *v, *ve; 1055 char *str = strdup(p); 1056 1057 if (str == NULL) { 1058 fetch_syserr(); 1059 return (-1); 1060 } 1061 v = str; 1062 1063 if ((ve = strchr(v, ':')) == NULL) 1064 goto out; 1065 1066 *ve = 0; 1067 if ((parms->scheme = strdup(v)) == NULL) { 1068 fetch_syserr(); 1069 goto out; 1070 } 1071 v = ve + 1; 1072 1073 if ((ve = strchr(v, ':')) == NULL) 1074 goto out; 1075 1076 *ve = 0; 1077 if ((parms->realm = strdup(v)) == NULL) { 1078 fetch_syserr(); 1079 goto out; 1080 } 1081 v = ve + 1; 1082 1083 if ((ve = strchr(v, ':')) == NULL) 1084 goto out; 1085 1086 *ve = 0; 1087 if ((parms->user = strdup(v)) == NULL) { 1088 fetch_syserr(); 1089 goto out; 1090 } 1091 v = ve + 1; 1092 1093 1094 if ((parms->password = strdup(v)) == NULL) { 1095 fetch_syserr(); 1096 goto out; 1097 } 1098 ret = 0; 1099 out: 1100 if (ret == -1) 1101 clean_http_auth_params(parms); 1102 if (str) 1103 free(str); 1104 return (ret); 1105 } 1106 1107 1108 /* 1109 * Digest response: the code to compute the digest is taken from the 1110 * sample implementation in RFC2616 1111 */ 1112 #define IN const 1113 #define OUT 1114 1115 #define HASHLEN 16 1116 typedef char HASH[HASHLEN]; 1117 #define HASHHEXLEN 32 1118 typedef char HASHHEX[HASHHEXLEN+1]; 1119 1120 static const char *hexchars = "0123456789abcdef"; 1121 static void 1122 CvtHex(IN HASH Bin, OUT HASHHEX Hex) 1123 { 1124 unsigned short i; 1125 unsigned char j; 1126 1127 for (i = 0; i < HASHLEN; i++) { 1128 j = (Bin[i] >> 4) & 0xf; 1129 Hex[i*2] = hexchars[j]; 1130 j = Bin[i] & 0xf; 1131 Hex[i*2+1] = hexchars[j]; 1132 } 1133 Hex[HASHHEXLEN] = '\0'; 1134 }; 1135 1136 /* calculate H(A1) as per spec */ 1137 static void 1138 DigestCalcHA1( 1139 IN char * pszAlg, 1140 IN char * pszUserName, 1141 IN char * pszRealm, 1142 IN char * pszPassword, 1143 IN char * pszNonce, 1144 IN char * pszCNonce, 1145 OUT HASHHEX SessionKey 1146 ) 1147 { 1148 MD5_CTX Md5Ctx; 1149 HASH HA1; 1150 1151 MD5Init(&Md5Ctx); 1152 MD5Update(&Md5Ctx, pszUserName, strlen(pszUserName)); 1153 MD5Update(&Md5Ctx, ":", 1); 1154 MD5Update(&Md5Ctx, pszRealm, strlen(pszRealm)); 1155 MD5Update(&Md5Ctx, ":", 1); 1156 MD5Update(&Md5Ctx, pszPassword, strlen(pszPassword)); 1157 MD5Final(HA1, &Md5Ctx); 1158 if (strcasecmp(pszAlg, "md5-sess") == 0) { 1159 1160 MD5Init(&Md5Ctx); 1161 MD5Update(&Md5Ctx, HA1, HASHLEN); 1162 MD5Update(&Md5Ctx, ":", 1); 1163 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce)); 1164 MD5Update(&Md5Ctx, ":", 1); 1165 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce)); 1166 MD5Final(HA1, &Md5Ctx); 1167 } 1168 CvtHex(HA1, SessionKey); 1169 } 1170 1171 /* calculate request-digest/response-digest as per HTTP Digest spec */ 1172 static void 1173 DigestCalcResponse( 1174 IN HASHHEX HA1, /* H(A1) */ 1175 IN char * pszNonce, /* nonce from server */ 1176 IN char * pszNonceCount, /* 8 hex digits */ 1177 IN char * pszCNonce, /* client nonce */ 1178 IN char * pszQop, /* qop-value: "", "auth", "auth-int" */ 1179 IN char * pszMethod, /* method from the request */ 1180 IN char * pszDigestUri, /* requested URL */ 1181 IN HASHHEX HEntity, /* H(entity body) if qop="auth-int" */ 1182 OUT HASHHEX Response /* request-digest or response-digest */ 1183 ) 1184 { 1185 /* DEBUG(fprintf(stderr, 1186 "Calc: HA1[%s] Nonce[%s] qop[%s] method[%s] URI[%s]\n", 1187 HA1, pszNonce, pszQop, pszMethod, pszDigestUri));*/ 1188 MD5_CTX Md5Ctx; 1189 HASH HA2; 1190 HASH RespHash; 1191 HASHHEX HA2Hex; 1192 1193 // calculate H(A2) 1194 MD5Init(&Md5Ctx); 1195 MD5Update(&Md5Ctx, pszMethod, strlen(pszMethod)); 1196 MD5Update(&Md5Ctx, ":", 1); 1197 MD5Update(&Md5Ctx, pszDigestUri, strlen(pszDigestUri)); 1198 if (strcasecmp(pszQop, "auth-int") == 0) { 1199 MD5Update(&Md5Ctx, ":", 1); 1200 MD5Update(&Md5Ctx, HEntity, HASHHEXLEN); 1201 } 1202 MD5Final(HA2, &Md5Ctx); 1203 CvtHex(HA2, HA2Hex); 1204 1205 // calculate response 1206 MD5Init(&Md5Ctx); 1207 MD5Update(&Md5Ctx, HA1, HASHHEXLEN); 1208 MD5Update(&Md5Ctx, ":", 1); 1209 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce)); 1210 MD5Update(&Md5Ctx, ":", 1); 1211 if (*pszQop) { 1212 MD5Update(&Md5Ctx, pszNonceCount, strlen(pszNonceCount)); 1213 MD5Update(&Md5Ctx, ":", 1); 1214 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce)); 1215 MD5Update(&Md5Ctx, ":", 1); 1216 MD5Update(&Md5Ctx, pszQop, strlen(pszQop)); 1217 MD5Update(&Md5Ctx, ":", 1); 1218 } 1219 MD5Update(&Md5Ctx, HA2Hex, HASHHEXLEN); 1220 MD5Final(RespHash, &Md5Ctx); 1221 CvtHex(RespHash, Response); 1222 } 1223 1224 /* 1225 * Generate/Send a Digest authorization header 1226 * This looks like: [Proxy-]Authorization: credentials 1227 * 1228 * credentials = "Digest" digest-response 1229 * digest-response = 1#( username | realm | nonce | digest-uri 1230 * | response | [ algorithm ] | [cnonce] | 1231 * [opaque] | [message-qop] | 1232 * [nonce-count] | [auth-param] ) 1233 * username = "username" "=" username-value 1234 * username-value = quoted-string 1235 * digest-uri = "uri" "=" digest-uri-value 1236 * digest-uri-value = request-uri ; As specified by HTTP/1.1 1237 * message-qop = "qop" "=" qop-value 1238 * cnonce = "cnonce" "=" cnonce-value 1239 * cnonce-value = nonce-value 1240 * nonce-count = "nc" "=" nc-value 1241 * nc-value = 8LHEX 1242 * response = "response" "=" request-digest 1243 * request-digest = <"> 32LHEX <"> 1244 */ 1245 static int 1246 http_digest_auth(conn_t *conn, const char *hdr, http_auth_challenge_t *c, 1247 http_auth_params_t *parms, struct url *url) 1248 { 1249 int r; 1250 char noncecount[10]; 1251 char cnonce[40]; 1252 char *options = NULL; 1253 1254 if (!c->realm || !c->nonce) { 1255 DEBUG(fprintf(stderr, "realm/nonce not set in challenge\n")); 1256 return(-1); 1257 } 1258 if (!c->algo) 1259 c->algo = strdup(""); 1260 1261 if (asprintf(&options, "%s%s%s%s", 1262 *c->algo? ",algorithm=" : "", c->algo, 1263 c->opaque? ",opaque=" : "", c->opaque?c->opaque:"")== -1) 1264 return (-1); 1265 1266 if (!c->qop) { 1267 c->qop = strdup(""); 1268 *noncecount = 0; 1269 *cnonce = 0; 1270 } else { 1271 c->nc++; 1272 sprintf(noncecount, "%08x", c->nc); 1273 /* We don't try very hard with the cnonce ... */ 1274 sprintf(cnonce, "%x%lx", getpid(), (unsigned long)time(0)); 1275 } 1276 1277 HASHHEX HA1; 1278 DigestCalcHA1(c->algo, parms->user, c->realm, 1279 parms->password, c->nonce, cnonce, HA1); 1280 DEBUG(fprintf(stderr, "HA1: [%s]\n", HA1)); 1281 HASHHEX digest; 1282 DigestCalcResponse(HA1, c->nonce, noncecount, cnonce, c->qop, 1283 "GET", url->doc, "", digest); 1284 1285 if (c->qop[0]) { 1286 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\"," 1287 "nonce=\"%s\",uri=\"%s\",response=\"%s\"," 1288 "qop=\"auth\", cnonce=\"%s\", nc=%s%s", 1289 hdr, parms->user, c->realm, 1290 c->nonce, url->doc, digest, 1291 cnonce, noncecount, options); 1292 } else { 1293 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\"," 1294 "nonce=\"%s\",uri=\"%s\",response=\"%s\"%s", 1295 hdr, parms->user, c->realm, 1296 c->nonce, url->doc, digest, options); 1297 } 1298 if (options) 1299 free(options); 1300 return (r); 1301 } 1302 1303 /* 1304 * Encode username and password 1305 */ 1306 static int 1307 http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd) 1308 { 1309 char *upw, *auth; 1310 int r; 1311 1312 DEBUG(fprintf(stderr, "basic: usr: [%s]\n", usr)); 1313 DEBUG(fprintf(stderr, "basic: pwd: [%s]\n", pwd)); 1314 if (asprintf(&upw, "%s:%s", usr, pwd) == -1) 1315 return (-1); 1316 auth = http_base64(upw); 1317 free(upw); 1318 if (auth == NULL) 1319 return (-1); 1320 r = http_cmd(conn, "%s: Basic %s", hdr, auth); 1321 free(auth); 1322 return (r); 1323 } 1324 1325 /* 1326 * Chose the challenge to answer and call the appropriate routine to 1327 * produce the header. 1328 */ 1329 static int 1330 http_authorize(conn_t *conn, const char *hdr, http_auth_challenges_t *cs, 1331 http_auth_params_t *parms, struct url *url) 1332 { 1333 http_auth_challenge_t *digest = NULL; 1334 int i; 1335 1336 /* If user or pass are null we're not happy */ 1337 if (!parms->user || !parms->password) { 1338 DEBUG(fprintf(stderr, "NULL usr or pass\n")); 1339 return (-1); 1340 } 1341 1342 /* Look for a Digest */ 1343 for (i = 0; i < cs->count; i++) { 1344 if (cs->challenges[i]->scheme == HTTPAS_DIGEST) 1345 digest = cs->challenges[i]; 1346 } 1347 1348 /* Error if "Digest" was specified and there is no Digest challenge */ 1349 if (!digest && (parms->scheme && 1350 !strcasecmp(parms->scheme, "digest"))) { 1351 DEBUG(fprintf(stderr, 1352 "Digest auth in env, not supported by peer\n")); 1353 return (-1); 1354 } 1355 /* 1356 * If "basic" was specified in the environment, or there is no Digest 1357 * challenge, do the basic thing. Don't need a challenge for this, 1358 * so no need to check basic!=NULL 1359 */ 1360 if (!digest || (parms->scheme && !strcasecmp(parms->scheme,"basic"))) 1361 return (http_basic_auth(conn,hdr,parms->user,parms->password)); 1362 1363 /* Else, prefer digest. We just checked that it's not NULL */ 1364 return (http_digest_auth(conn, hdr, digest, parms, url)); 1365 } 1366 1367 /***************************************************************************** 1368 * Helper functions for connecting to a server or proxy 1369 */ 1370 1371 /* 1372 * Connect to the correct HTTP server or proxy. 1373 */ 1374 static conn_t * 1375 http_connect(struct url *URL, struct url *purl, const char *flags) 1376 { 1377 struct url *curl; 1378 conn_t *conn; 1379 hdr_t h; 1380 http_headerbuf_t headerbuf; 1381 const char *p; 1382 int verbose; 1383 int af, val; 1384 int serrno; 1385 1386 #ifdef INET6 1387 af = AF_UNSPEC; 1388 #else 1389 af = AF_INET; 1390 #endif 1391 1392 verbose = CHECK_FLAG('v'); 1393 if (CHECK_FLAG('4')) 1394 af = AF_INET; 1395 #ifdef INET6 1396 else if (CHECK_FLAG('6')) 1397 af = AF_INET6; 1398 #endif 1399 1400 curl = (purl != NULL) ? purl : URL; 1401 1402 if ((conn = fetch_connect(curl->host, curl->port, af, verbose)) == NULL) 1403 /* fetch_connect() has already set an error code */ 1404 return (NULL); 1405 init_http_headerbuf(&headerbuf); 1406 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 && purl) { 1407 http_cmd(conn, "CONNECT %s:%d HTTP/1.1", 1408 URL->host, URL->port); 1409 http_cmd(conn, "Host: %s:%d", 1410 URL->host, URL->port); 1411 http_cmd(conn, ""); 1412 if (http_get_reply(conn) != HTTP_OK) { 1413 http_seterr(conn->err); 1414 goto ouch; 1415 } 1416 /* Read and discard the rest of the proxy response */ 1417 if (fetch_getln(conn) < 0) { 1418 fetch_syserr(); 1419 goto ouch; 1420 } 1421 do { 1422 switch ((h = http_next_header(conn, &headerbuf, &p))) { 1423 case hdr_syserror: 1424 fetch_syserr(); 1425 goto ouch; 1426 case hdr_error: 1427 http_seterr(HTTP_PROTOCOL_ERROR); 1428 goto ouch; 1429 default: 1430 /* ignore */ ; 1431 } 1432 } while (h < hdr_end); 1433 } 1434 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 && 1435 fetch_ssl(conn, URL, verbose) == -1) { 1436 fetch_close(conn); 1437 /* grrr */ 1438 errno = EAUTH; 1439 fetch_syserr(); 1440 goto ouch; 1441 } 1442 1443 val = 1; 1444 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val)); 1445 1446 clean_http_headerbuf(&headerbuf); 1447 return (conn); 1448 ouch: 1449 serrno = errno; 1450 clean_http_headerbuf(&headerbuf); 1451 fetch_close(conn); 1452 errno = serrno; 1453 return (NULL); 1454 } 1455 1456 static struct url * 1457 http_get_proxy(struct url * url, const char *flags) 1458 { 1459 struct url *purl; 1460 char *p; 1461 1462 if (flags != NULL && strchr(flags, 'd') != NULL) 1463 return (NULL); 1464 if (fetch_no_proxy_match(url->host)) 1465 return (NULL); 1466 if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) && 1467 *p && (purl = fetchParseURL(p))) { 1468 if (!*purl->scheme) 1469 strcpy(purl->scheme, SCHEME_HTTP); 1470 if (!purl->port) 1471 purl->port = fetch_default_proxy_port(purl->scheme); 1472 if (strcasecmp(purl->scheme, SCHEME_HTTP) == 0) 1473 return (purl); 1474 fetchFreeURL(purl); 1475 } 1476 return (NULL); 1477 } 1478 1479 static void 1480 http_print_html(FILE *out, FILE *in) 1481 { 1482 size_t len; 1483 char *line, *p, *q; 1484 int comment, tag; 1485 1486 comment = tag = 0; 1487 while ((line = fgetln(in, &len)) != NULL) { 1488 while (len && isspace((unsigned char)line[len - 1])) 1489 --len; 1490 for (p = q = line; q < line + len; ++q) { 1491 if (comment && *q == '-') { 1492 if (q + 2 < line + len && 1493 strcmp(q, "-->") == 0) { 1494 tag = comment = 0; 1495 q += 2; 1496 } 1497 } else if (tag && !comment && *q == '>') { 1498 p = q + 1; 1499 tag = 0; 1500 } else if (!tag && *q == '<') { 1501 if (q > p) 1502 fwrite(p, q - p, 1, out); 1503 tag = 1; 1504 if (q + 3 < line + len && 1505 strcmp(q, "<!--") == 0) { 1506 comment = 1; 1507 q += 3; 1508 } 1509 } 1510 } 1511 if (!tag && q > p) 1512 fwrite(p, q - p, 1, out); 1513 fputc('\n', out); 1514 } 1515 } 1516 1517 1518 /***************************************************************************** 1519 * Core 1520 */ 1521 1522 FILE * 1523 http_request(struct url *URL, const char *op, struct url_stat *us, 1524 struct url *purl, const char *flags) 1525 { 1526 1527 return (http_request_body(URL, op, us, purl, flags, NULL, NULL)); 1528 } 1529 1530 /* 1531 * Send a request and process the reply 1532 * 1533 * XXX This function is way too long, the do..while loop should be split 1534 * XXX off into a separate function. 1535 */ 1536 FILE * 1537 http_request_body(struct url *URL, const char *op, struct url_stat *us, 1538 struct url *purl, const char *flags, const char *content_type, 1539 const char *body) 1540 { 1541 char timebuf[80]; 1542 char hbuf[MAXHOSTNAMELEN + 7], *host; 1543 conn_t *conn; 1544 struct url *url, *new; 1545 int chunked, direct, ims, noredirect, verbose; 1546 int e, i, n, val; 1547 off_t offset, clength, length, size; 1548 time_t mtime; 1549 const char *p; 1550 FILE *f; 1551 hdr_t h; 1552 struct tm *timestruct; 1553 http_headerbuf_t headerbuf; 1554 http_auth_challenges_t server_challenges; 1555 http_auth_challenges_t proxy_challenges; 1556 size_t body_len; 1557 1558 /* The following calls don't allocate anything */ 1559 init_http_headerbuf(&headerbuf); 1560 init_http_auth_challenges(&server_challenges); 1561 init_http_auth_challenges(&proxy_challenges); 1562 1563 direct = CHECK_FLAG('d'); 1564 noredirect = CHECK_FLAG('A'); 1565 verbose = CHECK_FLAG('v'); 1566 ims = CHECK_FLAG('i'); 1567 1568 if (direct && purl) { 1569 fetchFreeURL(purl); 1570 purl = NULL; 1571 } 1572 1573 /* try the provided URL first */ 1574 url = URL; 1575 1576 n = MAX_REDIRECT; 1577 i = 0; 1578 1579 e = HTTP_PROTOCOL_ERROR; 1580 do { 1581 new = NULL; 1582 chunked = 0; 1583 offset = 0; 1584 clength = -1; 1585 length = -1; 1586 size = -1; 1587 mtime = 0; 1588 1589 /* check port */ 1590 if (!url->port) 1591 url->port = fetch_default_port(url->scheme); 1592 1593 /* were we redirected to an FTP URL? */ 1594 if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) { 1595 if (strcmp(op, "GET") == 0) 1596 return (ftp_request(url, "RETR", us, purl, flags)); 1597 else if (strcmp(op, "HEAD") == 0) 1598 return (ftp_request(url, "STAT", us, purl, flags)); 1599 } 1600 1601 /* connect to server or proxy */ 1602 if ((conn = http_connect(url, purl, flags)) == NULL) 1603 goto ouch; 1604 1605 host = url->host; 1606 #ifdef INET6 1607 if (strchr(url->host, ':')) { 1608 snprintf(hbuf, sizeof(hbuf), "[%s]", url->host); 1609 host = hbuf; 1610 } 1611 #endif 1612 if (url->port != fetch_default_port(url->scheme)) { 1613 if (host != hbuf) { 1614 strcpy(hbuf, host); 1615 host = hbuf; 1616 } 1617 snprintf(hbuf + strlen(hbuf), 1618 sizeof(hbuf) - strlen(hbuf), ":%d", url->port); 1619 } 1620 1621 /* send request */ 1622 if (verbose) 1623 fetch_info("requesting %s://%s%s", 1624 url->scheme, host, url->doc); 1625 if (purl && strcasecmp(URL->scheme, SCHEME_HTTPS) != 0) { 1626 http_cmd(conn, "%s %s://%s%s HTTP/1.1", 1627 op, url->scheme, host, url->doc); 1628 } else { 1629 http_cmd(conn, "%s %s HTTP/1.1", 1630 op, url->doc); 1631 } 1632 1633 if (ims && url->ims_time) { 1634 timestruct = gmtime((time_t *)&url->ims_time); 1635 (void)strftime(timebuf, 80, "%a, %d %b %Y %T GMT", 1636 timestruct); 1637 if (verbose) 1638 fetch_info("If-Modified-Since: %s", timebuf); 1639 http_cmd(conn, "If-Modified-Since: %s", timebuf); 1640 } 1641 /* virtual host */ 1642 http_cmd(conn, "Host: %s", host); 1643 1644 /* 1645 * Proxy authorization: we only send auth after we received 1646 * a 407 error. We do not first try basic anyway (changed 1647 * when support was added for digest-auth) 1648 */ 1649 if (purl && proxy_challenges.valid) { 1650 http_auth_params_t aparams; 1651 init_http_auth_params(&aparams); 1652 if (*purl->user || *purl->pwd) { 1653 aparams.user = strdup(purl->user); 1654 aparams.password = strdup(purl->pwd); 1655 } else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL && 1656 *p != '\0') { 1657 if (http_authfromenv(p, &aparams) < 0) { 1658 http_seterr(HTTP_NEED_PROXY_AUTH); 1659 goto ouch; 1660 } 1661 } 1662 http_authorize(conn, "Proxy-Authorization", 1663 &proxy_challenges, &aparams, url); 1664 clean_http_auth_params(&aparams); 1665 } 1666 1667 /* 1668 * Server authorization: we never send "a priori" 1669 * Basic auth, which used to be done if user/pass were 1670 * set in the url. This would be weird because we'd send the 1671 * password in the clear even if Digest is finally to be 1672 * used (it would have made more sense for the 1673 * pre-digest version to do this when Basic was specified 1674 * in the environment) 1675 */ 1676 if (server_challenges.valid) { 1677 http_auth_params_t aparams; 1678 init_http_auth_params(&aparams); 1679 if (*url->user || *url->pwd) { 1680 aparams.user = strdup(url->user); 1681 aparams.password = strdup(url->pwd); 1682 } else if ((p = getenv("HTTP_AUTH")) != NULL && 1683 *p != '\0') { 1684 if (http_authfromenv(p, &aparams) < 0) { 1685 http_seterr(HTTP_NEED_AUTH); 1686 goto ouch; 1687 } 1688 } else if (fetchAuthMethod && 1689 fetchAuthMethod(url) == 0) { 1690 aparams.user = strdup(url->user); 1691 aparams.password = strdup(url->pwd); 1692 } else { 1693 http_seterr(HTTP_NEED_AUTH); 1694 goto ouch; 1695 } 1696 http_authorize(conn, "Authorization", 1697 &server_challenges, &aparams, url); 1698 clean_http_auth_params(&aparams); 1699 } 1700 1701 /* other headers */ 1702 if ((p = getenv("HTTP_ACCEPT")) != NULL) { 1703 if (*p != '\0') 1704 http_cmd(conn, "Accept: %s", p); 1705 } else { 1706 http_cmd(conn, "Accept: */*"); 1707 } 1708 if ((p = getenv("HTTP_REFERER")) != NULL && *p != '\0') { 1709 if (strcasecmp(p, "auto") == 0) 1710 http_cmd(conn, "Referer: %s://%s%s", 1711 url->scheme, host, url->doc); 1712 else 1713 http_cmd(conn, "Referer: %s", p); 1714 } 1715 if ((p = getenv("HTTP_USER_AGENT")) != NULL) { 1716 /* no User-Agent if defined but empty */ 1717 if (*p != '\0') 1718 http_cmd(conn, "User-Agent: %s", p); 1719 } else { 1720 /* default User-Agent */ 1721 http_cmd(conn, "User-Agent: %s " _LIBFETCH_VER, 1722 getprogname()); 1723 } 1724 if (url->offset > 0) 1725 http_cmd(conn, "Range: bytes=%lld-", (long long)url->offset); 1726 http_cmd(conn, "Connection: close"); 1727 1728 if (body) { 1729 body_len = strlen(body); 1730 http_cmd(conn, "Content-Length: %zu", body_len); 1731 if (content_type != NULL) 1732 http_cmd(conn, "Content-Type: %s", content_type); 1733 } 1734 1735 http_cmd(conn, ""); 1736 1737 if (body) 1738 fetch_write(conn, body, body_len); 1739 1740 /* 1741 * Force the queued request to be dispatched. Normally, one 1742 * would do this with shutdown(2) but squid proxies can be 1743 * configured to disallow such half-closed connections. To 1744 * be compatible with such configurations, fiddle with socket 1745 * options to force the pending data to be written. 1746 */ 1747 val = 0; 1748 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, 1749 sizeof(val)); 1750 val = 1; 1751 setsockopt(conn->sd, IPPROTO_TCP, TCP_NODELAY, &val, 1752 sizeof(val)); 1753 1754 /* get reply */ 1755 switch (http_get_reply(conn)) { 1756 case HTTP_OK: 1757 case HTTP_PARTIAL: 1758 case HTTP_NOT_MODIFIED: 1759 /* fine */ 1760 break; 1761 case HTTP_MOVED_PERM: 1762 case HTTP_MOVED_TEMP: 1763 case HTTP_SEE_OTHER: 1764 case HTTP_USE_PROXY: 1765 /* 1766 * Not so fine, but we still have to read the 1767 * headers to get the new location. 1768 */ 1769 break; 1770 case HTTP_NEED_AUTH: 1771 if (server_challenges.valid) { 1772 /* 1773 * We already sent out authorization code, 1774 * so there's nothing more we can do. 1775 */ 1776 http_seterr(conn->err); 1777 goto ouch; 1778 } 1779 /* try again, but send the password this time */ 1780 if (verbose) 1781 fetch_info("server requires authorization"); 1782 break; 1783 case HTTP_NEED_PROXY_AUTH: 1784 if (proxy_challenges.valid) { 1785 /* 1786 * We already sent our proxy 1787 * authorization code, so there's 1788 * nothing more we can do. */ 1789 http_seterr(conn->err); 1790 goto ouch; 1791 } 1792 /* try again, but send the password this time */ 1793 if (verbose) 1794 fetch_info("proxy requires authorization"); 1795 break; 1796 case HTTP_BAD_RANGE: 1797 /* 1798 * This can happen if we ask for 0 bytes because 1799 * we already have the whole file. Consider this 1800 * a success for now, and check sizes later. 1801 */ 1802 break; 1803 case HTTP_PROTOCOL_ERROR: 1804 /* fall through */ 1805 case -1: 1806 fetch_syserr(); 1807 goto ouch; 1808 default: 1809 http_seterr(conn->err); 1810 if (!verbose) 1811 goto ouch; 1812 /* fall through so we can get the full error message */ 1813 } 1814 1815 /* get headers. http_next_header expects one line readahead */ 1816 if (fetch_getln(conn) == -1) { 1817 fetch_syserr(); 1818 goto ouch; 1819 } 1820 do { 1821 switch ((h = http_next_header(conn, &headerbuf, &p))) { 1822 case hdr_syserror: 1823 fetch_syserr(); 1824 goto ouch; 1825 case hdr_error: 1826 http_seterr(HTTP_PROTOCOL_ERROR); 1827 goto ouch; 1828 case hdr_content_length: 1829 http_parse_length(p, &clength); 1830 break; 1831 case hdr_content_range: 1832 http_parse_range(p, &offset, &length, &size); 1833 break; 1834 case hdr_last_modified: 1835 http_parse_mtime(p, &mtime); 1836 break; 1837 case hdr_location: 1838 if (!HTTP_REDIRECT(conn->err)) 1839 break; 1840 /* 1841 * if the A flag is set, we don't follow 1842 * temporary redirects. 1843 */ 1844 if (noredirect && 1845 conn->err != HTTP_MOVED_PERM && 1846 conn->err != HTTP_PERM_REDIRECT && 1847 conn->err != HTTP_USE_PROXY) { 1848 n = 1; 1849 break; 1850 } 1851 if (new) 1852 free(new); 1853 if (verbose) 1854 fetch_info("%d redirect to %s", conn->err, p); 1855 if (*p == '/') 1856 /* absolute path */ 1857 new = fetchMakeURL(url->scheme, url->host, url->port, p, 1858 url->user, url->pwd); 1859 else 1860 new = fetchParseURL(p); 1861 if (new == NULL) { 1862 /* XXX should set an error code */ 1863 DEBUG(fprintf(stderr, "failed to parse new URL\n")); 1864 goto ouch; 1865 } 1866 1867 /* Only copy credentials if the host matches */ 1868 if (!strcmp(new->host, url->host) && !*new->user && !*new->pwd) { 1869 strcpy(new->user, url->user); 1870 strcpy(new->pwd, url->pwd); 1871 } 1872 new->offset = url->offset; 1873 new->length = url->length; 1874 break; 1875 case hdr_transfer_encoding: 1876 /* XXX weak test*/ 1877 chunked = (strcasecmp(p, "chunked") == 0); 1878 break; 1879 case hdr_www_authenticate: 1880 if (conn->err != HTTP_NEED_AUTH) 1881 break; 1882 if (http_parse_authenticate(p, &server_challenges) == 0) 1883 ++n; 1884 break; 1885 case hdr_proxy_authenticate: 1886 if (conn->err != HTTP_NEED_PROXY_AUTH) 1887 break; 1888 if (http_parse_authenticate(p, &proxy_challenges) == 0) 1889 ++n; 1890 break; 1891 case hdr_end: 1892 /* fall through */ 1893 case hdr_unknown: 1894 /* ignore */ 1895 break; 1896 } 1897 } while (h > hdr_end); 1898 1899 /* we need to provide authentication */ 1900 if (conn->err == HTTP_NEED_AUTH || 1901 conn->err == HTTP_NEED_PROXY_AUTH) { 1902 e = conn->err; 1903 if ((conn->err == HTTP_NEED_AUTH && 1904 !server_challenges.valid) || 1905 (conn->err == HTTP_NEED_PROXY_AUTH && 1906 !proxy_challenges.valid)) { 1907 /* 401/7 but no www/proxy-authenticate ?? */ 1908 DEBUG(fprintf(stderr, "401/7 and no auth header\n")); 1909 goto ouch; 1910 } 1911 fetch_close(conn); 1912 conn = NULL; 1913 continue; 1914 } 1915 1916 /* requested range not satisfiable */ 1917 if (conn->err == HTTP_BAD_RANGE) { 1918 if (url->offset == size && url->length == 0) { 1919 /* asked for 0 bytes; fake it */ 1920 offset = url->offset; 1921 clength = -1; 1922 conn->err = HTTP_OK; 1923 break; 1924 } else { 1925 http_seterr(conn->err); 1926 goto ouch; 1927 } 1928 } 1929 1930 /* we have a hit or an error */ 1931 if (conn->err == HTTP_OK 1932 || conn->err == HTTP_NOT_MODIFIED 1933 || conn->err == HTTP_PARTIAL 1934 || HTTP_ERROR(conn->err)) 1935 break; 1936 1937 /* all other cases: we got a redirect */ 1938 e = conn->err; 1939 clean_http_auth_challenges(&server_challenges); 1940 fetch_close(conn); 1941 conn = NULL; 1942 if (!new) { 1943 DEBUG(fprintf(stderr, "redirect with no new location\n")); 1944 break; 1945 } 1946 if (url != URL) 1947 fetchFreeURL(url); 1948 url = new; 1949 } while (++i < n); 1950 1951 /* we failed, or ran out of retries */ 1952 if (conn == NULL) { 1953 http_seterr(e); 1954 goto ouch; 1955 } 1956 1957 DEBUG(fprintf(stderr, "offset %lld, length %lld," 1958 " size %lld, clength %lld\n", 1959 (long long)offset, (long long)length, 1960 (long long)size, (long long)clength)); 1961 1962 if (conn->err == HTTP_NOT_MODIFIED) { 1963 http_seterr(HTTP_NOT_MODIFIED); 1964 return (NULL); 1965 } 1966 1967 /* check for inconsistencies */ 1968 if (clength != -1 && length != -1 && clength != length) { 1969 http_seterr(HTTP_PROTOCOL_ERROR); 1970 goto ouch; 1971 } 1972 if (clength == -1) 1973 clength = length; 1974 if (clength != -1) 1975 length = offset + clength; 1976 if (length != -1 && size != -1 && length != size) { 1977 http_seterr(HTTP_PROTOCOL_ERROR); 1978 goto ouch; 1979 } 1980 if (size == -1) 1981 size = length; 1982 1983 /* fill in stats */ 1984 if (us) { 1985 us->size = size; 1986 us->atime = us->mtime = mtime; 1987 } 1988 1989 /* too far? */ 1990 if (URL->offset > 0 && offset > URL->offset) { 1991 http_seterr(HTTP_PROTOCOL_ERROR); 1992 goto ouch; 1993 } 1994 1995 /* report back real offset and size */ 1996 URL->offset = offset; 1997 URL->length = clength; 1998 1999 /* wrap it up in a FILE */ 2000 if ((f = http_funopen(conn, chunked)) == NULL) { 2001 fetch_syserr(); 2002 goto ouch; 2003 } 2004 2005 if (url != URL) 2006 fetchFreeURL(url); 2007 if (purl) 2008 fetchFreeURL(purl); 2009 2010 if (HTTP_ERROR(conn->err)) { 2011 http_print_html(stderr, f); 2012 fclose(f); 2013 f = NULL; 2014 } 2015 clean_http_headerbuf(&headerbuf); 2016 clean_http_auth_challenges(&server_challenges); 2017 clean_http_auth_challenges(&proxy_challenges); 2018 return (f); 2019 2020 ouch: 2021 if (url != URL) 2022 fetchFreeURL(url); 2023 if (purl) 2024 fetchFreeURL(purl); 2025 if (conn != NULL) 2026 fetch_close(conn); 2027 clean_http_headerbuf(&headerbuf); 2028 clean_http_auth_challenges(&server_challenges); 2029 clean_http_auth_challenges(&proxy_challenges); 2030 return (NULL); 2031 } 2032 2033 2034 /***************************************************************************** 2035 * Entry points 2036 */ 2037 2038 /* 2039 * Retrieve and stat a file by HTTP 2040 */ 2041 FILE * 2042 fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags) 2043 { 2044 return (http_request(URL, "GET", us, http_get_proxy(URL, flags), flags)); 2045 } 2046 2047 /* 2048 * Retrieve a file by HTTP 2049 */ 2050 FILE * 2051 fetchGetHTTP(struct url *URL, const char *flags) 2052 { 2053 return (fetchXGetHTTP(URL, NULL, flags)); 2054 } 2055 2056 /* 2057 * Store a file by HTTP 2058 */ 2059 FILE * 2060 fetchPutHTTP(struct url *URL __unused, const char *flags __unused) 2061 { 2062 warnx("fetchPutHTTP(): not implemented"); 2063 return (NULL); 2064 } 2065 2066 /* 2067 * Get an HTTP document's metadata 2068 */ 2069 int 2070 fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags) 2071 { 2072 FILE *f; 2073 2074 f = http_request(URL, "HEAD", us, http_get_proxy(URL, flags), flags); 2075 if (f == NULL) 2076 return (-1); 2077 fclose(f); 2078 return (0); 2079 } 2080 2081 /* 2082 * List a directory 2083 */ 2084 struct url_ent * 2085 fetchListHTTP(struct url *url __unused, const char *flags __unused) 2086 { 2087 warnx("fetchListHTTP(): not implemented"); 2088 return (NULL); 2089 } 2090 2091 FILE * 2092 fetchReqHTTP(struct url *URL, const char *method, const char *flags, 2093 const char *content_type, const char *body) 2094 { 2095 2096 return (http_request_body(URL, method, NULL, http_get_proxy(URL, flags), 2097 flags, content_type, body)); 2098 } 2099