1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 2000-2014 Dag-Erling Smørgrav 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer 12 * in this position and unchanged. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. The name of the author may not be used to endorse or promote products 17 * derived from this software without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 20 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 21 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 22 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, 23 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 24 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 25 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 26 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 27 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF 28 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 29 * 30 * $FreeBSD: head/lib/libfetch/http.c 341072 2018-11-27 16:23:17Z des $ 31 */ 32 33 /* 34 * The following copyright applies to the base64 code: 35 * 36 *- 37 * Copyright 1997 Massachusetts Institute of Technology 38 * 39 * Permission to use, copy, modify, and distribute this software and 40 * its documentation for any purpose and without fee is hereby 41 * granted, provided that both the above copyright notice and this 42 * permission notice appear in all copies, that both the above 43 * copyright notice and this permission notice appear in all 44 * supporting documentation, and that the name of M.I.T. not be used 45 * in advertising or publicity pertaining to distribution of the 46 * software without specific, written prior permission. M.I.T. makes 47 * no representations about the suitability of this software for any 48 * purpose. It is provided "as is" without express or implied 49 * warranty. 50 * 51 * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''. M.I.T. DISCLAIMS 52 * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE, 53 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 54 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT 55 * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 56 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 57 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF 58 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND 59 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 60 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 61 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 62 * SUCH DAMAGE. 63 */ 64 65 #include <sys/param.h> 66 #include <sys/socket.h> 67 #include <sys/time.h> 68 69 #include <ctype.h> 70 #include <err.h> 71 #include <errno.h> 72 #include <locale.h> 73 #include <netdb.h> 74 #include <stdarg.h> 75 #include <stdio.h> 76 #include <stdlib.h> 77 #include <string.h> 78 #include <time.h> 79 #include <unistd.h> 80 81 #ifdef WITH_SSL 82 #include <openssl/md5.h> 83 #define MD5Init(c) MD5_Init(c) 84 #define MD5Update(c, data, len) MD5_Update(c, data, len) 85 #define MD5Final(md, c) MD5_Final(md, c) 86 #else 87 #include <md5.h> 88 #endif 89 90 #include <netinet/in.h> 91 #include <netinet/tcp.h> 92 93 #include "fetch.h" 94 #include "common.h" 95 #include "httperr.h" 96 97 /* Maximum number of redirects to follow */ 98 #define MAX_REDIRECT 20 99 100 /* Symbolic names for reply codes we care about */ 101 #define HTTP_OK 200 102 #define HTTP_PARTIAL 206 103 #define HTTP_MOVED_PERM 301 104 #define HTTP_MOVED_TEMP 302 105 #define HTTP_SEE_OTHER 303 106 #define HTTP_NOT_MODIFIED 304 107 #define HTTP_USE_PROXY 305 108 #define HTTP_TEMP_REDIRECT 307 109 #define HTTP_PERM_REDIRECT 308 110 #define HTTP_NEED_AUTH 401 111 #define HTTP_NEED_PROXY_AUTH 407 112 #define HTTP_BAD_RANGE 416 113 #define HTTP_PROTOCOL_ERROR 999 114 115 #define HTTP_REDIRECT(xyz) ((xyz) == HTTP_MOVED_PERM \ 116 || (xyz) == HTTP_MOVED_TEMP \ 117 || (xyz) == HTTP_TEMP_REDIRECT \ 118 || (xyz) == HTTP_PERM_REDIRECT \ 119 || (xyz) == HTTP_USE_PROXY \ 120 || (xyz) == HTTP_SEE_OTHER) 121 122 #define HTTP_ERROR(xyz) ((xyz) >= 400 && (xyz) <= 599) 123 124 125 /***************************************************************************** 126 * I/O functions for decoding chunked streams 127 */ 128 129 struct httpio 130 { 131 conn_t *conn; /* connection */ 132 int chunked; /* chunked mode */ 133 char *buf; /* chunk buffer */ 134 size_t bufsize; /* size of chunk buffer */ 135 size_t buflen; /* amount of data currently in buffer */ 136 size_t bufpos; /* current read offset in buffer */ 137 int eof; /* end-of-file flag */ 138 int error; /* error flag */ 139 size_t chunksize; /* remaining size of current chunk */ 140 #ifndef NDEBUG 141 size_t total; 142 #endif 143 }; 144 145 /* 146 * Get next chunk header 147 */ 148 static int 149 http_new_chunk(struct httpio *io) 150 { 151 char *p; 152 153 if (fetch_getln(io->conn) == -1) 154 return (-1); 155 156 if (io->conn->buflen < 2 || !isxdigit((unsigned char)*io->conn->buf)) 157 return (-1); 158 159 for (p = io->conn->buf; *p && !isspace((unsigned char)*p); ++p) { 160 if (*p == ';') 161 break; 162 if (!isxdigit((unsigned char)*p)) 163 return (-1); 164 if (isdigit((unsigned char)*p)) { 165 io->chunksize = io->chunksize * 16 + 166 *p - '0'; 167 } else { 168 io->chunksize = io->chunksize * 16 + 169 10 + tolower((unsigned char)*p) - 'a'; 170 } 171 } 172 173 #ifndef NDEBUG 174 if (fetchDebug) { 175 io->total += io->chunksize; 176 if (io->chunksize == 0) 177 fprintf(stderr, "%s(): end of last chunk\n", __func__); 178 else 179 fprintf(stderr, "%s(): new chunk: %lu (%lu)\n", 180 __func__, (unsigned long)io->chunksize, 181 (unsigned long)io->total); 182 } 183 #endif 184 185 return (io->chunksize); 186 } 187 188 /* 189 * Grow the input buffer to at least len bytes 190 */ 191 static inline int 192 http_growbuf(struct httpio *io, size_t len) 193 { 194 char *tmp; 195 196 if (io->bufsize >= len) 197 return (0); 198 199 if ((tmp = realloc(io->buf, len)) == NULL) 200 return (-1); 201 io->buf = tmp; 202 io->bufsize = len; 203 return (0); 204 } 205 206 /* 207 * Fill the input buffer, do chunk decoding on the fly 208 */ 209 static ssize_t 210 http_fillbuf(struct httpio *io, size_t len) 211 { 212 ssize_t nbytes; 213 char ch; 214 215 if (io->error) 216 return (-1); 217 if (io->eof) 218 return (0); 219 220 /* not chunked: just fetch the requested amount */ 221 if (io->chunked == 0) { 222 if (http_growbuf(io, len) == -1) 223 return (-1); 224 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) { 225 io->error = errno; 226 return (-1); 227 } 228 io->buflen = nbytes; 229 io->bufpos = 0; 230 return (io->buflen); 231 } 232 233 /* chunked, but we ran out: get the next chunk header */ 234 if (io->chunksize == 0) { 235 switch (http_new_chunk(io)) { 236 case -1: 237 io->error = EPROTO; 238 return (-1); 239 case 0: 240 io->eof = 1; 241 return (0); 242 } 243 } 244 245 /* fetch the requested amount, but no more than the current chunk */ 246 if (len > io->chunksize) 247 len = io->chunksize; 248 if (http_growbuf(io, len) == -1) 249 return (-1); 250 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) { 251 io->error = errno; 252 return (-1); 253 } 254 io->bufpos = 0; 255 io->buflen = nbytes; 256 io->chunksize -= nbytes; 257 258 if (io->chunksize == 0) { 259 if (fetch_read(io->conn, &ch, 1) != 1 || ch != '\r' || 260 fetch_read(io->conn, &ch, 1) != 1 || ch != '\n') 261 return (-1); 262 } 263 264 return (io->buflen); 265 } 266 267 /* 268 * Read function 269 */ 270 static int 271 http_readfn(void *v, char *buf, int len) 272 { 273 struct httpio *io = (struct httpio *)v; 274 int rlen; 275 276 if (io->error) 277 return (-1); 278 if (io->eof) 279 return (0); 280 281 /* empty buffer */ 282 if (!io->buf || io->bufpos == io->buflen) { 283 if ((rlen = http_fillbuf(io, len)) < 0) { 284 if ((errno = io->error) == EINTR) 285 io->error = 0; 286 return (-1); 287 } else if (rlen == 0) { 288 return (0); 289 } 290 } 291 292 rlen = io->buflen - io->bufpos; 293 if (len < rlen) 294 rlen = len; 295 memcpy(buf, io->buf + io->bufpos, rlen); 296 io->bufpos += rlen; 297 return (rlen); 298 } 299 300 /* 301 * Write function 302 */ 303 static int 304 http_writefn(void *v, const char *buf, int len) 305 { 306 struct httpio *io = (struct httpio *)v; 307 308 return (fetch_write(io->conn, buf, len)); 309 } 310 311 /* 312 * Close function 313 */ 314 static int 315 http_closefn(void *v) 316 { 317 struct httpio *io = (struct httpio *)v; 318 int r; 319 320 r = fetch_close(io->conn); 321 if (io->buf) 322 free(io->buf); 323 free(io); 324 return (r); 325 } 326 327 /* 328 * Wrap a file descriptor up 329 */ 330 static FILE * 331 http_funopen(conn_t *conn, int chunked) 332 { 333 struct httpio *io; 334 FILE *f; 335 336 if ((io = calloc(1, sizeof(*io))) == NULL) { 337 fetch_syserr(); 338 return (NULL); 339 } 340 io->conn = conn; 341 io->chunked = chunked; 342 f = funopen(io, http_readfn, http_writefn, NULL, http_closefn); 343 if (f == NULL) { 344 fetch_syserr(); 345 free(io); 346 return (NULL); 347 } 348 return (f); 349 } 350 351 352 /***************************************************************************** 353 * Helper functions for talking to the server and parsing its replies 354 */ 355 356 /* Header types */ 357 typedef enum { 358 hdr_syserror = -2, 359 hdr_error = -1, 360 hdr_end = 0, 361 hdr_unknown = 1, 362 hdr_content_length, 363 hdr_content_range, 364 hdr_last_modified, 365 hdr_location, 366 hdr_transfer_encoding, 367 hdr_www_authenticate, 368 hdr_proxy_authenticate, 369 } hdr_t; 370 371 /* Names of interesting headers */ 372 static struct { 373 hdr_t num; 374 const char *name; 375 } hdr_names[] = { 376 { hdr_content_length, "Content-Length" }, 377 { hdr_content_range, "Content-Range" }, 378 { hdr_last_modified, "Last-Modified" }, 379 { hdr_location, "Location" }, 380 { hdr_transfer_encoding, "Transfer-Encoding" }, 381 { hdr_www_authenticate, "WWW-Authenticate" }, 382 { hdr_proxy_authenticate, "Proxy-Authenticate" }, 383 { hdr_unknown, NULL }, 384 }; 385 386 /* 387 * Send a formatted line; optionally echo to terminal 388 */ 389 static int 390 http_cmd(conn_t *conn, const char *fmt, ...) 391 { 392 va_list ap; 393 size_t len; 394 char *msg; 395 int r; 396 397 va_start(ap, fmt); 398 len = vasprintf(&msg, fmt, ap); 399 va_end(ap); 400 401 if (msg == NULL) { 402 errno = ENOMEM; 403 fetch_syserr(); 404 return (-1); 405 } 406 407 r = fetch_putln(conn, msg, len); 408 free(msg); 409 410 if (r == -1) { 411 fetch_syserr(); 412 return (-1); 413 } 414 415 return (0); 416 } 417 418 /* 419 * Get and parse status line 420 */ 421 static int 422 http_get_reply(conn_t *conn) 423 { 424 char *p; 425 426 if (fetch_getln(conn) == -1) 427 return (-1); 428 /* 429 * A valid status line looks like "HTTP/m.n xyz reason" where m 430 * and n are the major and minor protocol version numbers and xyz 431 * is the reply code. 432 * Unfortunately, there are servers out there (NCSA 1.5.1, to name 433 * just one) that do not send a version number, so we can't rely 434 * on finding one, but if we do, insist on it being 1.0 or 1.1. 435 * We don't care about the reason phrase. 436 */ 437 if (strncmp(conn->buf, "HTTP", 4) != 0) 438 return (HTTP_PROTOCOL_ERROR); 439 p = conn->buf + 4; 440 if (*p == '/') { 441 if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1')) 442 return (HTTP_PROTOCOL_ERROR); 443 p += 4; 444 } 445 if (*p != ' ' || 446 !isdigit((unsigned char)p[1]) || 447 !isdigit((unsigned char)p[2]) || 448 !isdigit((unsigned char)p[3])) 449 return (HTTP_PROTOCOL_ERROR); 450 451 conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0'); 452 return (conn->err); 453 } 454 455 /* 456 * Check a header; if the type matches the given string, return a pointer 457 * to the beginning of the value. 458 */ 459 static const char * 460 http_match(const char *str, const char *hdr) 461 { 462 while (*str && *hdr && 463 tolower((unsigned char)*str++) == tolower((unsigned char)*hdr++)) 464 /* nothing */; 465 if (*str || *hdr != ':') 466 return (NULL); 467 while (*hdr && isspace((unsigned char)*++hdr)) 468 /* nothing */; 469 return (hdr); 470 } 471 472 473 /* 474 * Get the next header and return the appropriate symbolic code. We 475 * need to read one line ahead for checking for a continuation line 476 * belonging to the current header (continuation lines start with 477 * white space). 478 * 479 * We get called with a fresh line already in the conn buffer, either 480 * from the previous http_next_header() invocation, or, the first 481 * time, from a fetch_getln() performed by our caller. 482 * 483 * This stops when we encounter an empty line (we dont read beyond the header 484 * area). 485 * 486 * Note that the "headerbuf" is just a place to return the result. Its 487 * contents are not used for the next call. This means that no cleanup 488 * is needed when ie doing another connection, just call the cleanup when 489 * fully done to deallocate memory. 490 */ 491 492 /* Limit the max number of continuation lines to some reasonable value */ 493 #define HTTP_MAX_CONT_LINES 10 494 495 /* Place into which to build a header from one or several lines */ 496 typedef struct { 497 char *buf; /* buffer */ 498 size_t bufsize; /* buffer size */ 499 size_t buflen; /* length of buffer contents */ 500 } http_headerbuf_t; 501 502 static void 503 init_http_headerbuf(http_headerbuf_t *buf) 504 { 505 buf->buf = NULL; 506 buf->bufsize = 0; 507 buf->buflen = 0; 508 } 509 510 static void 511 clean_http_headerbuf(http_headerbuf_t *buf) 512 { 513 if (buf->buf) 514 free(buf->buf); 515 init_http_headerbuf(buf); 516 } 517 518 /* Remove whitespace at the end of the buffer */ 519 static void 520 http_conn_trimright(conn_t *conn) 521 { 522 while (conn->buflen && 523 isspace((unsigned char)conn->buf[conn->buflen - 1])) 524 conn->buflen--; 525 conn->buf[conn->buflen] = '\0'; 526 } 527 528 static hdr_t 529 http_next_header(conn_t *conn, http_headerbuf_t *hbuf, const char **p) 530 { 531 unsigned int i, len; 532 533 /* 534 * Have to do the stripping here because of the first line. So 535 * it's done twice for the subsequent lines. No big deal 536 */ 537 http_conn_trimright(conn); 538 if (conn->buflen == 0) 539 return (hdr_end); 540 541 /* Copy the line to the headerbuf */ 542 if (hbuf->bufsize < conn->buflen + 1) { 543 if ((hbuf->buf = realloc(hbuf->buf, conn->buflen + 1)) == NULL) 544 return (hdr_syserror); 545 hbuf->bufsize = conn->buflen + 1; 546 } 547 strcpy(hbuf->buf, conn->buf); 548 hbuf->buflen = conn->buflen; 549 550 /* 551 * Fetch possible continuation lines. Stop at 1st non-continuation 552 * and leave it in the conn buffer 553 */ 554 for (i = 0; i < HTTP_MAX_CONT_LINES; i++) { 555 if (fetch_getln(conn) == -1) 556 return (hdr_syserror); 557 558 /* 559 * Note: we carry on the idea from the previous version 560 * that a pure whitespace line is equivalent to an empty 561 * one (so it's not continuation and will be handled when 562 * we are called next) 563 */ 564 http_conn_trimright(conn); 565 if (conn->buf[0] != ' ' && conn->buf[0] != "\t"[0]) 566 break; 567 568 /* Got a continuation line. Concatenate to previous */ 569 len = hbuf->buflen + conn->buflen; 570 if (hbuf->bufsize < len + 1) { 571 len *= 2; 572 if ((hbuf->buf = realloc(hbuf->buf, len + 1)) == NULL) 573 return (hdr_syserror); 574 hbuf->bufsize = len + 1; 575 } 576 strcpy(hbuf->buf + hbuf->buflen, conn->buf); 577 hbuf->buflen += conn->buflen; 578 } 579 580 /* 581 * We could check for malformed headers but we don't really care. 582 * A valid header starts with a token immediately followed by a 583 * colon; a token is any sequence of non-control, non-whitespace 584 * characters except "()<>@,;:\\\"{}". 585 */ 586 for (i = 0; hdr_names[i].num != hdr_unknown; i++) 587 if ((*p = http_match(hdr_names[i].name, hbuf->buf)) != NULL) 588 return (hdr_names[i].num); 589 590 return (hdr_unknown); 591 } 592 593 /************************** 594 * [Proxy-]Authenticate header parsing 595 */ 596 597 /* 598 * Read doublequote-delimited string into output buffer obuf (allocated 599 * by caller, whose responsibility it is to ensure that it's big enough) 600 * cp points to the first char after the initial '"' 601 * Handles \ quoting 602 * Returns pointer to the first char after the terminating double quote, or 603 * NULL for error. 604 */ 605 static const char * 606 http_parse_headerstring(const char *cp, char *obuf) 607 { 608 for (;;) { 609 switch (*cp) { 610 case 0: /* Unterminated string */ 611 *obuf = 0; 612 return (NULL); 613 case '"': /* Ending quote */ 614 *obuf = 0; 615 return (++cp); 616 case '\\': 617 if (*++cp == 0) { 618 *obuf = 0; 619 return (NULL); 620 } 621 /* FALLTHROUGH */ 622 default: 623 *obuf++ = *cp++; 624 } 625 } 626 } 627 628 /* Http auth challenge schemes */ 629 typedef enum {HTTPAS_UNKNOWN, HTTPAS_BASIC,HTTPAS_DIGEST} http_auth_schemes_t; 630 631 /* Data holder for a Basic or Digest challenge. */ 632 typedef struct { 633 http_auth_schemes_t scheme; 634 char *realm; 635 char *qop; 636 char *nonce; 637 char *opaque; 638 char *algo; 639 int stale; 640 int nc; /* Nonce count */ 641 } http_auth_challenge_t; 642 643 static void 644 init_http_auth_challenge(http_auth_challenge_t *b) 645 { 646 b->scheme = HTTPAS_UNKNOWN; 647 b->realm = b->qop = b->nonce = b->opaque = b->algo = NULL; 648 b->stale = b->nc = 0; 649 } 650 651 static void 652 clean_http_auth_challenge(http_auth_challenge_t *b) 653 { 654 if (b->realm) 655 free(b->realm); 656 if (b->qop) 657 free(b->qop); 658 if (b->nonce) 659 free(b->nonce); 660 if (b->opaque) 661 free(b->opaque); 662 if (b->algo) 663 free(b->algo); 664 init_http_auth_challenge(b); 665 } 666 667 /* Data holder for an array of challenges offered in an http response. */ 668 #define MAX_CHALLENGES 10 669 typedef struct { 670 http_auth_challenge_t *challenges[MAX_CHALLENGES]; 671 int count; /* Number of parsed challenges in the array */ 672 int valid; /* We did parse an authenticate header */ 673 } http_auth_challenges_t; 674 675 static void 676 init_http_auth_challenges(http_auth_challenges_t *cs) 677 { 678 int i; 679 for (i = 0; i < MAX_CHALLENGES; i++) 680 cs->challenges[i] = NULL; 681 cs->count = cs->valid = 0; 682 } 683 684 static void 685 clean_http_auth_challenges(http_auth_challenges_t *cs) 686 { 687 int i; 688 /* We rely on non-zero pointers being allocated, not on the count */ 689 for (i = 0; i < MAX_CHALLENGES; i++) { 690 if (cs->challenges[i] != NULL) { 691 clean_http_auth_challenge(cs->challenges[i]); 692 free(cs->challenges[i]); 693 } 694 } 695 init_http_auth_challenges(cs); 696 } 697 698 /* 699 * Enumeration for lexical elements. Separators will be returned as their own 700 * ascii value 701 */ 702 typedef enum {HTTPHL_WORD=256, HTTPHL_STRING=257, HTTPHL_END=258, 703 HTTPHL_ERROR = 259} http_header_lex_t; 704 705 /* 706 * Determine what kind of token comes next and return possible value 707 * in buf, which is supposed to have been allocated big enough by 708 * caller. Advance input pointer and return element type. 709 */ 710 static int 711 http_header_lex(const char **cpp, char *buf) 712 { 713 size_t l; 714 /* Eat initial whitespace */ 715 *cpp += strspn(*cpp, " \t"); 716 if (**cpp == 0) 717 return (HTTPHL_END); 718 719 /* Separator ? */ 720 if (**cpp == ',' || **cpp == '=') 721 return (*((*cpp)++)); 722 723 /* String ? */ 724 if (**cpp == '"') { 725 *cpp = http_parse_headerstring(++*cpp, buf); 726 if (*cpp == NULL) 727 return (HTTPHL_ERROR); 728 return (HTTPHL_STRING); 729 } 730 731 /* Read other token, until separator or whitespace */ 732 l = strcspn(*cpp, " \t,="); 733 memcpy(buf, *cpp, l); 734 buf[l] = 0; 735 *cpp += l; 736 return (HTTPHL_WORD); 737 } 738 739 /* 740 * Read challenges from http xxx-authenticate header and accumulate them 741 * in the challenges list structure. 742 * 743 * Headers with multiple challenges are specified by rfc2617, but 744 * servers (ie: squid) often send them in separate headers instead, 745 * which in turn is forbidden by the http spec (multiple headers with 746 * the same name are only allowed for pure comma-separated lists, see 747 * rfc2616 sec 4.2). 748 * 749 * We support both approaches anyway 750 */ 751 static int 752 http_parse_authenticate(const char *cp, http_auth_challenges_t *cs) 753 { 754 int ret = -1; 755 http_header_lex_t lex; 756 char *key = malloc(strlen(cp) + 1); 757 char *value = malloc(strlen(cp) + 1); 758 char *buf = malloc(strlen(cp) + 1); 759 760 if (key == NULL || value == NULL || buf == NULL) { 761 fetch_syserr(); 762 goto out; 763 } 764 765 /* In any case we've seen the header and we set the valid bit */ 766 cs->valid = 1; 767 768 /* Need word first */ 769 lex = http_header_lex(&cp, key); 770 if (lex != HTTPHL_WORD) 771 goto out; 772 773 /* Loop on challenges */ 774 for (; cs->count < MAX_CHALLENGES; cs->count++) { 775 cs->challenges[cs->count] = 776 malloc(sizeof(http_auth_challenge_t)); 777 if (cs->challenges[cs->count] == NULL) { 778 fetch_syserr(); 779 goto out; 780 } 781 init_http_auth_challenge(cs->challenges[cs->count]); 782 if (strcasecmp(key, "basic") == 0) { 783 cs->challenges[cs->count]->scheme = HTTPAS_BASIC; 784 } else if (strcasecmp(key, "digest") == 0) { 785 cs->challenges[cs->count]->scheme = HTTPAS_DIGEST; 786 } else { 787 cs->challenges[cs->count]->scheme = HTTPAS_UNKNOWN; 788 /* 789 * Continue parsing as basic or digest may 790 * follow, and the syntax is the same for 791 * all. We'll just ignore this one when 792 * looking at the list 793 */ 794 } 795 796 /* Loop on attributes */ 797 for (;;) { 798 /* Key */ 799 lex = http_header_lex(&cp, key); 800 if (lex != HTTPHL_WORD) 801 goto out; 802 803 /* Equal sign */ 804 lex = http_header_lex(&cp, buf); 805 if (lex != '=') 806 goto out; 807 808 /* Value */ 809 lex = http_header_lex(&cp, value); 810 if (lex != HTTPHL_WORD && lex != HTTPHL_STRING) 811 goto out; 812 813 if (strcasecmp(key, "realm") == 0) { 814 cs->challenges[cs->count]->realm = 815 strdup(value); 816 } else if (strcasecmp(key, "qop") == 0) { 817 cs->challenges[cs->count]->qop = 818 strdup(value); 819 } else if (strcasecmp(key, "nonce") == 0) { 820 cs->challenges[cs->count]->nonce = 821 strdup(value); 822 } else if (strcasecmp(key, "opaque") == 0) { 823 cs->challenges[cs->count]->opaque = 824 strdup(value); 825 } else if (strcasecmp(key, "algorithm") == 0) { 826 cs->challenges[cs->count]->algo = 827 strdup(value); 828 } else if (strcasecmp(key, "stale") == 0) { 829 cs->challenges[cs->count]->stale = 830 strcasecmp(value, "no"); 831 } else { 832 /* ignore unknown attributes */ 833 } 834 835 /* Comma or Next challenge or End */ 836 lex = http_header_lex(&cp, key); 837 /* 838 * If we get a word here, this is the beginning of the 839 * next challenge. Break the attributes loop 840 */ 841 if (lex == HTTPHL_WORD) 842 break; 843 844 if (lex == HTTPHL_END) { 845 /* End while looking for ',' is normal exit */ 846 cs->count++; 847 ret = 0; 848 goto out; 849 } 850 /* Anything else is an error */ 851 if (lex != ',') 852 goto out; 853 854 } /* End attributes loop */ 855 } /* End challenge loop */ 856 857 /* 858 * Challenges max count exceeded. This really can't happen 859 * with normal data, something's fishy -> error 860 */ 861 862 out: 863 if (key) 864 free(key); 865 if (value) 866 free(value); 867 if (buf) 868 free(buf); 869 return (ret); 870 } 871 872 873 /* 874 * Parse a last-modified header 875 */ 876 static int 877 http_parse_mtime(const char *p, time_t *mtime) 878 { 879 char locale[64], *r; 880 struct tm tm; 881 882 strlcpy(locale, setlocale(LC_TIME, NULL), sizeof(locale)); 883 setlocale(LC_TIME, "C"); 884 r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm); 885 /* 886 * Some proxies use UTC in response, but it should still be 887 * parsed. RFC2616 states GMT and UTC are exactly equal for HTTP. 888 */ 889 if (r == NULL) 890 r = strptime(p, "%a, %d %b %Y %H:%M:%S UTC", &tm); 891 /* XXX should add support for date-2 and date-3 */ 892 setlocale(LC_TIME, locale); 893 if (r == NULL) 894 return (-1); 895 DEBUGF("last modified: [%04d-%02d-%02d %02d:%02d:%02d]\n", 896 tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday, 897 tm.tm_hour, tm.tm_min, tm.tm_sec); 898 *mtime = timegm(&tm); 899 return (0); 900 } 901 902 /* 903 * Parse a content-length header 904 */ 905 static int 906 http_parse_length(const char *p, off_t *length) 907 { 908 off_t len; 909 910 for (len = 0; *p && isdigit((unsigned char)*p); ++p) 911 len = len * 10 + (*p - '0'); 912 if (*p) 913 return (-1); 914 DEBUGF("content length: [%lld]\n", (long long)len); 915 *length = len; 916 return (0); 917 } 918 919 /* 920 * Parse a content-range header 921 */ 922 static int 923 http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size) 924 { 925 off_t first, last, len; 926 927 if (strncasecmp(p, "bytes ", 6) != 0) 928 return (-1); 929 p += 6; 930 if (*p == '*') { 931 first = last = -1; 932 ++p; 933 } else { 934 for (first = 0; *p && isdigit((unsigned char)*p); ++p) 935 first = first * 10 + *p - '0'; 936 if (*p != '-') 937 return (-1); 938 for (last = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 939 last = last * 10 + *p - '0'; 940 } 941 if (first > last || *p != '/') 942 return (-1); 943 for (len = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 944 len = len * 10 + *p - '0'; 945 if (*p || len < last - first + 1) 946 return (-1); 947 if (first == -1) { 948 DEBUGF("content range: [*/%lld]\n", (long long)len); 949 *length = 0; 950 } else { 951 DEBUGF("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 #if 0 1189 DEBUGF("Calc: HA1[%s] Nonce[%s] qop[%s] method[%s] URI[%s]\n", 1190 HA1, pszNonce, pszQop, pszMethod, pszDigestUri); 1191 #endif 1192 MD5_CTX Md5Ctx; 1193 HASH HA2; 1194 HASH RespHash; 1195 HASHHEX HA2Hex; 1196 1197 // calculate H(A2) 1198 MD5Init(&Md5Ctx); 1199 MD5Update(&Md5Ctx, pszMethod, strlen(pszMethod)); 1200 MD5Update(&Md5Ctx, ":", 1); 1201 MD5Update(&Md5Ctx, pszDigestUri, strlen(pszDigestUri)); 1202 if (strcasecmp(pszQop, "auth-int") == 0) { 1203 MD5Update(&Md5Ctx, ":", 1); 1204 MD5Update(&Md5Ctx, HEntity, HASHHEXLEN); 1205 } 1206 MD5Final(HA2, &Md5Ctx); 1207 CvtHex(HA2, HA2Hex); 1208 1209 // calculate response 1210 MD5Init(&Md5Ctx); 1211 MD5Update(&Md5Ctx, HA1, HASHHEXLEN); 1212 MD5Update(&Md5Ctx, ":", 1); 1213 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce)); 1214 MD5Update(&Md5Ctx, ":", 1); 1215 if (*pszQop) { 1216 MD5Update(&Md5Ctx, pszNonceCount, strlen(pszNonceCount)); 1217 MD5Update(&Md5Ctx, ":", 1); 1218 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce)); 1219 MD5Update(&Md5Ctx, ":", 1); 1220 MD5Update(&Md5Ctx, pszQop, strlen(pszQop)); 1221 MD5Update(&Md5Ctx, ":", 1); 1222 } 1223 MD5Update(&Md5Ctx, HA2Hex, HASHHEXLEN); 1224 MD5Final(RespHash, &Md5Ctx); 1225 CvtHex(RespHash, Response); 1226 } 1227 1228 /* 1229 * Generate/Send a Digest authorization header 1230 * This looks like: [Proxy-]Authorization: credentials 1231 * 1232 * credentials = "Digest" digest-response 1233 * digest-response = 1#( username | realm | nonce | digest-uri 1234 * | response | [ algorithm ] | [cnonce] | 1235 * [opaque] | [message-qop] | 1236 * [nonce-count] | [auth-param] ) 1237 * username = "username" "=" username-value 1238 * username-value = quoted-string 1239 * digest-uri = "uri" "=" digest-uri-value 1240 * digest-uri-value = request-uri ; As specified by HTTP/1.1 1241 * message-qop = "qop" "=" qop-value 1242 * cnonce = "cnonce" "=" cnonce-value 1243 * cnonce-value = nonce-value 1244 * nonce-count = "nc" "=" nc-value 1245 * nc-value = 8LHEX 1246 * response = "response" "=" request-digest 1247 * request-digest = <"> 32LHEX <"> 1248 */ 1249 static int 1250 http_digest_auth(conn_t *conn, const char *hdr, http_auth_challenge_t *c, 1251 http_auth_params_t *parms, struct url *url) 1252 { 1253 int r; 1254 char noncecount[10]; 1255 char cnonce[40]; 1256 char *options = NULL; 1257 1258 if (!c->realm || !c->nonce) { 1259 DEBUGF("realm/nonce not set in challenge\n"); 1260 return(-1); 1261 } 1262 if (!c->algo) 1263 c->algo = strdup(""); 1264 1265 if (asprintf(&options, "%s%s%s%s", 1266 *c->algo? ",algorithm=" : "", c->algo, 1267 c->opaque? ",opaque=" : "", c->opaque?c->opaque:"") < 0) 1268 return (-1); 1269 1270 if (!c->qop) { 1271 c->qop = strdup(""); 1272 *noncecount = 0; 1273 *cnonce = 0; 1274 } else { 1275 c->nc++; 1276 sprintf(noncecount, "%08x", c->nc); 1277 /* We don't try very hard with the cnonce ... */ 1278 sprintf(cnonce, "%x%lx", getpid(), (unsigned long)time(0)); 1279 } 1280 1281 HASHHEX HA1; 1282 DigestCalcHA1(c->algo, parms->user, c->realm, 1283 parms->password, c->nonce, cnonce, HA1); 1284 DEBUGF("HA1: [%s]\n", HA1); 1285 HASHHEX digest; 1286 DigestCalcResponse(HA1, c->nonce, noncecount, cnonce, c->qop, 1287 "GET", url->doc, "", digest); 1288 1289 if (c->qop[0]) { 1290 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\"," 1291 "nonce=\"%s\",uri=\"%s\",response=\"%s\"," 1292 "qop=\"auth\", cnonce=\"%s\", nc=%s%s", 1293 hdr, parms->user, c->realm, 1294 c->nonce, url->doc, digest, 1295 cnonce, noncecount, options); 1296 } else { 1297 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\"," 1298 "nonce=\"%s\",uri=\"%s\",response=\"%s\"%s", 1299 hdr, parms->user, c->realm, 1300 c->nonce, url->doc, digest, options); 1301 } 1302 if (options) 1303 free(options); 1304 return (r); 1305 } 1306 1307 /* 1308 * Encode username and password 1309 */ 1310 static int 1311 http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd) 1312 { 1313 char *upw, *auth; 1314 int r; 1315 1316 DEBUGF("basic: usr: [%s]\n", usr); 1317 DEBUGF("basic: pwd: [%s]\n", pwd); 1318 if (asprintf(&upw, "%s:%s", usr, pwd) == -1) 1319 return (-1); 1320 auth = http_base64(upw); 1321 free(upw); 1322 if (auth == NULL) 1323 return (-1); 1324 r = http_cmd(conn, "%s: Basic %s", hdr, auth); 1325 free(auth); 1326 return (r); 1327 } 1328 1329 /* 1330 * Chose the challenge to answer and call the appropriate routine to 1331 * produce the header. 1332 */ 1333 static int 1334 http_authorize(conn_t *conn, const char *hdr, http_auth_challenges_t *cs, 1335 http_auth_params_t *parms, struct url *url) 1336 { 1337 http_auth_challenge_t *digest = NULL; 1338 int i; 1339 1340 /* If user or pass are null we're not happy */ 1341 if (!parms->user || !parms->password) { 1342 DEBUGF("NULL usr or pass\n"); 1343 return (-1); 1344 } 1345 1346 /* Look for a Digest */ 1347 for (i = 0; i < cs->count; i++) { 1348 if (cs->challenges[i]->scheme == HTTPAS_DIGEST) 1349 digest = cs->challenges[i]; 1350 } 1351 1352 /* Error if "Digest" was specified and there is no Digest challenge */ 1353 if (!digest && 1354 (parms->scheme && strcasecmp(parms->scheme, "digest") == 0)) { 1355 DEBUGF("Digest auth in env, not supported by peer\n"); 1356 return (-1); 1357 } 1358 /* 1359 * If "basic" was specified in the environment, or there is no Digest 1360 * challenge, do the basic thing. Don't need a challenge for this, 1361 * so no need to check basic!=NULL 1362 */ 1363 if (!digest || 1364 (parms->scheme && strcasecmp(parms->scheme, "basic") == 0)) 1365 return (http_basic_auth(conn,hdr,parms->user,parms->password)); 1366 1367 /* Else, prefer digest. We just checked that it's not NULL */ 1368 return (http_digest_auth(conn, hdr, digest, parms, url)); 1369 } 1370 1371 /***************************************************************************** 1372 * Helper functions for connecting to a server or proxy 1373 */ 1374 1375 /* 1376 * Connect to the correct HTTP server or proxy. 1377 */ 1378 static conn_t * 1379 http_connect(struct url *URL, struct url *purl, const char *flags) 1380 { 1381 struct url *curl; 1382 conn_t *conn; 1383 hdr_t h; 1384 http_headerbuf_t headerbuf; 1385 const char *p; 1386 int verbose; 1387 int af, val; 1388 int serrno; 1389 1390 #ifdef INET6 1391 af = AF_UNSPEC; 1392 #else 1393 af = AF_INET; 1394 #endif 1395 1396 verbose = CHECK_FLAG('v'); 1397 if (CHECK_FLAG('4')) 1398 af = AF_INET; 1399 #ifdef INET6 1400 else if (CHECK_FLAG('6')) 1401 af = AF_INET6; 1402 #endif 1403 1404 curl = (purl != NULL) ? purl : URL; 1405 1406 if ((conn = fetch_connect(curl->host, curl->port, af, verbose)) == NULL) 1407 /* fetch_connect() has already set an error code */ 1408 return (NULL); 1409 init_http_headerbuf(&headerbuf); 1410 if (strcmp(URL->scheme, SCHEME_HTTPS) == 0 && purl) { 1411 http_cmd(conn, "CONNECT %s:%d HTTP/1.1", 1412 URL->host, URL->port); 1413 http_cmd(conn, "Host: %s:%d", 1414 URL->host, URL->port); 1415 http_cmd(conn, ""); 1416 if (http_get_reply(conn) != HTTP_OK) { 1417 http_seterr(conn->err); 1418 goto ouch; 1419 } 1420 /* Read and discard the rest of the proxy response */ 1421 if (fetch_getln(conn) < 0) { 1422 fetch_syserr(); 1423 goto ouch; 1424 } 1425 do { 1426 switch ((h = http_next_header(conn, &headerbuf, &p))) { 1427 case hdr_syserror: 1428 fetch_syserr(); 1429 goto ouch; 1430 case hdr_error: 1431 http_seterr(HTTP_PROTOCOL_ERROR); 1432 goto ouch; 1433 default: 1434 /* ignore */ ; 1435 } 1436 } while (h > hdr_end); 1437 } 1438 if (strcmp(URL->scheme, SCHEME_HTTPS) == 0 && 1439 fetch_ssl(conn, URL, verbose) == -1) { 1440 /* grrr */ 1441 errno = EAUTH; 1442 fetch_syserr(); 1443 goto ouch; 1444 } 1445 1446 val = 1; 1447 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val)); 1448 1449 clean_http_headerbuf(&headerbuf); 1450 return (conn); 1451 ouch: 1452 serrno = errno; 1453 clean_http_headerbuf(&headerbuf); 1454 fetch_close(conn); 1455 errno = serrno; 1456 return (NULL); 1457 } 1458 1459 static struct url * 1460 http_get_proxy(struct url * url, const char *flags) 1461 { 1462 struct url *purl; 1463 char *p; 1464 1465 if (flags != NULL && strchr(flags, 'd') != NULL) 1466 return (NULL); 1467 if (fetch_no_proxy_match(url->host)) 1468 return (NULL); 1469 if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) && 1470 *p && (purl = fetchParseURL(p))) { 1471 if (!*purl->scheme) 1472 strcpy(purl->scheme, SCHEME_HTTP); 1473 if (!purl->port) 1474 purl->port = fetch_default_proxy_port(purl->scheme); 1475 if (strcmp(purl->scheme, SCHEME_HTTP) == 0) 1476 return (purl); 1477 fetchFreeURL(purl); 1478 } 1479 return (NULL); 1480 } 1481 1482 static void 1483 http_print_html(FILE *out, FILE *in) 1484 { 1485 size_t len; 1486 char *line, *p, *q; 1487 int comment, tag; 1488 1489 comment = tag = 0; 1490 while ((line = fgetln(in, &len)) != NULL) { 1491 while (len && isspace((unsigned char)line[len - 1])) 1492 --len; 1493 for (p = q = line; q < line + len; ++q) { 1494 if (comment && *q == '-') { 1495 if (q + 2 < line + len && 1496 strcmp(q, "-->") == 0) { 1497 tag = comment = 0; 1498 q += 2; 1499 } 1500 } else if (tag && !comment && *q == '>') { 1501 p = q + 1; 1502 tag = 0; 1503 } else if (!tag && *q == '<') { 1504 if (q > p) 1505 fwrite(p, q - p, 1, out); 1506 tag = 1; 1507 if (q + 3 < line + len && 1508 strcmp(q, "<!--") == 0) { 1509 comment = 1; 1510 q += 3; 1511 } 1512 } 1513 } 1514 if (!tag && q > p) 1515 fwrite(p, q - p, 1, out); 1516 fputc('\n', out); 1517 } 1518 } 1519 1520 1521 /***************************************************************************** 1522 * Core 1523 */ 1524 1525 FILE * 1526 http_request(struct url *URL, const char *op, struct url_stat *us, 1527 struct url *purl, const char *flags) 1528 { 1529 1530 return (http_request_body(URL, op, us, purl, flags, NULL, NULL)); 1531 } 1532 1533 /* 1534 * Send a request and process the reply 1535 * 1536 * XXX This function is way too long, the do..while loop should be split 1537 * XXX off into a separate function. 1538 */ 1539 FILE * 1540 http_request_body(struct url *URL, const char *op, struct url_stat *us, 1541 struct url *purl, const char *flags, const char *content_type, 1542 const char *body) 1543 { 1544 char timebuf[80]; 1545 char hbuf[MAXHOSTNAMELEN + 7], *host; 1546 conn_t *conn; 1547 struct url *url, *new; 1548 int chunked, direct, ims, noredirect, verbose; 1549 int e, i, n, val; 1550 off_t offset, clength, length, size; 1551 time_t mtime; 1552 const char *p; 1553 FILE *f; 1554 hdr_t h; 1555 struct tm *timestruct; 1556 http_headerbuf_t headerbuf; 1557 http_auth_challenges_t server_challenges; 1558 http_auth_challenges_t proxy_challenges; 1559 size_t body_len; 1560 1561 /* The following calls don't allocate anything */ 1562 init_http_headerbuf(&headerbuf); 1563 init_http_auth_challenges(&server_challenges); 1564 init_http_auth_challenges(&proxy_challenges); 1565 1566 direct = CHECK_FLAG('d'); 1567 noredirect = CHECK_FLAG('A'); 1568 verbose = CHECK_FLAG('v'); 1569 ims = CHECK_FLAG('i'); 1570 1571 if (direct && purl) { 1572 fetchFreeURL(purl); 1573 purl = NULL; 1574 } 1575 1576 /* try the provided URL first */ 1577 url = URL; 1578 1579 n = MAX_REDIRECT; 1580 i = 0; 1581 1582 e = HTTP_PROTOCOL_ERROR; 1583 do { 1584 new = NULL; 1585 chunked = 0; 1586 offset = 0; 1587 clength = -1; 1588 length = -1; 1589 size = -1; 1590 mtime = 0; 1591 1592 /* check port */ 1593 if (!url->port) 1594 url->port = fetch_default_port(url->scheme); 1595 1596 /* were we redirected to an FTP URL? */ 1597 if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) { 1598 if (strcmp(op, "GET") == 0) 1599 return (ftp_request(url, "RETR", us, purl, flags)); 1600 else if (strcmp(op, "HEAD") == 0) 1601 return (ftp_request(url, "STAT", us, purl, flags)); 1602 } 1603 1604 /* connect to server or proxy */ 1605 if ((conn = http_connect(url, purl, flags)) == NULL) 1606 goto ouch; 1607 1608 /* append port number only if necessary */ 1609 host = url->host; 1610 if (url->port != fetch_default_port(url->scheme)) { 1611 snprintf(hbuf, sizeof(hbuf), "%s:%d", host, url->port); 1612 host = hbuf; 1613 } 1614 1615 /* send request */ 1616 if (verbose) 1617 fetch_info("requesting %s://%s%s", 1618 url->scheme, host, url->doc); 1619 if (purl && strcmp(url->scheme, SCHEME_HTTPS) != 0) { 1620 http_cmd(conn, "%s %s://%s%s HTTP/1.1", 1621 op, url->scheme, host, url->doc); 1622 } else { 1623 http_cmd(conn, "%s %s HTTP/1.1", 1624 op, url->doc); 1625 } 1626 1627 if (ims && url->ims_time) { 1628 timestruct = gmtime((time_t *)&url->ims_time); 1629 (void)strftime(timebuf, 80, "%a, %d %b %Y %T GMT", 1630 timestruct); 1631 if (verbose) 1632 fetch_info("If-Modified-Since: %s", timebuf); 1633 http_cmd(conn, "If-Modified-Since: %s", timebuf); 1634 } 1635 /* virtual host */ 1636 http_cmd(conn, "Host: %s", host); 1637 1638 /* 1639 * Proxy authorization: we only send auth after we received 1640 * a 407 error. We do not first try basic anyway (changed 1641 * when support was added for digest-auth) 1642 */ 1643 if (purl && proxy_challenges.valid) { 1644 http_auth_params_t aparams; 1645 init_http_auth_params(&aparams); 1646 if (*purl->user || *purl->pwd) { 1647 aparams.user = strdup(purl->user); 1648 aparams.password = strdup(purl->pwd); 1649 } else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL && 1650 *p != '\0') { 1651 if (http_authfromenv(p, &aparams) < 0) { 1652 http_seterr(HTTP_NEED_PROXY_AUTH); 1653 goto ouch; 1654 } 1655 } else if (fetch_netrc_auth(purl) == 0) { 1656 aparams.user = strdup(purl->user); 1657 aparams.password = strdup(purl->pwd); 1658 } 1659 http_authorize(conn, "Proxy-Authorization", 1660 &proxy_challenges, &aparams, url); 1661 clean_http_auth_params(&aparams); 1662 } 1663 1664 /* 1665 * Server authorization: we never send "a priori" 1666 * Basic auth, which used to be done if user/pass were 1667 * set in the url. This would be weird because we'd send the 1668 * password in the clear even if Digest is finally to be 1669 * used (it would have made more sense for the 1670 * pre-digest version to do this when Basic was specified 1671 * in the environment) 1672 */ 1673 if (server_challenges.valid) { 1674 http_auth_params_t aparams; 1675 init_http_auth_params(&aparams); 1676 if (*url->user || *url->pwd) { 1677 aparams.user = strdup(url->user); 1678 aparams.password = strdup(url->pwd); 1679 } else if ((p = getenv("HTTP_AUTH")) != NULL && 1680 *p != '\0') { 1681 if (http_authfromenv(p, &aparams) < 0) { 1682 http_seterr(HTTP_NEED_AUTH); 1683 goto ouch; 1684 } 1685 } else if (fetch_netrc_auth(url) == 0) { 1686 aparams.user = strdup(url->user); 1687 aparams.password = strdup(url->pwd); 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_TEMP_REDIRECT: 1764 case HTTP_PERM_REDIRECT: 1765 case HTTP_SEE_OTHER: 1766 case HTTP_USE_PROXY: 1767 /* 1768 * Not so fine, but we still have to read the 1769 * headers to get the new location. 1770 */ 1771 break; 1772 case HTTP_NEED_AUTH: 1773 if (server_challenges.valid) { 1774 /* 1775 * We already sent out authorization code, 1776 * so there's nothing more we can do. 1777 */ 1778 http_seterr(conn->err); 1779 goto ouch; 1780 } 1781 /* try again, but send the password this time */ 1782 if (verbose) 1783 fetch_info("server requires authorization"); 1784 break; 1785 case HTTP_NEED_PROXY_AUTH: 1786 if (proxy_challenges.valid) { 1787 /* 1788 * We already sent our proxy 1789 * authorization code, so there's 1790 * nothing more we can do. */ 1791 http_seterr(conn->err); 1792 goto ouch; 1793 } 1794 /* try again, but send the password this time */ 1795 if (verbose) 1796 fetch_info("proxy requires authorization"); 1797 break; 1798 case HTTP_BAD_RANGE: 1799 /* 1800 * This can happen if we ask for 0 bytes because 1801 * we already have the whole file. Consider this 1802 * a success for now, and check sizes later. 1803 */ 1804 break; 1805 case HTTP_PROTOCOL_ERROR: 1806 /* fall through */ 1807 case -1: 1808 fetch_syserr(); 1809 goto ouch; 1810 default: 1811 http_seterr(conn->err); 1812 if (!verbose) 1813 goto ouch; 1814 /* fall through so we can get the full error message */ 1815 } 1816 1817 /* get headers. http_next_header expects one line readahead */ 1818 if (fetch_getln(conn) == -1) { 1819 fetch_syserr(); 1820 goto ouch; 1821 } 1822 do { 1823 switch ((h = http_next_header(conn, &headerbuf, &p))) { 1824 case hdr_syserror: 1825 fetch_syserr(); 1826 goto ouch; 1827 case hdr_error: 1828 http_seterr(HTTP_PROTOCOL_ERROR); 1829 goto ouch; 1830 case hdr_content_length: 1831 http_parse_length(p, &clength); 1832 break; 1833 case hdr_content_range: 1834 http_parse_range(p, &offset, &length, &size); 1835 break; 1836 case hdr_last_modified: 1837 http_parse_mtime(p, &mtime); 1838 break; 1839 case hdr_location: 1840 if (!HTTP_REDIRECT(conn->err)) 1841 break; 1842 /* 1843 * if the A flag is set, we don't follow 1844 * temporary redirects. 1845 */ 1846 if (noredirect && 1847 conn->err != HTTP_MOVED_PERM && 1848 conn->err != HTTP_PERM_REDIRECT && 1849 conn->err != HTTP_USE_PROXY) { 1850 n = 1; 1851 break; 1852 } 1853 if (new) 1854 free(new); 1855 if (verbose) 1856 fetch_info("%d redirect to %s", 1857 conn->err, p); 1858 if (*p == '/') 1859 /* absolute path */ 1860 new = fetchMakeURL(url->scheme, url->host, 1861 url->port, p, url->user, url->pwd); 1862 else 1863 new = fetchParseURL(p); 1864 if (new == NULL) { 1865 /* XXX should set an error code */ 1866 DEBUGF("failed to parse new URL\n"); 1867 goto ouch; 1868 } 1869 1870 /* Only copy credentials if the host matches */ 1871 if (strcmp(new->host, url->host) == 0 && 1872 !*new->user && !*new->pwd) { 1873 strcpy(new->user, url->user); 1874 strcpy(new->pwd, url->pwd); 1875 } 1876 new->offset = url->offset; 1877 new->length = url->length; 1878 new->ims_time = url->ims_time; 1879 break; 1880 case hdr_transfer_encoding: 1881 /* XXX weak test*/ 1882 chunked = (strcasecmp(p, "chunked") == 0); 1883 break; 1884 case hdr_www_authenticate: 1885 if (conn->err != HTTP_NEED_AUTH) 1886 break; 1887 if (http_parse_authenticate(p, &server_challenges) == 0) 1888 ++n; 1889 break; 1890 case hdr_proxy_authenticate: 1891 if (conn->err != HTTP_NEED_PROXY_AUTH) 1892 break; 1893 if (http_parse_authenticate(p, &proxy_challenges) == 0) 1894 ++n; 1895 break; 1896 case hdr_end: 1897 /* fall through */ 1898 case hdr_unknown: 1899 /* ignore */ 1900 break; 1901 } 1902 } while (h > hdr_end); 1903 1904 /* we need to provide authentication */ 1905 if (conn->err == HTTP_NEED_AUTH || 1906 conn->err == HTTP_NEED_PROXY_AUTH) { 1907 e = conn->err; 1908 if ((conn->err == HTTP_NEED_AUTH && 1909 !server_challenges.valid) || 1910 (conn->err == HTTP_NEED_PROXY_AUTH && 1911 !proxy_challenges.valid)) { 1912 /* 401/7 but no www/proxy-authenticate ?? */ 1913 DEBUGF("%03d without auth header\n", conn->err); 1914 goto ouch; 1915 } 1916 fetch_close(conn); 1917 conn = NULL; 1918 continue; 1919 } 1920 1921 /* requested range not satisfiable */ 1922 if (conn->err == HTTP_BAD_RANGE) { 1923 if (url->offset > 0 && url->length == 0) { 1924 /* asked for 0 bytes; fake it */ 1925 offset = url->offset; 1926 clength = -1; 1927 conn->err = HTTP_OK; 1928 break; 1929 } else { 1930 http_seterr(conn->err); 1931 goto ouch; 1932 } 1933 } 1934 1935 /* we have a hit or an error */ 1936 if (conn->err == HTTP_OK 1937 || conn->err == HTTP_NOT_MODIFIED 1938 || conn->err == HTTP_PARTIAL 1939 || HTTP_ERROR(conn->err)) 1940 break; 1941 1942 /* all other cases: we got a redirect */ 1943 e = conn->err; 1944 clean_http_auth_challenges(&server_challenges); 1945 fetch_close(conn); 1946 conn = NULL; 1947 if (!new) { 1948 DEBUGF("redirect with no new location\n"); 1949 break; 1950 } 1951 if (url != URL) 1952 fetchFreeURL(url); 1953 url = new; 1954 } while (++i < n); 1955 1956 /* we failed, or ran out of retries */ 1957 if (conn == NULL) { 1958 http_seterr(e); 1959 goto ouch; 1960 } 1961 1962 DEBUGF("offset %lld, length %lld, size %lld, clength %lld\n", 1963 (long long)offset, (long long)length, 1964 (long long)size, (long long)clength); 1965 1966 if (conn->err == HTTP_NOT_MODIFIED) { 1967 http_seterr(HTTP_NOT_MODIFIED); 1968 return (NULL); 1969 } 1970 1971 /* check for inconsistencies */ 1972 if (clength != -1 && length != -1 && clength != length) { 1973 http_seterr(HTTP_PROTOCOL_ERROR); 1974 goto ouch; 1975 } 1976 if (clength == -1) 1977 clength = length; 1978 if (clength != -1) 1979 length = offset + clength; 1980 if (length != -1 && size != -1 && length != size) { 1981 http_seterr(HTTP_PROTOCOL_ERROR); 1982 goto ouch; 1983 } 1984 if (size == -1) 1985 size = length; 1986 1987 /* fill in stats */ 1988 if (us) { 1989 us->size = size; 1990 us->atime = us->mtime = mtime; 1991 } 1992 1993 /* too far? */ 1994 if (URL->offset > 0 && offset > URL->offset) { 1995 http_seterr(HTTP_PROTOCOL_ERROR); 1996 goto ouch; 1997 } 1998 1999 /* report back real offset and size */ 2000 URL->offset = offset; 2001 URL->length = clength; 2002 2003 /* wrap it up in a FILE */ 2004 if ((f = http_funopen(conn, chunked)) == NULL) { 2005 fetch_syserr(); 2006 goto ouch; 2007 } 2008 2009 if (url != URL) 2010 fetchFreeURL(url); 2011 if (purl) 2012 fetchFreeURL(purl); 2013 2014 if (HTTP_ERROR(conn->err)) { 2015 http_print_html(stderr, f); 2016 fclose(f); 2017 f = NULL; 2018 } 2019 clean_http_headerbuf(&headerbuf); 2020 clean_http_auth_challenges(&server_challenges); 2021 clean_http_auth_challenges(&proxy_challenges); 2022 return (f); 2023 2024 ouch: 2025 if (url != URL) 2026 fetchFreeURL(url); 2027 if (purl) 2028 fetchFreeURL(purl); 2029 if (conn != NULL) 2030 fetch_close(conn); 2031 clean_http_headerbuf(&headerbuf); 2032 clean_http_auth_challenges(&server_challenges); 2033 clean_http_auth_challenges(&proxy_challenges); 2034 return (NULL); 2035 } 2036 2037 2038 /***************************************************************************** 2039 * Entry points 2040 */ 2041 2042 /* 2043 * Retrieve and stat a file by HTTP 2044 */ 2045 FILE * 2046 fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags) 2047 { 2048 return (http_request(URL, "GET", us, http_get_proxy(URL, flags), flags)); 2049 } 2050 2051 /* 2052 * Retrieve a file by HTTP 2053 */ 2054 FILE * 2055 fetchGetHTTP(struct url *URL, const char *flags) 2056 { 2057 return (fetchXGetHTTP(URL, NULL, flags)); 2058 } 2059 2060 /* 2061 * Store a file by HTTP 2062 */ 2063 FILE * 2064 fetchPutHTTP(struct url *URL __unused, const char *flags __unused) 2065 { 2066 warnx("fetchPutHTTP(): not implemented"); 2067 return (NULL); 2068 } 2069 2070 /* 2071 * Get an HTTP document's metadata 2072 */ 2073 int 2074 fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags) 2075 { 2076 FILE *f; 2077 2078 f = http_request(URL, "HEAD", us, http_get_proxy(URL, flags), flags); 2079 if (f == NULL) 2080 return (-1); 2081 fclose(f); 2082 return (0); 2083 } 2084 2085 /* 2086 * List a directory 2087 */ 2088 struct url_ent * 2089 fetchListHTTP(struct url *url __unused, const char *flags __unused) 2090 { 2091 warnx("fetchListHTTP(): not implemented"); 2092 return (NULL); 2093 } 2094 2095 FILE * 2096 fetchReqHTTP(struct url *URL, const char *method, const char *flags, 2097 const char *content_type, const char *body) 2098 { 2099 2100 return (http_request_body(URL, method, NULL, http_get_proxy(URL, flags), 2101 flags, content_type, body)); 2102 } 2103