http.c revision 241839
1/*- 2 * Copyright (c) 2000-2011 Dag-Erling Sm��rgrav 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: head/lib/libfetch/http.c 241839 2012-10-22 03:00:04Z eadler $"); 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_TEMP_REDIRECT 307 107#define HTTP_NEED_AUTH 401 108#define HTTP_NEED_PROXY_AUTH 407 109#define HTTP_BAD_RANGE 416 110#define HTTP_PROTOCOL_ERROR 999 111 112#define HTTP_REDIRECT(xyz) ((xyz) == HTTP_MOVED_PERM \ 113 || (xyz) == HTTP_MOVED_TEMP \ 114 || (xyz) == HTTP_TEMP_REDIRECT \ 115 || (xyz) == HTTP_SEE_OTHER) 116 117#define HTTP_ERROR(xyz) ((xyz) > 400 && (xyz) < 599) 118 119 120/***************************************************************************** 121 * I/O functions for decoding chunked streams 122 */ 123 124struct httpio 125{ 126 conn_t *conn; /* connection */ 127 int chunked; /* chunked mode */ 128 char *buf; /* chunk buffer */ 129 size_t bufsize; /* size of chunk buffer */ 130 ssize_t buflen; /* amount of data currently in buffer */ 131 int bufpos; /* current read offset in buffer */ 132 int eof; /* end-of-file flag */ 133 int error; /* error flag */ 134 size_t chunksize; /* remaining size of current chunk */ 135#ifndef NDEBUG 136 size_t total; 137#endif 138}; 139 140/* 141 * Get next chunk header 142 */ 143static int 144http_new_chunk(struct httpio *io) 145{ 146 char *p; 147 148 if (fetch_getln(io->conn) == -1) 149 return (-1); 150 151 if (io->conn->buflen < 2 || !isxdigit((unsigned char)*io->conn->buf)) 152 return (-1); 153 154 for (p = io->conn->buf; *p && !isspace((unsigned char)*p); ++p) { 155 if (*p == ';') 156 break; 157 if (!isxdigit((unsigned char)*p)) 158 return (-1); 159 if (isdigit((unsigned char)*p)) { 160 io->chunksize = io->chunksize * 16 + 161 *p - '0'; 162 } else { 163 io->chunksize = io->chunksize * 16 + 164 10 + tolower((unsigned char)*p) - 'a'; 165 } 166 } 167 168#ifndef NDEBUG 169 if (fetchDebug) { 170 io->total += io->chunksize; 171 if (io->chunksize == 0) 172 fprintf(stderr, "%s(): end of last chunk\n", __func__); 173 else 174 fprintf(stderr, "%s(): new chunk: %lu (%lu)\n", 175 __func__, (unsigned long)io->chunksize, 176 (unsigned long)io->total); 177 } 178#endif 179 180 return (io->chunksize); 181} 182 183/* 184 * Grow the input buffer to at least len bytes 185 */ 186static inline int 187http_growbuf(struct httpio *io, size_t len) 188{ 189 char *tmp; 190 191 if (io->bufsize >= len) 192 return (0); 193 194 if ((tmp = realloc(io->buf, len)) == NULL) 195 return (-1); 196 io->buf = tmp; 197 io->bufsize = len; 198 return (0); 199} 200 201/* 202 * Fill the input buffer, do chunk decoding on the fly 203 */ 204static int 205http_fillbuf(struct httpio *io, size_t len) 206{ 207 ssize_t nbytes; 208 209 if (io->error) 210 return (-1); 211 if (io->eof) 212 return (0); 213 214 if (io->chunked == 0) { 215 if (http_growbuf(io, len) == -1) 216 return (-1); 217 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) { 218 io->error = errno; 219 return (-1); 220 } 221 io->buflen = nbytes; 222 io->bufpos = 0; 223 return (io->buflen); 224 } 225 226 if (io->chunksize == 0) { 227 switch (http_new_chunk(io)) { 228 case -1: 229 io->error = 1; 230 return (-1); 231 case 0: 232 io->eof = 1; 233 return (0); 234 } 235 } 236 237 if (len > io->chunksize) 238 len = io->chunksize; 239 if (http_growbuf(io, len) == -1) 240 return (-1); 241 if ((nbytes = fetch_read(io->conn, io->buf, len)) == -1) { 242 io->error = errno; 243 return (-1); 244 } 245 io->buflen = nbytes; 246 io->chunksize -= io->buflen; 247 248 if (io->chunksize == 0) { 249 char endl[2]; 250 251 if (fetch_read(io->conn, endl, 2) != 2 || 252 endl[0] != '\r' || endl[1] != '\n') 253 return (-1); 254 } 255 256 io->bufpos = 0; 257 258 return (io->buflen); 259} 260 261/* 262 * Read function 263 */ 264static int 265http_readfn(void *v, char *buf, int len) 266{ 267 struct httpio *io = (struct httpio *)v; 268 int l, pos; 269 270 if (io->error) 271 return (-1); 272 if (io->eof) 273 return (0); 274 275 for (pos = 0; len > 0; pos += l, len -= l) { 276 /* empty buffer */ 277 if (!io->buf || io->bufpos == io->buflen) 278 if (http_fillbuf(io, len) < 1) 279 break; 280 l = io->buflen - io->bufpos; 281 if (len < l) 282 l = len; 283 memcpy(buf + pos, io->buf + io->bufpos, l); 284 io->bufpos += l; 285 } 286 287 if (!pos && io->error) { 288 if (io->error == EINTR) 289 io->error = 0; 290 return (-1); 291 } 292 return (pos); 293} 294 295/* 296 * Write function 297 */ 298static int 299http_writefn(void *v, const char *buf, int len) 300{ 301 struct httpio *io = (struct httpio *)v; 302 303 return (fetch_write(io->conn, buf, len)); 304} 305 306/* 307 * Close function 308 */ 309static int 310http_closefn(void *v) 311{ 312 struct httpio *io = (struct httpio *)v; 313 int r; 314 315 r = fetch_close(io->conn); 316 if (io->buf) 317 free(io->buf); 318 free(io); 319 return (r); 320} 321 322/* 323 * Wrap a file descriptor up 324 */ 325static FILE * 326http_funopen(conn_t *conn, int chunked) 327{ 328 struct httpio *io; 329 FILE *f; 330 331 if ((io = calloc(1, sizeof(*io))) == NULL) { 332 fetch_syserr(); 333 return (NULL); 334 } 335 io->conn = conn; 336 io->chunked = chunked; 337 f = funopen(io, http_readfn, http_writefn, NULL, http_closefn); 338 if (f == NULL) { 339 fetch_syserr(); 340 free(io); 341 return (NULL); 342 } 343 return (f); 344} 345 346 347/***************************************************************************** 348 * Helper functions for talking to the server and parsing its replies 349 */ 350 351/* Header types */ 352typedef enum { 353 hdr_syserror = -2, 354 hdr_error = -1, 355 hdr_end = 0, 356 hdr_unknown = 1, 357 hdr_content_length, 358 hdr_content_range, 359 hdr_last_modified, 360 hdr_location, 361 hdr_transfer_encoding, 362 hdr_www_authenticate, 363 hdr_proxy_authenticate, 364} hdr_t; 365 366/* Names of interesting headers */ 367static struct { 368 hdr_t num; 369 const char *name; 370} hdr_names[] = { 371 { hdr_content_length, "Content-Length" }, 372 { hdr_content_range, "Content-Range" }, 373 { hdr_last_modified, "Last-Modified" }, 374 { hdr_location, "Location" }, 375 { hdr_transfer_encoding, "Transfer-Encoding" }, 376 { hdr_www_authenticate, "WWW-Authenticate" }, 377 { hdr_proxy_authenticate, "Proxy-Authenticate" }, 378 { hdr_unknown, NULL }, 379}; 380 381/* 382 * Send a formatted line; optionally echo to terminal 383 */ 384static int 385http_cmd(conn_t *conn, const char *fmt, ...) 386{ 387 va_list ap; 388 size_t len; 389 char *msg; 390 int r; 391 392 va_start(ap, fmt); 393 len = vasprintf(&msg, fmt, ap); 394 va_end(ap); 395 396 if (msg == NULL) { 397 errno = ENOMEM; 398 fetch_syserr(); 399 return (-1); 400 } 401 402 r = fetch_putln(conn, msg, len); 403 free(msg); 404 405 if (r == -1) { 406 fetch_syserr(); 407 return (-1); 408 } 409 410 return (0); 411} 412 413/* 414 * Get and parse status line 415 */ 416static int 417http_get_reply(conn_t *conn) 418{ 419 char *p; 420 421 if (fetch_getln(conn) == -1) 422 return (-1); 423 /* 424 * A valid status line looks like "HTTP/m.n xyz reason" where m 425 * and n are the major and minor protocol version numbers and xyz 426 * is the reply code. 427 * Unfortunately, there are servers out there (NCSA 1.5.1, to name 428 * just one) that do not send a version number, so we can't rely 429 * on finding one, but if we do, insist on it being 1.0 or 1.1. 430 * We don't care about the reason phrase. 431 */ 432 if (strncmp(conn->buf, "HTTP", 4) != 0) 433 return (HTTP_PROTOCOL_ERROR); 434 p = conn->buf + 4; 435 if (*p == '/') { 436 if (p[1] != '1' || p[2] != '.' || (p[3] != '0' && p[3] != '1')) 437 return (HTTP_PROTOCOL_ERROR); 438 p += 4; 439 } 440 if (*p != ' ' || 441 !isdigit((unsigned char)p[1]) || 442 !isdigit((unsigned char)p[2]) || 443 !isdigit((unsigned char)p[3])) 444 return (HTTP_PROTOCOL_ERROR); 445 446 conn->err = (p[1] - '0') * 100 + (p[2] - '0') * 10 + (p[3] - '0'); 447 return (conn->err); 448} 449 450/* 451 * Check a header; if the type matches the given string, return a pointer 452 * to the beginning of the value. 453 */ 454static const char * 455http_match(const char *str, const char *hdr) 456{ 457 while (*str && *hdr && 458 tolower((unsigned char)*str++) == tolower((unsigned char)*hdr++)) 459 /* nothing */; 460 if (*str || *hdr != ':') 461 return (NULL); 462 while (*hdr && isspace((unsigned char)*++hdr)) 463 /* nothing */; 464 return (hdr); 465} 466 467 468/* 469 * Get the next header and return the appropriate symbolic code. We 470 * need to read one line ahead for checking for a continuation line 471 * belonging to the current header (continuation lines start with 472 * white space). 473 * 474 * We get called with a fresh line already in the conn buffer, either 475 * from the previous http_next_header() invocation, or, the first 476 * time, from a fetch_getln() performed by our caller. 477 * 478 * This stops when we encounter an empty line (we dont read beyond the header 479 * area). 480 * 481 * Note that the "headerbuf" is just a place to return the result. Its 482 * contents are not used for the next call. This means that no cleanup 483 * is needed when ie doing another connection, just call the cleanup when 484 * fully done to deallocate memory. 485 */ 486 487/* Limit the max number of continuation lines to some reasonable value */ 488#define HTTP_MAX_CONT_LINES 10 489 490/* Place into which to build a header from one or several lines */ 491typedef struct { 492 char *buf; /* buffer */ 493 size_t bufsize; /* buffer size */ 494 size_t buflen; /* length of buffer contents */ 495} http_headerbuf_t; 496 497static void 498init_http_headerbuf(http_headerbuf_t *buf) 499{ 500 buf->buf = NULL; 501 buf->bufsize = 0; 502 buf->buflen = 0; 503} 504 505static void 506clean_http_headerbuf(http_headerbuf_t *buf) 507{ 508 if (buf->buf) 509 free(buf->buf); 510 init_http_headerbuf(buf); 511} 512 513/* Remove whitespace at the end of the buffer */ 514static void 515http_conn_trimright(conn_t *conn) 516{ 517 while (conn->buflen && 518 isspace((unsigned char)conn->buf[conn->buflen - 1])) 519 conn->buflen--; 520 conn->buf[conn->buflen] = '\0'; 521} 522 523static hdr_t 524http_next_header(conn_t *conn, http_headerbuf_t *hbuf, const char **p) 525{ 526 unsigned int i, len; 527 528 /* 529 * Have to do the stripping here because of the first line. So 530 * it's done twice for the subsequent lines. No big deal 531 */ 532 http_conn_trimright(conn); 533 if (conn->buflen == 0) 534 return (hdr_end); 535 536 /* Copy the line to the headerbuf */ 537 if (hbuf->bufsize < conn->buflen + 1) { 538 if ((hbuf->buf = realloc(hbuf->buf, conn->buflen + 1)) == NULL) 539 return (hdr_syserror); 540 hbuf->bufsize = conn->buflen + 1; 541 } 542 strcpy(hbuf->buf, conn->buf); 543 hbuf->buflen = conn->buflen; 544 545 /* 546 * Fetch possible continuation lines. Stop at 1st non-continuation 547 * and leave it in the conn buffer 548 */ 549 for (i = 0; i < HTTP_MAX_CONT_LINES; i++) { 550 if (fetch_getln(conn) == -1) 551 return (hdr_syserror); 552 553 /* 554 * Note: we carry on the idea from the previous version 555 * that a pure whitespace line is equivalent to an empty 556 * one (so it's not continuation and will be handled when 557 * we are called next) 558 */ 559 http_conn_trimright(conn); 560 if (conn->buf[0] != ' ' && conn->buf[0] != "\t"[0]) 561 break; 562 563 /* Got a continuation line. Concatenate to previous */ 564 len = hbuf->buflen + conn->buflen; 565 if (hbuf->bufsize < len + 1) { 566 len *= 2; 567 if ((hbuf->buf = realloc(hbuf->buf, len + 1)) == NULL) 568 return (hdr_syserror); 569 hbuf->bufsize = len + 1; 570 } 571 strcpy(hbuf->buf + hbuf->buflen, conn->buf); 572 hbuf->buflen += conn->buflen; 573 } 574 575 /* 576 * We could check for malformed headers but we don't really care. 577 * A valid header starts with a token immediately followed by a 578 * colon; a token is any sequence of non-control, non-whitespace 579 * characters except "()<>@,;:\\\"{}". 580 */ 581 for (i = 0; hdr_names[i].num != hdr_unknown; i++) 582 if ((*p = http_match(hdr_names[i].name, hbuf->buf)) != NULL) 583 return (hdr_names[i].num); 584 585 return (hdr_unknown); 586} 587 588/************************** 589 * [Proxy-]Authenticate header parsing 590 */ 591 592/* 593 * Read doublequote-delimited string into output buffer obuf (allocated 594 * by caller, whose responsibility it is to ensure that it's big enough) 595 * cp points to the first char after the initial '"' 596 * Handles \ quoting 597 * Returns pointer to the first char after the terminating double quote, or 598 * NULL for error. 599 */ 600static const char * 601http_parse_headerstring(const char *cp, char *obuf) 602{ 603 for (;;) { 604 switch (*cp) { 605 case 0: /* Unterminated string */ 606 *obuf = 0; 607 return (NULL); 608 case '"': /* Ending quote */ 609 *obuf = 0; 610 return (++cp); 611 case '\\': 612 if (*++cp == 0) { 613 *obuf = 0; 614 return (NULL); 615 } 616 /* FALLTHROUGH */ 617 default: 618 *obuf++ = *cp++; 619 } 620 } 621} 622 623/* Http auth challenge schemes */ 624typedef enum {HTTPAS_UNKNOWN, HTTPAS_BASIC,HTTPAS_DIGEST} http_auth_schemes_t; 625 626/* Data holder for a Basic or Digest challenge. */ 627typedef struct { 628 http_auth_schemes_t scheme; 629 char *realm; 630 char *qop; 631 char *nonce; 632 char *opaque; 633 char *algo; 634 int stale; 635 int nc; /* Nonce count */ 636} http_auth_challenge_t; 637 638static void 639init_http_auth_challenge(http_auth_challenge_t *b) 640{ 641 b->scheme = HTTPAS_UNKNOWN; 642 b->realm = b->qop = b->nonce = b->opaque = b->algo = NULL; 643 b->stale = b->nc = 0; 644} 645 646static void 647clean_http_auth_challenge(http_auth_challenge_t *b) 648{ 649 if (b->realm) 650 free(b->realm); 651 if (b->qop) 652 free(b->qop); 653 if (b->nonce) 654 free(b->nonce); 655 if (b->opaque) 656 free(b->opaque); 657 if (b->algo) 658 free(b->algo); 659 init_http_auth_challenge(b); 660} 661 662/* Data holder for an array of challenges offered in an http response. */ 663#define MAX_CHALLENGES 10 664typedef struct { 665 http_auth_challenge_t *challenges[MAX_CHALLENGES]; 666 int count; /* Number of parsed challenges in the array */ 667 int valid; /* We did parse an authenticate header */ 668} http_auth_challenges_t; 669 670static void 671init_http_auth_challenges(http_auth_challenges_t *cs) 672{ 673 int i; 674 for (i = 0; i < MAX_CHALLENGES; i++) 675 cs->challenges[i] = NULL; 676 cs->count = cs->valid = 0; 677} 678 679static void 680clean_http_auth_challenges(http_auth_challenges_t *cs) 681{ 682 int i; 683 /* We rely on non-zero pointers being allocated, not on the count */ 684 for (i = 0; i < MAX_CHALLENGES; i++) { 685 if (cs->challenges[i] != NULL) { 686 clean_http_auth_challenge(cs->challenges[i]); 687 free(cs->challenges[i]); 688 } 689 } 690 init_http_auth_challenges(cs); 691} 692 693/* 694 * Enumeration for lexical elements. Separators will be returned as their own 695 * ascii value 696 */ 697typedef enum {HTTPHL_WORD=256, HTTPHL_STRING=257, HTTPHL_END=258, 698 HTTPHL_ERROR = 259} http_header_lex_t; 699 700/* 701 * Determine what kind of token comes next and return possible value 702 * in buf, which is supposed to have been allocated big enough by 703 * caller. Advance input pointer and return element type. 704 */ 705static int 706http_header_lex(const char **cpp, char *buf) 707{ 708 size_t l; 709 /* Eat initial whitespace */ 710 *cpp += strspn(*cpp, " \t"); 711 if (**cpp == 0) 712 return (HTTPHL_END); 713 714 /* Separator ? */ 715 if (**cpp == ',' || **cpp == '=') 716 return (*((*cpp)++)); 717 718 /* String ? */ 719 if (**cpp == '"') { 720 *cpp = http_parse_headerstring(++*cpp, buf); 721 if (*cpp == NULL) 722 return (HTTPHL_ERROR); 723 return (HTTPHL_STRING); 724 } 725 726 /* Read other token, until separator or whitespace */ 727 l = strcspn(*cpp, " \t,="); 728 memcpy(buf, *cpp, l); 729 buf[l] = 0; 730 *cpp += l; 731 return (HTTPHL_WORD); 732} 733 734/* 735 * Read challenges from http xxx-authenticate header and accumulate them 736 * in the challenges list structure. 737 * 738 * Headers with multiple challenges are specified by rfc2617, but 739 * servers (ie: squid) often send them in separate headers instead, 740 * which in turn is forbidden by the http spec (multiple headers with 741 * the same name are only allowed for pure comma-separated lists, see 742 * rfc2616 sec 4.2). 743 * 744 * We support both approaches anyway 745 */ 746static int 747http_parse_authenticate(const char *cp, http_auth_challenges_t *cs) 748{ 749 int ret = -1; 750 http_header_lex_t lex; 751 char *key = malloc(strlen(cp) + 1); 752 char *value = malloc(strlen(cp) + 1); 753 char *buf = malloc(strlen(cp) + 1); 754 755 if (key == NULL || value == NULL || buf == NULL) { 756 fetch_syserr(); 757 goto out; 758 } 759 760 /* In any case we've seen the header and we set the valid bit */ 761 cs->valid = 1; 762 763 /* Need word first */ 764 lex = http_header_lex(&cp, key); 765 if (lex != HTTPHL_WORD) 766 goto out; 767 768 /* Loop on challenges */ 769 for (; cs->count < MAX_CHALLENGES; cs->count++) { 770 cs->challenges[cs->count] = 771 malloc(sizeof(http_auth_challenge_t)); 772 if (cs->challenges[cs->count] == NULL) { 773 fetch_syserr(); 774 goto out; 775 } 776 init_http_auth_challenge(cs->challenges[cs->count]); 777 if (!strcasecmp(key, "basic")) { 778 cs->challenges[cs->count]->scheme = HTTPAS_BASIC; 779 } else if (!strcasecmp(key, "digest")) { 780 cs->challenges[cs->count]->scheme = HTTPAS_DIGEST; 781 } else { 782 cs->challenges[cs->count]->scheme = HTTPAS_UNKNOWN; 783 /* 784 * Continue parsing as basic or digest may 785 * follow, and the syntax is the same for 786 * all. We'll just ignore this one when 787 * looking at the list 788 */ 789 } 790 791 /* Loop on attributes */ 792 for (;;) { 793 /* Key */ 794 lex = http_header_lex(&cp, key); 795 if (lex != HTTPHL_WORD) 796 goto out; 797 798 /* Equal sign */ 799 lex = http_header_lex(&cp, buf); 800 if (lex != '=') 801 goto out; 802 803 /* Value */ 804 lex = http_header_lex(&cp, value); 805 if (lex != HTTPHL_WORD && lex != HTTPHL_STRING) 806 goto out; 807 808 if (!strcasecmp(key, "realm")) 809 cs->challenges[cs->count]->realm = 810 strdup(value); 811 else if (!strcasecmp(key, "qop")) 812 cs->challenges[cs->count]->qop = 813 strdup(value); 814 else if (!strcasecmp(key, "nonce")) 815 cs->challenges[cs->count]->nonce = 816 strdup(value); 817 else if (!strcasecmp(key, "opaque")) 818 cs->challenges[cs->count]->opaque = 819 strdup(value); 820 else if (!strcasecmp(key, "algorithm")) 821 cs->challenges[cs->count]->algo = 822 strdup(value); 823 else if (!strcasecmp(key, "stale")) 824 cs->challenges[cs->count]->stale = 825 strcasecmp(value, "no"); 826 /* Else ignore unknown attributes */ 827 828 /* Comma or Next challenge or End */ 829 lex = http_header_lex(&cp, key); 830 /* 831 * If we get a word here, this is the beginning of the 832 * next challenge. Break the attributes loop 833 */ 834 if (lex == HTTPHL_WORD) 835 break; 836 837 if (lex == HTTPHL_END) { 838 /* End while looking for ',' is normal exit */ 839 cs->count++; 840 ret = 0; 841 goto out; 842 } 843 /* Anything else is an error */ 844 if (lex != ',') 845 goto out; 846 847 } /* End attributes loop */ 848 } /* End challenge loop */ 849 850 /* 851 * Challenges max count exceeded. This really can't happen 852 * with normal data, something's fishy -> error 853 */ 854 855out: 856 if (key) 857 free(key); 858 if (value) 859 free(value); 860 if (buf) 861 free(buf); 862 return (ret); 863} 864 865 866/* 867 * Parse a last-modified header 868 */ 869static int 870http_parse_mtime(const char *p, time_t *mtime) 871{ 872 char locale[64], *r; 873 struct tm tm; 874 875 strncpy(locale, setlocale(LC_TIME, NULL), sizeof(locale)); 876 setlocale(LC_TIME, "C"); 877 r = strptime(p, "%a, %d %b %Y %H:%M:%S GMT", &tm); 878 /* XXX should add support for date-2 and date-3 */ 879 setlocale(LC_TIME, locale); 880 if (r == NULL) 881 return (-1); 882 DEBUG(fprintf(stderr, "last modified: [%04d-%02d-%02d " 883 "%02d:%02d:%02d]\n", 884 tm.tm_year + 1900, tm.tm_mon + 1, tm.tm_mday, 885 tm.tm_hour, tm.tm_min, tm.tm_sec)); 886 *mtime = timegm(&tm); 887 return (0); 888} 889 890/* 891 * Parse a content-length header 892 */ 893static int 894http_parse_length(const char *p, off_t *length) 895{ 896 off_t len; 897 898 for (len = 0; *p && isdigit((unsigned char)*p); ++p) 899 len = len * 10 + (*p - '0'); 900 if (*p) 901 return (-1); 902 DEBUG(fprintf(stderr, "content length: [%lld]\n", 903 (long long)len)); 904 *length = len; 905 return (0); 906} 907 908/* 909 * Parse a content-range header 910 */ 911static int 912http_parse_range(const char *p, off_t *offset, off_t *length, off_t *size) 913{ 914 off_t first, last, len; 915 916 if (strncasecmp(p, "bytes ", 6) != 0) 917 return (-1); 918 p += 6; 919 if (*p == '*') { 920 first = last = -1; 921 ++p; 922 } else { 923 for (first = 0; *p && isdigit((unsigned char)*p); ++p) 924 first = first * 10 + *p - '0'; 925 if (*p != '-') 926 return (-1); 927 for (last = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 928 last = last * 10 + *p - '0'; 929 } 930 if (first > last || *p != '/') 931 return (-1); 932 for (len = 0, ++p; *p && isdigit((unsigned char)*p); ++p) 933 len = len * 10 + *p - '0'; 934 if (*p || len < last - first + 1) 935 return (-1); 936 if (first == -1) { 937 DEBUG(fprintf(stderr, "content range: [*/%lld]\n", 938 (long long)len)); 939 *length = 0; 940 } else { 941 DEBUG(fprintf(stderr, "content range: [%lld-%lld/%lld]\n", 942 (long long)first, (long long)last, (long long)len)); 943 *length = last - first + 1; 944 } 945 *offset = first; 946 *size = len; 947 return (0); 948} 949 950 951/***************************************************************************** 952 * Helper functions for authorization 953 */ 954 955/* 956 * Base64 encoding 957 */ 958static char * 959http_base64(const char *src) 960{ 961 static const char base64[] = 962 "ABCDEFGHIJKLMNOPQRSTUVWXYZ" 963 "abcdefghijklmnopqrstuvwxyz" 964 "0123456789+/"; 965 char *str, *dst; 966 size_t l; 967 int t, r; 968 969 l = strlen(src); 970 if ((str = malloc(((l + 2) / 3) * 4 + 1)) == NULL) 971 return (NULL); 972 dst = str; 973 r = 0; 974 975 while (l >= 3) { 976 t = (src[0] << 16) | (src[1] << 8) | src[2]; 977 dst[0] = base64[(t >> 18) & 0x3f]; 978 dst[1] = base64[(t >> 12) & 0x3f]; 979 dst[2] = base64[(t >> 6) & 0x3f]; 980 dst[3] = base64[(t >> 0) & 0x3f]; 981 src += 3; l -= 3; 982 dst += 4; r += 4; 983 } 984 985 switch (l) { 986 case 2: 987 t = (src[0] << 16) | (src[1] << 8); 988 dst[0] = base64[(t >> 18) & 0x3f]; 989 dst[1] = base64[(t >> 12) & 0x3f]; 990 dst[2] = base64[(t >> 6) & 0x3f]; 991 dst[3] = '='; 992 dst += 4; 993 r += 4; 994 break; 995 case 1: 996 t = src[0] << 16; 997 dst[0] = base64[(t >> 18) & 0x3f]; 998 dst[1] = base64[(t >> 12) & 0x3f]; 999 dst[2] = dst[3] = '='; 1000 dst += 4; 1001 r += 4; 1002 break; 1003 case 0: 1004 break; 1005 } 1006 1007 *dst = 0; 1008 return (str); 1009} 1010 1011 1012/* 1013 * Extract authorization parameters from environment value. 1014 * The value is like scheme:realm:user:pass 1015 */ 1016typedef struct { 1017 char *scheme; 1018 char *realm; 1019 char *user; 1020 char *password; 1021} http_auth_params_t; 1022 1023static void 1024init_http_auth_params(http_auth_params_t *s) 1025{ 1026 s->scheme = s->realm = s->user = s->password = 0; 1027} 1028 1029static void 1030clean_http_auth_params(http_auth_params_t *s) 1031{ 1032 if (s->scheme) 1033 free(s->scheme); 1034 if (s->realm) 1035 free(s->realm); 1036 if (s->user) 1037 free(s->user); 1038 if (s->password) 1039 free(s->password); 1040 init_http_auth_params(s); 1041} 1042 1043static int 1044http_authfromenv(const char *p, http_auth_params_t *parms) 1045{ 1046 int ret = -1; 1047 char *v, *ve; 1048 char *str = strdup(p); 1049 1050 if (str == NULL) { 1051 fetch_syserr(); 1052 return (-1); 1053 } 1054 v = str; 1055 1056 if ((ve = strchr(v, ':')) == NULL) 1057 goto out; 1058 1059 *ve = 0; 1060 if ((parms->scheme = strdup(v)) == NULL) { 1061 fetch_syserr(); 1062 goto out; 1063 } 1064 v = ve + 1; 1065 1066 if ((ve = strchr(v, ':')) == NULL) 1067 goto out; 1068 1069 *ve = 0; 1070 if ((parms->realm = 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->user = strdup(v)) == NULL) { 1081 fetch_syserr(); 1082 goto out; 1083 } 1084 v = ve + 1; 1085 1086 1087 if ((parms->password = strdup(v)) == NULL) { 1088 fetch_syserr(); 1089 goto out; 1090 } 1091 ret = 0; 1092out: 1093 if (ret == -1) 1094 clean_http_auth_params(parms); 1095 if (str) 1096 free(str); 1097 return (ret); 1098} 1099 1100 1101/* 1102 * Digest response: the code to compute the digest is taken from the 1103 * sample implementation in RFC2616 1104 */ 1105#define IN const 1106#define OUT 1107 1108#define HASHLEN 16 1109typedef char HASH[HASHLEN]; 1110#define HASHHEXLEN 32 1111typedef char HASHHEX[HASHHEXLEN+1]; 1112 1113static const char *hexchars = "0123456789abcdef"; 1114static void 1115CvtHex(IN HASH Bin, OUT HASHHEX Hex) 1116{ 1117 unsigned short i; 1118 unsigned char j; 1119 1120 for (i = 0; i < HASHLEN; i++) { 1121 j = (Bin[i] >> 4) & 0xf; 1122 Hex[i*2] = hexchars[j]; 1123 j = Bin[i] & 0xf; 1124 Hex[i*2+1] = hexchars[j]; 1125 }; 1126 Hex[HASHHEXLEN] = '\0'; 1127}; 1128 1129/* calculate H(A1) as per spec */ 1130static void 1131DigestCalcHA1( 1132 IN char * pszAlg, 1133 IN char * pszUserName, 1134 IN char * pszRealm, 1135 IN char * pszPassword, 1136 IN char * pszNonce, 1137 IN char * pszCNonce, 1138 OUT HASHHEX SessionKey 1139 ) 1140{ 1141 MD5_CTX Md5Ctx; 1142 HASH HA1; 1143 1144 MD5Init(&Md5Ctx); 1145 MD5Update(&Md5Ctx, pszUserName, strlen(pszUserName)); 1146 MD5Update(&Md5Ctx, ":", 1); 1147 MD5Update(&Md5Ctx, pszRealm, strlen(pszRealm)); 1148 MD5Update(&Md5Ctx, ":", 1); 1149 MD5Update(&Md5Ctx, pszPassword, strlen(pszPassword)); 1150 MD5Final(HA1, &Md5Ctx); 1151 if (strcasecmp(pszAlg, "md5-sess") == 0) { 1152 1153 MD5Init(&Md5Ctx); 1154 MD5Update(&Md5Ctx, HA1, HASHLEN); 1155 MD5Update(&Md5Ctx, ":", 1); 1156 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce)); 1157 MD5Update(&Md5Ctx, ":", 1); 1158 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce)); 1159 MD5Final(HA1, &Md5Ctx); 1160 }; 1161 CvtHex(HA1, SessionKey); 1162} 1163 1164/* calculate request-digest/response-digest as per HTTP Digest spec */ 1165static void 1166DigestCalcResponse( 1167 IN HASHHEX HA1, /* H(A1) */ 1168 IN char * pszNonce, /* nonce from server */ 1169 IN char * pszNonceCount, /* 8 hex digits */ 1170 IN char * pszCNonce, /* client nonce */ 1171 IN char * pszQop, /* qop-value: "", "auth", "auth-int" */ 1172 IN char * pszMethod, /* method from the request */ 1173 IN char * pszDigestUri, /* requested URL */ 1174 IN HASHHEX HEntity, /* H(entity body) if qop="auth-int" */ 1175 OUT HASHHEX Response /* request-digest or response-digest */ 1176 ) 1177{ 1178/* DEBUG(fprintf(stderr, 1179 "Calc: HA1[%s] Nonce[%s] qop[%s] method[%s] URI[%s]\n", 1180 HA1, pszNonce, pszQop, pszMethod, pszDigestUri));*/ 1181 MD5_CTX Md5Ctx; 1182 HASH HA2; 1183 HASH RespHash; 1184 HASHHEX HA2Hex; 1185 1186 // calculate H(A2) 1187 MD5Init(&Md5Ctx); 1188 MD5Update(&Md5Ctx, pszMethod, strlen(pszMethod)); 1189 MD5Update(&Md5Ctx, ":", 1); 1190 MD5Update(&Md5Ctx, pszDigestUri, strlen(pszDigestUri)); 1191 if (strcasecmp(pszQop, "auth-int") == 0) { 1192 MD5Update(&Md5Ctx, ":", 1); 1193 MD5Update(&Md5Ctx, HEntity, HASHHEXLEN); 1194 }; 1195 MD5Final(HA2, &Md5Ctx); 1196 CvtHex(HA2, HA2Hex); 1197 1198 // calculate response 1199 MD5Init(&Md5Ctx); 1200 MD5Update(&Md5Ctx, HA1, HASHHEXLEN); 1201 MD5Update(&Md5Ctx, ":", 1); 1202 MD5Update(&Md5Ctx, pszNonce, strlen(pszNonce)); 1203 MD5Update(&Md5Ctx, ":", 1); 1204 if (*pszQop) { 1205 MD5Update(&Md5Ctx, pszNonceCount, strlen(pszNonceCount)); 1206 MD5Update(&Md5Ctx, ":", 1); 1207 MD5Update(&Md5Ctx, pszCNonce, strlen(pszCNonce)); 1208 MD5Update(&Md5Ctx, ":", 1); 1209 MD5Update(&Md5Ctx, pszQop, strlen(pszQop)); 1210 MD5Update(&Md5Ctx, ":", 1); 1211 }; 1212 MD5Update(&Md5Ctx, HA2Hex, HASHHEXLEN); 1213 MD5Final(RespHash, &Md5Ctx); 1214 CvtHex(RespHash, Response); 1215} 1216 1217/* 1218 * Generate/Send a Digest authorization header 1219 * This looks like: [Proxy-]Authorization: credentials 1220 * 1221 * credentials = "Digest" digest-response 1222 * digest-response = 1#( username | realm | nonce | digest-uri 1223 * | response | [ algorithm ] | [cnonce] | 1224 * [opaque] | [message-qop] | 1225 * [nonce-count] | [auth-param] ) 1226 * username = "username" "=" username-value 1227 * username-value = quoted-string 1228 * digest-uri = "uri" "=" digest-uri-value 1229 * digest-uri-value = request-uri ; As specified by HTTP/1.1 1230 * message-qop = "qop" "=" qop-value 1231 * cnonce = "cnonce" "=" cnonce-value 1232 * cnonce-value = nonce-value 1233 * nonce-count = "nc" "=" nc-value 1234 * nc-value = 8LHEX 1235 * response = "response" "=" request-digest 1236 * request-digest = <"> 32LHEX <"> 1237 */ 1238static int 1239http_digest_auth(conn_t *conn, const char *hdr, http_auth_challenge_t *c, 1240 http_auth_params_t *parms, struct url *url) 1241{ 1242 int r; 1243 char noncecount[10]; 1244 char cnonce[40]; 1245 char *options = 0; 1246 1247 if (!c->realm || !c->nonce) { 1248 DEBUG(fprintf(stderr, "realm/nonce not set in challenge\n")); 1249 return(-1); 1250 } 1251 if (!c->algo) 1252 c->algo = strdup(""); 1253 1254 if (asprintf(&options, "%s%s%s%s", 1255 *c->algo? ",algorithm=" : "", c->algo, 1256 c->opaque? ",opaque=" : "", c->opaque?c->opaque:"")== -1) 1257 return (-1); 1258 1259 if (!c->qop) { 1260 c->qop = strdup(""); 1261 *noncecount = 0; 1262 *cnonce = 0; 1263 } else { 1264 c->nc++; 1265 sprintf(noncecount, "%08x", c->nc); 1266 /* We don't try very hard with the cnonce ... */ 1267 sprintf(cnonce, "%x%lx", getpid(), (unsigned long)time(0)); 1268 } 1269 1270 HASHHEX HA1; 1271 DigestCalcHA1(c->algo, parms->user, c->realm, 1272 parms->password, c->nonce, cnonce, HA1); 1273 DEBUG(fprintf(stderr, "HA1: [%s]\n", HA1)); 1274 HASHHEX digest; 1275 DigestCalcResponse(HA1, c->nonce, noncecount, cnonce, c->qop, 1276 "GET", url->doc, "", digest); 1277 1278 if (c->qop[0]) { 1279 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\"," 1280 "nonce=\"%s\",uri=\"%s\",response=\"%s\"," 1281 "qop=\"auth\", cnonce=\"%s\", nc=%s%s", 1282 hdr, parms->user, c->realm, 1283 c->nonce, url->doc, digest, 1284 cnonce, noncecount, options); 1285 } else { 1286 r = http_cmd(conn, "%s: Digest username=\"%s\",realm=\"%s\"," 1287 "nonce=\"%s\",uri=\"%s\",response=\"%s\"%s", 1288 hdr, parms->user, c->realm, 1289 c->nonce, url->doc, digest, options); 1290 } 1291 if (options) 1292 free(options); 1293 return (r); 1294} 1295 1296/* 1297 * Encode username and password 1298 */ 1299static int 1300http_basic_auth(conn_t *conn, const char *hdr, const char *usr, const char *pwd) 1301{ 1302 char *upw, *auth; 1303 int r; 1304 1305 DEBUG(fprintf(stderr, "basic: usr: [%s]\n", usr)); 1306 DEBUG(fprintf(stderr, "basic: pwd: [%s]\n", pwd)); 1307 if (asprintf(&upw, "%s:%s", usr, pwd) == -1) 1308 return (-1); 1309 auth = http_base64(upw); 1310 free(upw); 1311 if (auth == NULL) 1312 return (-1); 1313 r = http_cmd(conn, "%s: Basic %s", hdr, auth); 1314 free(auth); 1315 return (r); 1316} 1317 1318/* 1319 * Chose the challenge to answer and call the appropriate routine to 1320 * produce the header. 1321 */ 1322static int 1323http_authorize(conn_t *conn, const char *hdr, http_auth_challenges_t *cs, 1324 http_auth_params_t *parms, struct url *url) 1325{ 1326 http_auth_challenge_t *basic = NULL; 1327 http_auth_challenge_t *digest = NULL; 1328 int i; 1329 1330 /* If user or pass are null we're not happy */ 1331 if (!parms->user || !parms->password) { 1332 DEBUG(fprintf(stderr, "NULL usr or pass\n")); 1333 return (-1); 1334 } 1335 1336 /* Look for a Digest and a Basic challenge */ 1337 for (i = 0; i < cs->count; i++) { 1338 if (cs->challenges[i]->scheme == HTTPAS_BASIC) 1339 basic = cs->challenges[i]; 1340 if (cs->challenges[i]->scheme == HTTPAS_DIGEST) 1341 digest = cs->challenges[i]; 1342 } 1343 1344 /* Error if "Digest" was specified and there is no Digest challenge */ 1345 if (!digest && (parms->scheme && 1346 !strcasecmp(parms->scheme, "digest"))) { 1347 DEBUG(fprintf(stderr, 1348 "Digest auth in env, not supported by peer\n")); 1349 return (-1); 1350 } 1351 /* 1352 * If "basic" was specified in the environment, or there is no Digest 1353 * challenge, do the basic thing. Don't need a challenge for this, 1354 * so no need to check basic!=NULL 1355 */ 1356 if (!digest || (parms->scheme && !strcasecmp(parms->scheme,"basic"))) 1357 return (http_basic_auth(conn,hdr,parms->user,parms->password)); 1358 1359 /* Else, prefer digest. We just checked that it's not NULL */ 1360 return (http_digest_auth(conn, hdr, digest, parms, url)); 1361} 1362 1363/***************************************************************************** 1364 * Helper functions for connecting to a server or proxy 1365 */ 1366 1367/* 1368 * Connect to the correct HTTP server or proxy. 1369 */ 1370static conn_t * 1371http_connect(struct url *URL, struct url *purl, const char *flags) 1372{ 1373 conn_t *conn; 1374 int verbose; 1375 int af, val; 1376 1377#ifdef INET6 1378 af = AF_UNSPEC; 1379#else 1380 af = AF_INET; 1381#endif 1382 1383 verbose = CHECK_FLAG('v'); 1384 if (CHECK_FLAG('4')) 1385 af = AF_INET; 1386#ifdef INET6 1387 else if (CHECK_FLAG('6')) 1388 af = AF_INET6; 1389#endif 1390 1391 if (purl && strcasecmp(URL->scheme, SCHEME_HTTPS) != 0) { 1392 URL = purl; 1393 } else if (strcasecmp(URL->scheme, SCHEME_FTP) == 0) { 1394 /* can't talk http to an ftp server */ 1395 /* XXX should set an error code */ 1396 return (NULL); 1397 } 1398 1399 if ((conn = fetch_connect(URL->host, URL->port, af, verbose)) == NULL) 1400 /* fetch_connect() has already set an error code */ 1401 return (NULL); 1402 if (strcasecmp(URL->scheme, SCHEME_HTTPS) == 0 && 1403 fetch_ssl(conn, verbose) == -1) { 1404 fetch_close(conn); 1405 /* grrr */ 1406 errno = EAUTH; 1407 fetch_syserr(); 1408 return (NULL); 1409 } 1410 1411 val = 1; 1412 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, sizeof(val)); 1413 1414 return (conn); 1415} 1416 1417static struct url * 1418http_get_proxy(struct url * url, const char *flags) 1419{ 1420 struct url *purl; 1421 char *p; 1422 1423 if (flags != NULL && strchr(flags, 'd') != NULL) 1424 return (NULL); 1425 if (fetch_no_proxy_match(url->host)) 1426 return (NULL); 1427 if (((p = getenv("HTTP_PROXY")) || (p = getenv("http_proxy"))) && 1428 *p && (purl = fetchParseURL(p))) { 1429 if (!*purl->scheme) 1430 strcpy(purl->scheme, SCHEME_HTTP); 1431 if (!purl->port) 1432 purl->port = fetch_default_proxy_port(purl->scheme); 1433 if (strcasecmp(purl->scheme, SCHEME_HTTP) == 0) 1434 return (purl); 1435 fetchFreeURL(purl); 1436 } 1437 return (NULL); 1438} 1439 1440static void 1441http_print_html(FILE *out, FILE *in) 1442{ 1443 size_t len; 1444 char *line, *p, *q; 1445 int comment, tag; 1446 1447 comment = tag = 0; 1448 while ((line = fgetln(in, &len)) != NULL) { 1449 while (len && isspace((unsigned char)line[len - 1])) 1450 --len; 1451 for (p = q = line; q < line + len; ++q) { 1452 if (comment && *q == '-') { 1453 if (q + 2 < line + len && 1454 strcmp(q, "-->") == 0) { 1455 tag = comment = 0; 1456 q += 2; 1457 } 1458 } else if (tag && !comment && *q == '>') { 1459 p = q + 1; 1460 tag = 0; 1461 } else if (!tag && *q == '<') { 1462 if (q > p) 1463 fwrite(p, q - p, 1, out); 1464 tag = 1; 1465 if (q + 3 < line + len && 1466 strcmp(q, "<!--") == 0) { 1467 comment = 1; 1468 q += 3; 1469 } 1470 } 1471 } 1472 if (!tag && q > p) 1473 fwrite(p, q - p, 1, out); 1474 fputc('\n', out); 1475 } 1476} 1477 1478 1479/***************************************************************************** 1480 * Core 1481 */ 1482 1483/* 1484 * Send a request and process the reply 1485 * 1486 * XXX This function is way too long, the do..while loop should be split 1487 * XXX off into a separate function. 1488 */ 1489FILE * 1490http_request(struct url *URL, const char *op, struct url_stat *us, 1491 struct url *purl, const char *flags) 1492{ 1493 char timebuf[80]; 1494 char hbuf[MAXHOSTNAMELEN + 7], *host; 1495 conn_t *conn; 1496 struct url *url, *new; 1497 int chunked, direct, ims, noredirect, verbose; 1498 int e, i, n, val; 1499 off_t offset, clength, length, size; 1500 time_t mtime; 1501 const char *p; 1502 FILE *f; 1503 hdr_t h; 1504 struct tm *timestruct; 1505 http_headerbuf_t headerbuf; 1506 http_auth_challenges_t server_challenges; 1507 http_auth_challenges_t proxy_challenges; 1508 1509 /* The following calls don't allocate anything */ 1510 init_http_headerbuf(&headerbuf); 1511 init_http_auth_challenges(&server_challenges); 1512 init_http_auth_challenges(&proxy_challenges); 1513 1514 direct = CHECK_FLAG('d'); 1515 noredirect = CHECK_FLAG('A'); 1516 verbose = CHECK_FLAG('v'); 1517 ims = CHECK_FLAG('i'); 1518 1519 if (direct && purl) { 1520 fetchFreeURL(purl); 1521 purl = NULL; 1522 } 1523 1524 /* try the provided URL first */ 1525 url = URL; 1526 1527 /* if the A flag is set, we only get one try */ 1528 n = noredirect ? 1 : MAX_REDIRECT; 1529 i = 0; 1530 1531 e = HTTP_PROTOCOL_ERROR; 1532 do { 1533 new = NULL; 1534 chunked = 0; 1535 offset = 0; 1536 clength = -1; 1537 length = -1; 1538 size = -1; 1539 mtime = 0; 1540 1541 /* check port */ 1542 if (!url->port) 1543 url->port = fetch_default_port(url->scheme); 1544 1545 /* were we redirected to an FTP URL? */ 1546 if (purl == NULL && strcmp(url->scheme, SCHEME_FTP) == 0) { 1547 if (strcmp(op, "GET") == 0) 1548 return (ftp_request(url, "RETR", us, purl, flags)); 1549 else if (strcmp(op, "HEAD") == 0) 1550 return (ftp_request(url, "STAT", us, purl, flags)); 1551 } 1552 1553 /* connect to server or proxy */ 1554 if ((conn = http_connect(url, purl, flags)) == NULL) 1555 goto ouch; 1556 1557 host = url->host; 1558#ifdef INET6 1559 if (strchr(url->host, ':')) { 1560 snprintf(hbuf, sizeof(hbuf), "[%s]", url->host); 1561 host = hbuf; 1562 } 1563#endif 1564 if (url->port != fetch_default_port(url->scheme)) { 1565 if (host != hbuf) { 1566 strcpy(hbuf, host); 1567 host = hbuf; 1568 } 1569 snprintf(hbuf + strlen(hbuf), 1570 sizeof(hbuf) - strlen(hbuf), ":%d", url->port); 1571 } 1572 1573 /* send request */ 1574 if (verbose) 1575 fetch_info("requesting %s://%s%s", 1576 url->scheme, host, url->doc); 1577 if (purl) { 1578 http_cmd(conn, "%s %s://%s%s HTTP/1.1", 1579 op, url->scheme, host, url->doc); 1580 } else { 1581 http_cmd(conn, "%s %s HTTP/1.1", 1582 op, url->doc); 1583 } 1584 1585 if (ims && url->ims_time) { 1586 timestruct = gmtime((time_t *)&url->ims_time); 1587 (void)strftime(timebuf, 80, "%a, %d %b %Y %T GMT", 1588 timestruct); 1589 if (verbose) 1590 fetch_info("If-Modified-Since: %s", timebuf); 1591 http_cmd(conn, "If-Modified-Since: %s", timebuf); 1592 } 1593 /* virtual host */ 1594 http_cmd(conn, "Host: %s", host); 1595 1596 /* 1597 * Proxy authorization: we only send auth after we received 1598 * a 407 error. We do not first try basic anyway (changed 1599 * when support was added for digest-auth) 1600 */ 1601 if (purl && proxy_challenges.valid) { 1602 http_auth_params_t aparams; 1603 init_http_auth_params(&aparams); 1604 if (*purl->user || *purl->pwd) { 1605 aparams.user = purl->user ? 1606 strdup(purl->user) : strdup(""); 1607 aparams.password = purl->pwd? 1608 strdup(purl->pwd) : strdup(""); 1609 } else if ((p = getenv("HTTP_PROXY_AUTH")) != NULL && 1610 *p != '\0') { 1611 if (http_authfromenv(p, &aparams) < 0) { 1612 http_seterr(HTTP_NEED_PROXY_AUTH); 1613 goto ouch; 1614 } 1615 } 1616 http_authorize(conn, "Proxy-Authorization", 1617 &proxy_challenges, &aparams, url); 1618 clean_http_auth_params(&aparams); 1619 } 1620 1621 /* 1622 * Server authorization: we never send "a priori" 1623 * Basic auth, which used to be done if user/pass were 1624 * set in the url. This would be weird because we'd send the 1625 * password in the clear even if Digest is finally to be 1626 * used (it would have made more sense for the 1627 * pre-digest version to do this when Basic was specified 1628 * in the environment) 1629 */ 1630 if (server_challenges.valid) { 1631 http_auth_params_t aparams; 1632 init_http_auth_params(&aparams); 1633 if (*url->user || *url->pwd) { 1634 aparams.user = url->user ? 1635 strdup(url->user) : strdup(""); 1636 aparams.password = url->pwd ? 1637 strdup(url->pwd) : strdup(""); 1638 } else if ((p = getenv("HTTP_AUTH")) != NULL && 1639 *p != '\0') { 1640 if (http_authfromenv(p, &aparams) < 0) { 1641 http_seterr(HTTP_NEED_AUTH); 1642 goto ouch; 1643 } 1644 } else if (fetchAuthMethod && 1645 fetchAuthMethod(url) == 0) { 1646 aparams.user = url->user ? 1647 strdup(url->user) : strdup(""); 1648 aparams.password = url->pwd ? 1649 strdup(url->pwd) : strdup(""); 1650 } else { 1651 http_seterr(HTTP_NEED_AUTH); 1652 goto ouch; 1653 } 1654 http_authorize(conn, "Authorization", 1655 &server_challenges, &aparams, url); 1656 clean_http_auth_params(&aparams); 1657 } 1658 1659 /* other headers */ 1660 if ((p = getenv("HTTP_REFERER")) != NULL && *p != '\0') { 1661 if (strcasecmp(p, "auto") == 0) 1662 http_cmd(conn, "Referer: %s://%s%s", 1663 url->scheme, host, url->doc); 1664 else 1665 http_cmd(conn, "Referer: %s", p); 1666 } 1667 if ((p = getenv("HTTP_USER_AGENT")) != NULL && *p != '\0') 1668 http_cmd(conn, "User-Agent: %s", p); 1669 else 1670 http_cmd(conn, "User-Agent: %s " _LIBFETCH_VER, getprogname()); 1671 if (url->offset > 0) 1672 http_cmd(conn, "Range: bytes=%lld-", (long long)url->offset); 1673 http_cmd(conn, "Connection: close"); 1674 http_cmd(conn, ""); 1675 1676 /* 1677 * Force the queued request to be dispatched. Normally, one 1678 * would do this with shutdown(2) but squid proxies can be 1679 * configured to disallow such half-closed connections. To 1680 * be compatible with such configurations, fiddle with socket 1681 * options to force the pending data to be written. 1682 */ 1683 val = 0; 1684 setsockopt(conn->sd, IPPROTO_TCP, TCP_NOPUSH, &val, 1685 sizeof(val)); 1686 val = 1; 1687 setsockopt(conn->sd, IPPROTO_TCP, TCP_NODELAY, &val, 1688 sizeof(val)); 1689 1690 /* get reply */ 1691 switch (http_get_reply(conn)) { 1692 case HTTP_OK: 1693 case HTTP_PARTIAL: 1694 case HTTP_NOT_MODIFIED: 1695 /* fine */ 1696 break; 1697 case HTTP_MOVED_PERM: 1698 case HTTP_MOVED_TEMP: 1699 case HTTP_SEE_OTHER: 1700 /* 1701 * Not so fine, but we still have to read the 1702 * headers to get the new location. 1703 */ 1704 break; 1705 case HTTP_NEED_AUTH: 1706 if (server_challenges.valid) { 1707 /* 1708 * We already sent out authorization code, 1709 * so there's nothing more we can do. 1710 */ 1711 http_seterr(conn->err); 1712 goto ouch; 1713 } 1714 /* try again, but send the password this time */ 1715 if (verbose) 1716 fetch_info("server requires authorization"); 1717 break; 1718 case HTTP_NEED_PROXY_AUTH: 1719 if (proxy_challenges.valid) { 1720 /* 1721 * We already sent our proxy 1722 * authorization code, so there's 1723 * nothing more we can do. */ 1724 http_seterr(conn->err); 1725 goto ouch; 1726 } 1727 /* try again, but send the password this time */ 1728 if (verbose) 1729 fetch_info("proxy requires authorization"); 1730 break; 1731 case HTTP_BAD_RANGE: 1732 /* 1733 * This can happen if we ask for 0 bytes because 1734 * we already have the whole file. Consider this 1735 * a success for now, and check sizes later. 1736 */ 1737 break; 1738 case HTTP_PROTOCOL_ERROR: 1739 /* fall through */ 1740 case -1: 1741 fetch_syserr(); 1742 goto ouch; 1743 default: 1744 http_seterr(conn->err); 1745 if (!verbose) 1746 goto ouch; 1747 /* fall through so we can get the full error message */ 1748 } 1749 1750 /* get headers. http_next_header expects one line readahead */ 1751 if (fetch_getln(conn) == -1) { 1752 fetch_syserr(); 1753 goto ouch; 1754 } 1755 do { 1756 switch ((h = http_next_header(conn, &headerbuf, &p))) { 1757 case hdr_syserror: 1758 fetch_syserr(); 1759 goto ouch; 1760 case hdr_error: 1761 http_seterr(HTTP_PROTOCOL_ERROR); 1762 goto ouch; 1763 case hdr_content_length: 1764 http_parse_length(p, &clength); 1765 break; 1766 case hdr_content_range: 1767 http_parse_range(p, &offset, &length, &size); 1768 break; 1769 case hdr_last_modified: 1770 http_parse_mtime(p, &mtime); 1771 break; 1772 case hdr_location: 1773 if (!HTTP_REDIRECT(conn->err)) 1774 break; 1775 if (new) 1776 free(new); 1777 if (verbose) 1778 fetch_info("%d redirect to %s", conn->err, p); 1779 if (*p == '/') 1780 /* absolute path */ 1781 new = fetchMakeURL(url->scheme, url->host, url->port, p, 1782 url->user, url->pwd); 1783 else 1784 new = fetchParseURL(p); 1785 if (new == NULL) { 1786 /* XXX should set an error code */ 1787 DEBUG(fprintf(stderr, "failed to parse new URL\n")); 1788 goto ouch; 1789 } 1790 1791 /* Only copy credentials if the host matches */ 1792 if (!strcmp(new->host, url->host) && !*new->user && !*new->pwd) { 1793 strcpy(new->user, url->user); 1794 strcpy(new->pwd, url->pwd); 1795 } 1796 new->offset = url->offset; 1797 new->length = url->length; 1798 break; 1799 case hdr_transfer_encoding: 1800 /* XXX weak test*/ 1801 chunked = (strcasecmp(p, "chunked") == 0); 1802 break; 1803 case hdr_www_authenticate: 1804 if (conn->err != HTTP_NEED_AUTH) 1805 break; 1806 if (http_parse_authenticate(p, &server_challenges) == 0) 1807 ++n; 1808 break; 1809 case hdr_proxy_authenticate: 1810 if (conn->err != HTTP_NEED_PROXY_AUTH) 1811 break; 1812 if (http_parse_authenticate(p, &proxy_challenges) == 0) 1813 ++n; 1814 break; 1815 case hdr_end: 1816 /* fall through */ 1817 case hdr_unknown: 1818 /* ignore */ 1819 break; 1820 } 1821 } while (h > hdr_end); 1822 1823 /* we need to provide authentication */ 1824 if (conn->err == HTTP_NEED_AUTH || 1825 conn->err == HTTP_NEED_PROXY_AUTH) { 1826 e = conn->err; 1827 if ((conn->err == HTTP_NEED_AUTH && 1828 !server_challenges.valid) || 1829 (conn->err == HTTP_NEED_PROXY_AUTH && 1830 !proxy_challenges.valid)) { 1831 /* 401/7 but no www/proxy-authenticate ?? */ 1832 DEBUG(fprintf(stderr, "401/7 and no auth header\n")); 1833 goto ouch; 1834 } 1835 fetch_close(conn); 1836 conn = NULL; 1837 continue; 1838 } 1839 1840 /* requested range not satisfiable */ 1841 if (conn->err == HTTP_BAD_RANGE) { 1842 if (url->offset == size && url->length == 0) { 1843 /* asked for 0 bytes; fake it */ 1844 offset = url->offset; 1845 clength = -1; 1846 conn->err = HTTP_OK; 1847 break; 1848 } else { 1849 http_seterr(conn->err); 1850 goto ouch; 1851 } 1852 } 1853 1854 /* we have a hit or an error */ 1855 if (conn->err == HTTP_OK 1856 || conn->err == HTTP_NOT_MODIFIED 1857 || conn->err == HTTP_PARTIAL 1858 || HTTP_ERROR(conn->err)) 1859 break; 1860 1861 /* all other cases: we got a redirect */ 1862 e = conn->err; 1863 clean_http_auth_challenges(&server_challenges); 1864 fetch_close(conn); 1865 conn = NULL; 1866 if (!new) { 1867 DEBUG(fprintf(stderr, "redirect with no new location\n")); 1868 break; 1869 } 1870 if (url != URL) 1871 fetchFreeURL(url); 1872 url = new; 1873 } while (++i < n); 1874 1875 /* we failed, or ran out of retries */ 1876 if (conn == NULL) { 1877 http_seterr(e); 1878 goto ouch; 1879 } 1880 1881 DEBUG(fprintf(stderr, "offset %lld, length %lld," 1882 " size %lld, clength %lld\n", 1883 (long long)offset, (long long)length, 1884 (long long)size, (long long)clength)); 1885 1886 if (conn->err == HTTP_NOT_MODIFIED) { 1887 http_seterr(HTTP_NOT_MODIFIED); 1888 return (NULL); 1889 } 1890 1891 /* check for inconsistencies */ 1892 if (clength != -1 && length != -1 && clength != length) { 1893 http_seterr(HTTP_PROTOCOL_ERROR); 1894 goto ouch; 1895 } 1896 if (clength == -1) 1897 clength = length; 1898 if (clength != -1) 1899 length = offset + clength; 1900 if (length != -1 && size != -1 && length != size) { 1901 http_seterr(HTTP_PROTOCOL_ERROR); 1902 goto ouch; 1903 } 1904 if (size == -1) 1905 size = length; 1906 1907 /* fill in stats */ 1908 if (us) { 1909 us->size = size; 1910 us->atime = us->mtime = mtime; 1911 } 1912 1913 /* too far? */ 1914 if (URL->offset > 0 && offset > URL->offset) { 1915 http_seterr(HTTP_PROTOCOL_ERROR); 1916 goto ouch; 1917 } 1918 1919 /* report back real offset and size */ 1920 URL->offset = offset; 1921 URL->length = clength; 1922 1923 /* wrap it up in a FILE */ 1924 if ((f = http_funopen(conn, chunked)) == NULL) { 1925 fetch_syserr(); 1926 goto ouch; 1927 } 1928 1929 if (url != URL) 1930 fetchFreeURL(url); 1931 if (purl) 1932 fetchFreeURL(purl); 1933 1934 if (HTTP_ERROR(conn->err)) { 1935 http_print_html(stderr, f); 1936 fclose(f); 1937 f = NULL; 1938 } 1939 clean_http_headerbuf(&headerbuf); 1940 clean_http_auth_challenges(&server_challenges); 1941 clean_http_auth_challenges(&proxy_challenges); 1942 return (f); 1943 1944ouch: 1945 if (url != URL) 1946 fetchFreeURL(url); 1947 if (purl) 1948 fetchFreeURL(purl); 1949 if (conn != NULL) 1950 fetch_close(conn); 1951 clean_http_headerbuf(&headerbuf); 1952 clean_http_auth_challenges(&server_challenges); 1953 clean_http_auth_challenges(&proxy_challenges); 1954 return (NULL); 1955} 1956 1957 1958/***************************************************************************** 1959 * Entry points 1960 */ 1961 1962/* 1963 * Retrieve and stat a file by HTTP 1964 */ 1965FILE * 1966fetchXGetHTTP(struct url *URL, struct url_stat *us, const char *flags) 1967{ 1968 return (http_request(URL, "GET", us, http_get_proxy(URL, flags), flags)); 1969} 1970 1971/* 1972 * Retrieve a file by HTTP 1973 */ 1974FILE * 1975fetchGetHTTP(struct url *URL, const char *flags) 1976{ 1977 return (fetchXGetHTTP(URL, NULL, flags)); 1978} 1979 1980/* 1981 * Store a file by HTTP 1982 */ 1983FILE * 1984fetchPutHTTP(struct url *URL __unused, const char *flags __unused) 1985{ 1986 warnx("fetchPutHTTP(): not implemented"); 1987 return (NULL); 1988} 1989 1990/* 1991 * Get an HTTP document's metadata 1992 */ 1993int 1994fetchStatHTTP(struct url *URL, struct url_stat *us, const char *flags) 1995{ 1996 FILE *f; 1997 1998 f = http_request(URL, "HEAD", us, http_get_proxy(URL, flags), flags); 1999 if (f == NULL) 2000 return (-1); 2001 fclose(f); 2002 return (0); 2003} 2004 2005/* 2006 * List a directory 2007 */ 2008struct url_ent * 2009fetchListHTTP(struct url *url __unused, const char *flags __unused) 2010{ 2011 warnx("fetchListHTTP(): not implemented"); 2012 return (NULL); 2013} 2014