src/url.cc

   1
   2 /*
   3  * $Id: url.cc,v 1.165 2008/02/03 10:00:30 amosjeffries Exp $
   4  *
   5  * DEBUG: section 23    URL Parsing
   6  * AUTHOR: Duane Wessels
   7  *
   8  * SQUID Web Proxy Cache          http://www.squid-cache.org/
   9  * ----------------------------------------------------------
  10  *
  11  *  Squid is the result of efforts by numerous individuals from
  12  *  the Internet community; see the CONTRIBUTORS file for full
  13  *  details.   Many organizations have provided support for Squid's
  14  *  development; see the SPONSORS file for full details.  Squid is
  15  *  Copyrighted (C) 2001 by the Regents of the University of
  16  *  California; see the COPYRIGHT file for full details.  Squid
  17  *  incorporates software developed and/or copyrighted by other
  18  *  sources; see the CREDITS file for full details.
  19  *
  20  *  This program is free software; you can redistribute it and/or modify
  21  *  it under the terms of the GNU General Public License as published by
  22  *  the Free Software Foundation; either version 2 of the License, or
  23  *  (at your option) any later version.
  24  *
  25  *  This program is distributed in the hope that it will be useful,
  26  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
  27  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  28  *  GNU General Public License for more details.
  29  *
  30  *  You should have received a copy of the GNU General Public License
  31  *  along with this program; if not, write to the Free Software
  32  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111, USA.
  33  *
  34  */
  35
  36 #include "URL.h"
  37 #include "HttpRequest.h"
  38 #include "URLScheme.h"
  39
  40 static HttpRequest *urnParse(const HttpRequestMethod& method, char *urn);
  41 static const char valid_hostname_chars_u[] =
  42     "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
  43     "abcdefghijklmnopqrstuvwxyz"
  44     "0123456789-._"
  45 #if USE_IPV6
  46     "[:]"
  47 #endif
  48     ;
  49 static const char valid_hostname_chars[] =
  50     "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
  51     "abcdefghijklmnopqrstuvwxyz"
  52     "0123456789-."
  53 #if USE_IPV6
  54     "[:]"
  55 #endif
  56     ;
  57
  58 void
  59 urlInitialize(void)
  60 {
  61     debugs(23, 5, "urlInitialize: Initializing...");
  62     /* this ensures that the number of protocol strings is the same as
  63      * the enum slots allocated because the last enum is always 'TOTAL'.
  64      */
  65     assert(strcmp(ProtocolStr[PROTO_MAX], "TOTAL") == 0);
  66     /*
  67      * These test that our matchDomainName() function works the
  68      * way we expect it to.
  69      */
  70     assert(0 == matchDomainName("foo.com", "foo.com"));
  71     assert(0 == matchDomainName(".foo.com", "foo.com"));
  72     assert(0 == matchDomainName("foo.com", ".foo.com"));
  73     assert(0 == matchDomainName(".foo.com", ".foo.com"));
  74     assert(0 == matchDomainName("x.foo.com", ".foo.com"));
  75     assert(0 != matchDomainName("x.foo.com", "foo.com"));
  76     assert(0 != matchDomainName("foo.com", "x.foo.com"));
  77     assert(0 != matchDomainName("bar.com", "foo.com"));
  78     assert(0 != matchDomainName(".bar.com", "foo.com"));
  79     assert(0 != matchDomainName(".bar.com", ".foo.com"));
  80     assert(0 != matchDomainName("bar.com", ".foo.com"));
  81     assert(0 < matchDomainName("zzz.com", "foo.com"));
  82     assert(0 > matchDomainName("aaa.com", "foo.com"));
  83     assert(0 == matchDomainName("FOO.com", "foo.COM"));
  84     assert(0 < matchDomainName("bfoo.com", "afoo.com"));
  85     assert(0 > matchDomainName("afoo.com", "bfoo.com"));
  86     assert(0 < matchDomainName("x-foo.com", ".foo.com"));
  87     /* more cases? */
  88 }
  89
  90 /**
  91  * urlParseProtocol() takes begin (b) and end (e) pointers, but for
  92  * backwards compatibility, e defaults to NULL, in which case we
  93  * assume b is NULL-terminated.
  94  */
  95 protocol_t
  96 urlParseProtocol(const char *b, const char *e)
  97 {
  98     /*
  99      * if e is NULL, b must be NULL terminated and we
 100      * make e point to the first whitespace character
 101      * after b.
 102      */
 103
 104     if (NULL == e)
 105         e = b + strcspn(b, ":");
 106
 107     int len = e - b;
 108
 109     /* test common stuff first */
 110
 111     if (strncasecmp(b, "http", len) == 0)
 112         return PROTO_HTTP;
 113
 114     if (strncasecmp(b, "ftp", len) == 0)
 115         return PROTO_FTP;
 116
 117     if (strncasecmp(b, "https", len) == 0)
 118         return PROTO_HTTPS;
 119
 120     if (strncasecmp(b, "file", len) == 0)
 121         return PROTO_FTP;
 122
 123     if (strncasecmp(b, "gopher", len) == 0)
 124         return PROTO_GOPHER;
 125
 126     if (strncasecmp(b, "wais", len) == 0)
 127         return PROTO_WAIS;
 128
 129     if (strncasecmp(b, "cache_object", len) == 0)
 130         return PROTO_CACHEOBJ;
 131
 132     if (strncasecmp(b, "urn", len) == 0)
 133         return PROTO_URN;
 134
 135     if (strncasecmp(b, "whois", len) == 0)
 136         return PROTO_WHOIS;
 137
 138     if (strncasecmp(b, "internal", len) == 0)
 139         return PROTO_INTERNAL;
 140
 141     return PROTO_NONE;
 142 }
 143
 144 int
 145 urlDefaultPort(protocol_t p)
 146 {
 147     switch (p) {
 148
 149     case PROTO_HTTP:
 150         return 80;
 151
 152     case PROTO_HTTPS:
 153         return 443;
 154
 155     case PROTO_FTP:
 156         return 21;
 157
 158     case PROTO_GOPHER:
 159         return 70;
 160
 161     case PROTO_WAIS:
 162         return 210;
 163
 164     case PROTO_CACHEOBJ:
 165
 166     case PROTO_INTERNAL:
 167         return CACHE_HTTP_PORT;
 168
 169     case PROTO_WHOIS:
 170         return 43;
 171
 172     default:
 173         return 0;
 174     }
 175 }
 176
 177 /*
 178  * Parse a URI/URL.
 179  *
 180  * If the 'request' arg is non-NULL, put parsed values there instead
 181  * of allocating a new HttpRequest.
 182  *
 183  * This abuses HttpRequest as a way of representing the parsed url
 184  * and its components.
 185  * method is used to switch parsers and to init the HttpRequest.
 186  * If method is METHOD_CONNECT, then rather than a URL a hostname:port is
 187  * looked for.
 188  * The url is non const so that if its too long we can NULL-terminate it in place.
 189  */
 190
 191 /*
 192  * This routine parses a URL. Its assumed that the URL is complete -
 193  * ie, the end of the string is the end of the URL. Don't pass a partial
 194  * URL here as this routine doesn't have any way of knowing whether
 195  * its partial or not (ie, it handles the case of no trailing slash as
 196  * being "end of host with implied path of /".
 197  */
 198 HttpRequest *
 199 urlParse(const HttpRequestMethod& method, char *url, HttpRequest *request)
 200 {
 201     LOCAL_ARRAY(char, proto, MAX_URL);
 202     LOCAL_ARRAY(char, login, MAX_URL);
 203     LOCAL_ARRAY(char, host, MAX_URL);
 204     LOCAL_ARRAY(char, urlpath, MAX_URL);
 205     char *t = NULL;
 206     char *q = NULL;
 207     int port;
 208     protocol_t protocol = PROTO_NONE;
 209     int l;
 210     int i;
 211     const char *src;
 212     char *dst;
 213     proto[0] = host[0] = urlpath[0] = login[0] = '\0';
 214
 215     if ((l = strlen(url)) + Config.appendDomainLen > (MAX_URL - 1)) {
 216         /* terminate so it doesn't overflow other buffers */
 217         *(url + (MAX_URL >> 1)) = '\0';
 218         debugs(23, 1, "urlParse: URL too large (" << l << " bytes)");
 219         return NULL;
 220     }
 221     if (method == METHOD_CONNECT) {
 222         port = CONNECT_PORT;
 223
 224         if (sscanf(url, "[%[^]]]:%d", host, &port) < 1)
 225             if (sscanf(url, "%[^:]:%d", host, &port) < 1)
 226                 return NULL;
 227
 228     } else if (!strncmp(url, "urn:", 4)) {
 229         return urnParse(method, url);
 230     } else {
 231         /* Parse the URL: */
 232         src = url;
 233         i = 0;
 234         /* Find first : - everything before is protocol */
 235         for (i = 0, dst = proto; i < l && *src != ':'; i++, src++, dst++) {
 236             *dst = *src;
 237         }
 238         if (i >= l)
 239            return NULL;
 240         *dst = '\0';
 241
 242         /* Then its :// */
 243         /* (XXX yah, I'm not checking we've got enough data left before checking the array..) */
 244         if (*src != ':' || *(src + 1) != '/' || *(src + 2) != '/')
 245             return NULL;
 246         i += 3;
 247         src += 3;
 248
 249         /* Then everything until first /; thats host (and port; which we'll look for here later) */
 250         /* bug 1881: If we don't get a "/" then we imply it was there */
 251         for (dst = host; i < l && *src != '/' && src != '\0'; i++, src++, dst++) {
 252             *dst = *src;
 253         }
 254
 255         /*
 256          * We can't check for "i >= l" here because we could be at the end of the line
 257          * and have a perfectly valid URL w/ no trailing '/'. In this case we assume we've
 258          * been -given- a valid URL and the path is just '/'.
 259          */
 260         if (i > l)
 261             return NULL;
 262         *dst = '\0';
 263
 264         /* Then everything from / (inclusive) until \r\n or \0 - thats urlpath */
 265         for (dst = urlpath; i < l && *src != '\r' && *src != '\n' && *src != '\0'; i++, src++, dst++) {
 266             *dst = *src;
 267         }
 268
 269         /* We -could- be at the end of the buffer here */
 270         if (i > l)
 271             return NULL;
 272         /* If the URL path is empty we set it to be "/" */
 273         if (dst == urlpath) {
 274             *(dst++) = '/';
 275         }
 276         *dst = '\0';
 277
 278         protocol = urlParseProtocol(proto);
 279         port = urlDefaultPort(protocol);
 280
 281         /* Is there any login information? (we should eventually parse it above) */
 282         if ((t = strrchr(host, '@'))) {
 283             strcpy((char *) login, (char *) host);
 284             t = strrchr(login, '@');
 285             *t = 0;
 286             strcpy((char *) host, t + 1);
 287         }
 288
 289         /* Is there any host information? (we should eventually parse it above) */
 290         if(*host == '[') {
 291             /* strip any IPA brackets. valid under IPv6. */
 292             dst = host;
 293 #if USE_IPV6
 294             /* only for IPv6 sadly, pre-IPv6/URL code can't handle the clean result properly anyway. */
 295             src = host; src++;
 296             l = strlen(host);
 297             i = 1;
 298             for (; i < l && *src != ']' && *src != '\0'; i++, src++, dst++) {
 299                 *dst = *src;
 300             }
 301
 302             /* we moved in-place, so truncate the actual hostname found */
 303             *(dst++) = '\0';
 304 #else
 305             /* IPv4-pure needs to skip the whole hostname to ']' inclusive for now */
 306             while(*dst != '\0' && *dst != ']') dst++;
 307 #endif
 308
 309             /* skip ahead to either start of port, or original EOS */
 310             while(*dst != '\0' && *dst != ':') dst++;
 311             t = dst;
 312         } else {
 313             t = strrchr(host, ':');
 314
 315             if(t != strchr(host,':') ) {
 316                 /* RFC 2732 states IPv6 "SHOULD" be bracketed. allowing for times when its not. */
 317                 /* RFC 3986 'update' simply modifies this to an "is" with no emphasis at all! */
 318                 /* therefore we MUST accept the case where they are not bracketed at all. */
 319                 t = NULL;
 320             }
 321         }
 322
 323         if (t && *t == ':') {
 324             *t = '\0'; t++;
 325             port = atoi(t);
 326         }
 327     }
 328
 329     for (t = host; *t; t++)
 330         *t = xtolower(*t);
 331
 332     if (stringHasWhitespace(host)) {
 333         if (URI_WHITESPACE_STRIP == Config.uri_whitespace) {
 334             t = q = host;
 335             while (*t) {
 336                 if (!xisspace(*t))
 337                     *q++ = *t;
 338                 t++;
 339             }
 340             *q = '\0';
 341         }
 342     }
 343
 344     debugs(23, 3, "urlParse: Split URL '" << url << "' into proto='" << proto << "', host='" << host << "', port='" << port << "', path='" << urlpath << "'");
 345
 346     if (Config.onoff.check_hostnames && strspn(host, Config.onoff.allow_underscore ? valid_hostname_chars_u : valid_hostname_chars) != strlen(host)) {
 347         debugs(23, 1, "urlParse: Illegal character in hostname '" << host << "'");
 348         return NULL;
 349     }
 350
 351     if (Config.appendDomain && !strchr(host, '.'))
 352         strncat(host, Config.appendDomain, SQUIDHOSTNAMELEN - strlen(host) - 1);
 353
 354     /* remove trailing dots from hostnames */
 355     while ((l = strlen(host)) > 0 && host[--l] == '.')
 356         host[l] = '\0';
 357
 358     /* reject duplicate or leading dots */
 359     if (strstr(host, "..") || *host == '.') {
 360         debugs(23, 1, "urlParse: Illegal hostname '" << host << "'");
 361         return NULL;
 362     }
 363
 364     if (port < 1 || port > 65535) {
 365         debugs(23, 3, "urlParse: Invalid port '" << port << "'");
 366         return NULL;
 367     }
 368
 369 #ifdef HARDCODE_DENY_PORTS
 370     /* These ports are filtered in the default squid.conf, but
 371      * maybe someone wants them hardcoded... */
 372     if (port == 7 || port == 9 || port == 19) {
 373         debugs(23, 0, "urlParse: Deny access to port " << port);
 374         return NULL;
 375     }
 376 #endif
 377
 378     if (stringHasWhitespace(urlpath)) {
 379         debugs(23, 2, "urlParse: URI has whitespace: {" << url << "}");
 380
 381         switch (Config.uri_whitespace) {
 382
 383         case URI_WHITESPACE_DENY:
 384             return NULL;
 385
 386         case URI_WHITESPACE_ALLOW:
 387             break;
 388
 389         case URI_WHITESPACE_ENCODE:
 390             t = rfc1738_escape_unescaped(urlpath);
 391             xstrncpy(urlpath, t, MAX_URL);
 392             break;
 393
 394         case URI_WHITESPACE_CHOP:
 395             *(urlpath + strcspn(urlpath, w_space)) = '\0';
 396             break;
 397
 398         case URI_WHITESPACE_STRIP:
 399         default:
 400             t = q = urlpath;
 401             while (*t) {
 402                 if (!xisspace(*t))
 403                     *q++ = *t;
 404                 t++;
 405             }
 406             *q = '\0';
 407         }
 408     }
 409
 410     if (NULL == request)
 411         request = new HttpRequest(method, protocol, urlpath);
 412     else {
 413         request->initHTTP(method, protocol, urlpath);
 414     }
 415
 416     request->SetHost(host);
 417     xstrncpy(request->login, login, MAX_LOGIN_SZ);
 418     request->port = (u_short) port;
 419     return request;
 420 }
 421
 422 static HttpRequest *
 423 urnParse(const HttpRequestMethod& method, char *urn)
 424 {
 425     debugs(50, 5, "urnParse: " << urn);
 426     return new HttpRequest(method, PROTO_URN, urn + 4);
 427 }
 428
 429 const char *
 430 urlCanonical(HttpRequest * request)
 431 {
 432     LOCAL_ARRAY(char, portbuf, 32);
 433 /// \todo AYJ: Performance: making this a ptr and allocating when needed will be better than a write and future xstrdup().
 434     LOCAL_ARRAY(char, urlbuf, MAX_URL);
 435
 436     if (request->canonical)
 437         return request->canonical;
 438
 439     if (request->protocol == PROTO_URN) {
 440         snprintf(urlbuf, MAX_URL, "urn:%s", request->urlpath.buf());
 441     } else {
 442 /// \todo AYJ: this could use "if..else and method == METHOD_CONNECT" easier.
 443         switch (request->method.id()) {
 444
 445         case METHOD_CONNECT:
 446             snprintf(urlbuf, MAX_URL, "%s:%d", request->GetHost(), request->port);
 447             break;
 448
 449         default:
 450             portbuf[0] = '\0';
 451
 452             if (request->port != urlDefaultPort(request->protocol))
 453                 snprintf(portbuf, 32, ":%d", request->port);
 454
 455             snprintf(urlbuf, MAX_URL, "%s://%s%s%s%s%s",
 456                      ProtocolStr[request->protocol],
 457                      request->login,
 458                      *request->login ? "@" : null_string,
 459                      request->GetHost(),
 460                      portbuf,
 461                      request->urlpath.buf());
 462
 463             break;
 464         }
 465     }
 466
 467     return (request->canonical = xstrdup(urlbuf));
 468 }
 469
 470 /** \todo AYJ: Performance: This is an *almost* duplicate of urlCanoncical. But elides the query-string.
 471  *        After copying it on in the first place! Would be less code to merge the two with a flag parameter.
 472  *        and never copy the query-string part in the first place
 473  */
 474 char *
 475 urlCanonicalClean(const HttpRequest * request)
 476 {
 477     LOCAL_ARRAY(char, buf, MAX_URL);
 478     LOCAL_ARRAY(char, portbuf, 32);
 479     LOCAL_ARRAY(char, loginbuf, MAX_LOGIN_SZ + 1);
 480     char *t;
 481
 482     if (request->protocol == PROTO_URN) {
 483         snprintf(buf, MAX_URL, "urn:%s", request->urlpath.buf());
 484     } else {
 485 /// \todo AYJ: this could use "if..else and method == METHOD_CONNECT" easier.
 486         switch (request->method.id()) {
 487
 488         case METHOD_CONNECT:
 489             snprintf(buf, MAX_URL, "%s:%d",
 490                      request->GetHost(),
 491                      request->port);
 492             break;
 493
 494         default:
 495             portbuf[0] = '\0';
 496
 497             if (request->port != urlDefaultPort(request->protocol))
 498                 snprintf(portbuf, 32, ":%d", request->port);
 499
 500             loginbuf[0] = '\0';
 501
 502             if ((int) strlen(request->login) > 0) {
 503                 strcpy(loginbuf, request->login);
 504
 505                 if ((t = strchr(loginbuf, ':')))
 506                     *t = '\0';
 507
 508                 strcat(loginbuf, "@");
 509             }
 510
 511             snprintf(buf, MAX_URL, "%s://%s%s%s%s",
 512                      ProtocolStr[request->protocol],
 513                      loginbuf,
 514                      request->GetHost(),
 515                      portbuf,
 516                      request->urlpath.buf());
 517             /*
 518              * strip arguments AFTER a question-mark
 519              */
 520
 521             if (Config.onoff.strip_query_terms)
 522                 if ((t = strchr(buf, '?')))
 523                     *(++t) = '\0';
 524
 525             break;
 526         }
 527     }
 528
 529     if (stringHasCntl(buf))
 530         xstrncpy(buf, rfc1738_escape_unescaped(buf), MAX_URL);
 531
 532     return buf;
 533 }
 534
 535 /*
 536  * Test if a URL is relative.
 537  *
 538  * RFC 2396, Section 5 (Page 17) implies that in a relative URL, a '/' will
 539  * appear before a ':'.
 540  */
 541 bool
 542 urlIsRelative(const char *url)
 543 {
 544     const char *p;
 545
 546     if (url == NULL) {
 547         return (false);
 548     }
 549     if (*url == '\0') {
 550         return (false);
 551     }
 552
 553     for (p = url; *p != '\0' && *p != ':' && *p != '/'; p++);
 554
 555     if (*p == ':') {
 556         return (false);
 557     }
 558     return (true);
 559 }
 560
 561 /*
 562  * Convert a relative URL to an absolute URL using the context of a given
 563  * request.
 564  *
 565  * It is assumed that you have already ensured that the URL is relative.
 566  *
 567  * If NULL is returned it is an indication that the method in use in the
 568  * request does not distinguish between relative and absolute and you should
 569  * use the url unchanged.
 570  *
 571  * If non-NULL is returned, it is up to the caller to free the resulting
 572  * memory using safe_free().
 573  */
 574 char *
 575 urlMakeAbsolute(const HttpRequest * req, const char *relUrl)
 576 {
 577
 578     if (req->method.id() == METHOD_CONNECT) {
 579         return (NULL);
 580     }
 581
 582     char *urlbuf = (char *)xmalloc(MAX_URL * sizeof(char));
 583
 584     if (req->protocol == PROTO_URN) {
 585         snprintf(urlbuf, MAX_URL, "urn:%s", req->urlpath.buf());
 586         return (urlbuf);
 587     }
 588
 589     size_t urllen;
 590
 591     if (req->port != urlDefaultPort(req->protocol)) {
 592         urllen = snprintf(urlbuf, MAX_URL, "%s://%s%s%s:%d",
 593             ProtocolStr[req->protocol],
 594             req->login,
 595             *req->login ? "@" : null_string,
 596             req->GetHost(),
 597             req->port
 598         );
 599     } else {
 600         urllen = snprintf(urlbuf, MAX_URL, "%s://%s%s%s",
 601             ProtocolStr[req->protocol],
 602             req->login,
 603             *req->login ? "@" : null_string,
 604             req->GetHost()
 605         );
 606     }
 607
 608     if (relUrl[0] == '/') {
 609         strncpy(&urlbuf[urllen], relUrl, MAX_URL - urllen - 1);
 610     } else {
 611         const char *path = req->urlpath.buf();
 612         const char *last_slash = strrchr(path, '/');
 613
 614         if (last_slash == NULL) {
 615                urlbuf[urllen++] = '/';
 616             strncpy(&urlbuf[urllen], relUrl, MAX_URL - urllen - 1);
 617         } else {
 618             last_slash++;
 619             size_t pathlen = last_slash - path;
 620             if (pathlen > MAX_URL - urllen - 1) {
 621                 pathlen = MAX_URL - urllen - 1;
 622             }
 623             strncpy(&urlbuf[urllen], path, pathlen);
 624             urllen += pathlen;
 625             if (urllen + 1 < MAX_URL) {
 626                 strncpy(&urlbuf[urllen], relUrl, MAX_URL - urllen - 1);
 627             }
 628         }
 629     }
 630
 631     return (urlbuf);
 632 }
 633
 634 /*
 635  * matchDomainName() compares a hostname with a domainname according
 636  * to the following rules:
 637  *
 638  *    HOST          DOMAIN        MATCH?
 639  * ------------- -------------    ------
 640  *    foo.com       foo.com         YES
 641  *   .foo.com       foo.com         YES
 642  *  x.foo.com       foo.com          NO
 643  *    foo.com      .foo.com         YES
 644  *   .foo.com      .foo.com         YES
 645  *  x.foo.com      .foo.com         YES
 646  *
 647  *  We strip leading dots on hosts (but not domains!) so that
 648  *  ".foo.com" is is always the same as "foo.com".
 649  *
 650  *  Return values:
 651  *     0 means the host matches the domain
 652  *     1 means the host is greater than the domain
 653  *    -1 means the host is less than the domain
 654  */
 655
 656 int
 657 matchDomainName(const char *h, const char *d)
 658 {
 659     int dl;
 660     int hl;
 661
 662     while ('.' == *h)
 663         h++;
 664
 665     hl = strlen(h);
 666
 667     dl = strlen(d);
 668
 669     /*
 670      * Start at the ends of the two strings and work towards the
 671      * beginning.
 672      */
 673     while (xtolower(h[--hl]) == xtolower(d[--dl])) {
 674         if (hl == 0 && dl == 0) {
 675             /*
 676              * We made it all the way to the beginning of both
 677              * strings without finding any difference.
 678              */
 679             return 0;
 680         }
 681
 682         if (0 == hl) {
 683             /*
 684              * The host string is shorter than the domain string.
 685              * There is only one case when this can be a match.
 686              * If the domain is just one character longer, and if
 687              * that character is a leading '.' then we call it a
 688              * match.
 689              */
 690
 691             if (1 == dl && '.' == d[0])
 692                 return 0;
 693             else
 694                 return -1;
 695         }
 696
 697         if (0 == dl) {
 698             /*
 699              * The domain string is shorter than the host string.
 700              * This is a match only if the first domain character
 701              * is a leading '.'.
 702              */
 703
 704             if ('.' == d[0])
 705                 return 0;
 706             else
 707                 return 1;
 708         }
 709     }
 710
 711     /*
 712      * We found different characters in the same position (from the end).
 713      */
 714     /*
 715      * If one of those character is '.' then its special.  In order
 716      * for splay tree sorting to work properly, "x-foo.com" must
 717      * be greater than ".foo.com" even though '-' is less than '.'.
 718      */
 719     if ('.' == d[dl])
 720         return 1;
 721
 722     if ('.' == h[hl])
 723         return -1;
 724
 725     return (xtolower(h[hl]) - xtolower(d[dl]));
 726 }
 727
 728
 729 /*
 730  * return true if we can serve requests for this method.
 731  */
 732 int
 733 urlCheckRequest(const HttpRequest * r)
 734 {
 735     int rc = 0;
 736     /* protocol "independent" methods
 737      *
 738      * actually these methods are specific to HTTP:
 739      * they are methods we recieve on our HTTP port,
 740      * and if we had a FTP listener would not be relevant
 741      * there.
 742      *
 743      * So, we should delegate them to HTTP. The problem is that we
 744      * do not have a default protocol from the client side of HTTP.
 745      */
 746
 747     if (r->method == METHOD_CONNECT)
 748         return 1;
 749
 750     if (r->method == METHOD_TRACE)
 751         return 1;
 752
 753     if (r->method == METHOD_PURGE)
 754         return 1;
 755
 756     /* does method match the protocol? */
 757     switch (r->protocol) {
 758
 759     case PROTO_URN:
 760
 761     case PROTO_HTTP:
 762
 763     case PROTO_CACHEOBJ:
 764         rc = 1;
 765         break;
 766
 767     case PROTO_FTP:
 768
 769         if (r->method == METHOD_PUT)
 770             rc = 1;
 771
 772     case PROTO_GOPHER:
 773
 774     case PROTO_WAIS:
 775
 776     case PROTO_WHOIS:
 777         if (r->method == METHOD_GET)
 778             rc = 1;
 779         else if (r->method == METHOD_HEAD)
 780             rc = 1;
 781
 782         break;
 783
 784     case PROTO_HTTPS:
 785 #ifdef USE_SSL
 786
 787         rc = 1;
 788
 789         break;
 790
 791 #else
 792         /*
 793         * Squid can't originate an SSL connection, so it should
 794         * never receive an "https:" URL.  It should always be
 795         * CONNECT instead.
 796         */
 797         rc = 0;
 798
 799 #endif
 800
 801     default:
 802         break;
 803     }
 804
 805     return rc;
 806 }
 807
 808 /*
 809  * Quick-n-dirty host extraction from a URL.  Steps:
 810  *      Look for a colon
 811  *      Skip any '/' after the colon
 812  *      Copy the next SQUID_MAXHOSTNAMELEN bytes to host[]
 813  *      Look for an ending '/' or ':' and terminate
 814  *      Look for login info preceeded by '@'
 815  */
 816
 817 class URLHostName
 818 {
 819
 820 public:
 821     char * extract(char const *url);
 822
 823 private:
 824     static char Host [SQUIDHOSTNAMELEN];
 825     void init(char const *);
 826     void findHostStart();
 827     void trimTrailingChars();
 828     void trimAuth();
 829     char const *hostStart;
 830     char const *url;
 831 };
 832
 833 char *
 834 urlHostname(const char *url)
 835 {
 836     return URLHostName().extract(url);
 837 }
 838
 839 char URLHostName::Host[SQUIDHOSTNAMELEN];
 840
 841 void
 842 URLHostName::init(char const *aUrl)
 843 {
 844     Host[0] = '\0';
 845     url = url;
 846 }
 847
 848 void
 849 URLHostName::findHostStart()
 850 {
 851     if (NULL == (hostStart = strchr(url, ':')))
 852         return;
 853
 854     ++hostStart;
 855
 856     while (*hostStart != '\0' && *hostStart == '/')
 857         ++hostStart;
 858
 859 #if USE_IPV6
 860     if (*hostStart == ']')
 861         ++hostStart;
 862 #endif
 863
 864 }
 865
 866 void
 867 URLHostName::trimTrailingChars()
 868 {
 869     char *t;
 870
 871     if ((t = strchr(Host, '/')))
 872         *t = '\0';
 873
 874     if ((t = strrchr(Host, ':')))
 875         *t = '\0';
 876
 877 #if USE_IPV6
 878     if ((t = strchr(Host, ']')))
 879         *t = '\0';
 880 #endif
 881
 882 }
 883
 884 void
 885 URLHostName::trimAuth()
 886 {
 887     char *t;
 888
 889     if ((t = strrchr(Host, '@'))) {
 890         t++;
 891         xmemmove(Host, t, strlen(t) + 1);
 892     }
 893 }
 894
 895 char *
 896 URLHostName::extract(char const *aUrl)
 897 {
 898     init(aUrl);
 899     findHostStart();
 900
 901     if (hostStart == NULL)
 902         return NULL;
 903
 904     xstrncpy(Host, hostStart, SQUIDHOSTNAMELEN);
 905
 906     trimTrailingChars();
 907
 908     trimAuth();
 909
 910     return Host;
 911 }
 912
 913 URL::URL() : scheme()
 914 {}
 915
 916 URL::URL(URLScheme const &aScheme): scheme(aScheme)
 917 {}