1 const char urlmatch_rcs[] = "$Id: urlmatch.c,v 1.48 2009/04/17 11:27:49 fabiankeil Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/urlmatch.c,v $
6 * Purpose : Declares functions to match URLs against URL
9 * Copyright : Written by and Copyright (C) 2001-2009
10 * the Privoxy team. http://www.privoxy.org/
12 * Based on the Internet Junkbuster originally written
13 * by and Copyright (C) 1997 Anonymous Coders and
14 * Junkbusters Corporation. http://www.junkbusters.com
16 * This program is free software; you can redistribute it
17 * and/or modify it under the terms of the GNU General
18 * Public License as published by the Free Software
19 * Foundation; either version 2 of the License, or (at
20 * your option) any later version.
22 * This program is distributed in the hope that it will
23 * be useful, but WITHOUT ANY WARRANTY; without even the
24 * implied warranty of MERCHANTABILITY or FITNESS FOR A
25 * PARTICULAR PURPOSE. See the GNU General Public
26 * License for more details.
28 * The GNU General Public License should be included with
29 * this file. If not, you can view it at
30 * http://www.gnu.org/copyleft/gpl.html
31 * or write to the Free Software Foundation, Inc., 59
32 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
35 * $Log: urlmatch.c,v $
36 * Revision 1.48 2009/04/17 11:27:49 fabiankeil
37 * Petr Pisar's privoxy-3.0.12-ipv6-3.diff.
39 * Revision 1.47 2009/03/02 19:18:10 fabiankeil
40 * Streamline parse_http_request()'s prototype. As
41 * cparser pointed out it doesn't actually use csp.
43 * Revision 1.46 2009/02/11 19:31:32 fabiankeil
44 * Reject request lines that end with neither HTTP/1.0 nor HTTP/1.1.
46 * Revision 1.45 2008/06/21 21:19:18 fabiankeil
47 * Silence bogus compiler warning.
49 * Revision 1.44 2008/05/04 16:18:32 fabiankeil
50 * Provide parse_http_url() with a third parameter to specify
51 * whether or not URLs without protocol are acceptable.
53 * Revision 1.43 2008/05/04 13:30:55 fabiankeil
54 * Streamline parse_http_url()'s prototype.
56 * Revision 1.42 2008/05/04 13:24:16 fabiankeil
57 * If the method isn't CONNECT, reject URLs without protocol.
59 * Revision 1.41 2008/05/02 09:51:34 fabiankeil
60 * In parse_http_url(), don't muck around with values
61 * that are none of its business: require an initialized
62 * http structure and never unset http->ssl.
64 * Revision 1.40 2008/04/23 16:12:28 fabiankeil
67 * Revision 1.39 2008/04/22 16:27:42 fabiankeil
68 * In parse_http_request(), remove a pointless
69 * temporary variable and free the buffer earlier.
71 * Revision 1.38 2008/04/18 05:17:18 fabiankeil
72 * Mark simplematch()'s parameters as immutable.
74 * Revision 1.37 2008/04/17 14:53:29 fabiankeil
75 * Move simplematch() into urlmatch.c as it's only
76 * used to match (old-school) domain patterns.
78 * Revision 1.36 2008/04/14 18:19:48 fabiankeil
79 * Remove now-pointless cast in create_url_spec().
81 * Revision 1.35 2008/04/14 18:11:21 fabiankeil
82 * The compiler might not notice it, but the buffer passed to
83 * create_url_spec() is modified later on and thus shouldn't
84 * be declared immutable.
86 * Revision 1.34 2008/04/13 13:32:07 fabiankeil
87 * Factor URL pattern compilation out of create_url_spec().
89 * Revision 1.33 2008/04/12 14:03:13 fabiankeil
90 * Remove an obvious comment and improve another one.
92 * Revision 1.32 2008/04/12 12:38:06 fabiankeil
93 * Factor out duplicated code to compile host, path and tag patterns.
95 * Revision 1.31 2008/04/10 14:41:04 fabiankeil
96 * Ditch url_spec's path member now that it's no longer used.
98 * Revision 1.30 2008/04/10 04:24:24 fabiankeil
99 * Stop duplicating the plain text representation of the path regex
100 * (and keeping the copy around). Once the regex is compiled it's no
103 * Revision 1.29 2008/04/10 04:17:56 fabiankeil
104 * In url_match(), check the right member for NULL when determining
105 * whether there's a path regex to execute. Looking for a plain-text
106 * representation works as well, but it looks "interesting" and that
107 * member will be removed soonish anyway.
109 * Revision 1.28 2008/04/08 16:07:39 fabiankeil
110 * Make it harder to mistake url_match()'s
111 * second parameter for an url_spec.
113 * Revision 1.27 2008/04/08 15:44:33 fabiankeil
114 * Save a bit of memory (and a few cpu cycles) by not bothering to
115 * compile slash-only path regexes that don't affect the result.
117 * Revision 1.26 2008/04/07 16:57:18 fabiankeil
118 * - Use free_url_spec() more consistently.
119 * - Let it reset url->dcount just in case.
121 * Revision 1.25 2008/04/06 15:18:38 fabiankeil
122 * Oh well, rename the --enable-pcre-host-patterns option to
123 * --enable-extended-host-patterns as it's not really PCRE syntax.
125 * Revision 1.24 2008/04/06 14:54:26 fabiankeil
126 * Use PCRE syntax in host patterns when configured
127 * with --enable-pcre-host-patterns.
129 * Revision 1.23 2008/04/05 12:19:20 fabiankeil
130 * Factor compile_host_pattern() out of create_url_spec().
132 * Revision 1.22 2008/03/30 15:02:32 fabiankeil
133 * SZitify unknown_method().
135 * Revision 1.21 2007/12/24 16:34:23 fabiankeil
136 * Band-aid (and micro-optimization) that makes it less likely to run out of
137 * stack space with overly-complex path patterns. Probably masks the problem
138 * reported by Lee in #1856679. Hohoho.
140 * Revision 1.20 2007/09/02 15:31:20 fabiankeil
141 * Move match_portlist() from filter.c to urlmatch.c.
142 * It's used for url matching, not for filtering.
144 * Revision 1.19 2007/09/02 13:42:11 fabiankeil
145 * - Allow port lists in url patterns.
146 * - Ditch unused url_spec member pathlen.
148 * Revision 1.18 2007/07/30 16:42:21 fabiankeil
149 * Move the method check into unknown_method()
150 * and loop through the known methods instead
151 * of using a screen-long OR chain.
153 * Revision 1.17 2007/04/15 16:39:21 fabiankeil
154 * Introduce tags as alternative way to specify which
155 * actions apply to a request. At the moment tags can be
156 * created based on client and server headers.
158 * Revision 1.16 2007/02/13 13:59:24 fabiankeil
159 * Remove redundant log message.
161 * Revision 1.15 2007/01/28 16:11:23 fabiankeil
162 * Accept WebDAV methods for subversion
163 * in parse_http_request(). Closes FR 1581425.
165 * Revision 1.14 2007/01/06 14:23:56 fabiankeil
166 * Fix gcc43 warnings. Mark *csp as immutable
167 * for parse_http_url() and url_match().
168 * Replace a sprintf call with snprintf.
170 * Revision 1.13 2006/12/06 19:50:54 fabiankeil
171 * parse_http_url() now handles intercepted
172 * HTTP request lines as well. Moved parts
173 * of parse_http_url()'s code into
174 * init_domain_components() so that it can
175 * be reused in chat().
177 * Revision 1.12 2006/07/18 14:48:47 david__schmidt
178 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
179 * with what was really the latest development (the v_3_0_branch branch)
181 * Revision 1.10.2.7 2003/05/17 15:57:24 oes
182 * - parse_http_url now checks memory allocation failure for
183 * duplication of "*" URL and rejects "*something" URLs
185 * - Added a comment to what might look like a bug in
186 * create_url_spec (see !bug #736931)
187 * - Comment cosmetics
189 * Revision 1.10.2.6 2003/05/07 12:39:48 oes
190 * Fix typo: Default port for https URLs is 443, not 143.
191 * Thanks to Scott Tregear for spotting this one.
193 * Revision 1.10.2.5 2003/02/28 13:09:29 oes
194 * Fixed a rare double free condition as per Bug #694713
196 * Revision 1.10.2.4 2003/02/28 12:57:44 oes
197 * Moved freeing of http request structure to its owner
198 * as per Dan Price's observations in Bug #694713
200 * Revision 1.10.2.3 2002/11/12 16:50:40 oes
201 * Fixed memory leak in parse_http_request() reported by Oliver Stoeneberg. Fixes bug #637073
203 * Revision 1.10.2.2 2002/09/25 14:53:15 oes
204 * Added basic support for OPTIONS and TRACE HTTP methods:
205 * parse_http_url now recognizes the "*" URI as well as
206 * the OPTIONS and TRACE method keywords.
208 * Revision 1.10.2.1 2002/06/06 19:06:44 jongfoster
209 * Adding support for proprietary Microsoft WebDAV extensions
211 * Revision 1.10 2002/05/12 21:40:37 jongfoster
212 * - Removing some unused code
214 * Revision 1.9 2002/04/04 00:36:36 gliptak
215 * always use pcre for matching
217 * Revision 1.8 2002/04/03 23:32:47 jongfoster
218 * Fixing memory leak on error
220 * Revision 1.7 2002/03/26 22:29:55 swa
221 * we have a new homepage!
223 * Revision 1.6 2002/03/24 13:25:43 swa
224 * name change related issues
226 * Revision 1.5 2002/03/13 00:27:05 jongfoster
229 * Revision 1.4 2002/03/07 03:46:17 oes
230 * Fixed compiler warnings
232 * Revision 1.3 2002/03/03 14:51:11 oes
233 * Fixed CLF logging: Added ocmd member for client's request to struct http_request
235 * Revision 1.2 2002/01/21 00:14:09 jongfoster
236 * Correcting comment style
237 * Fixing an uninitialized memory bug in create_url_spec()
239 * Revision 1.1 2002/01/17 20:53:46 jongfoster
240 * Moving all our URL and URL pattern parsing code to the same file - it
241 * was scattered around in filters.c, loaders.c and parsers.c.
243 * Providing a single, simple url_match(pattern,url) function - rather than
244 * the 3-line match routine which was repeated all over the place.
246 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
248 * Providing parse_http_url() so that URLs can be parsed without faking a
249 * HTTP request line for parse_http_request() or repeating the parsing
250 * code (both of which were techniques that were actually in use).
252 * Standardizing that struct http_request is used to represent a URL, and
253 * struct url_spec is used to represent a URL pattern. (Before, URLs were
254 * represented as seperate variables and a partially-filled-in url_spec).
257 *********************************************************************/
264 #include <sys/types.h>
272 #if !defined(_WIN32) && !defined(__OS2__)
277 #include "urlmatch.h"
279 #include "miscutil.h"
282 const char urlmatch_h_rcs[] = URLMATCH_H_VERSION;
284 enum regex_anchoring {NO_ANCHORING, LEFT_ANCHORED, RIGHT_ANCHORED};
285 static jb_err compile_host_pattern(struct url_spec *url, const char *host_pattern);
287 /*********************************************************************
289 * Function : free_http_request
291 * Description : Freez a http_request structure
294 * 1 : http = points to a http_request structure to free
298 *********************************************************************/
299 void free_http_request(struct http_request *http)
308 freez(http->hostport);
311 freez(http->host_ip_addr_str);
312 freez(http->dbuffer);
318 /*********************************************************************
320 * Function : init_domain_components
322 * Description : Splits the domain name so we can compare it
323 * against wildcards. It used to be part of
324 * parse_http_url, but was separated because the
325 * same code is required in chat in case of
326 * intercepted requests.
329 * 1 : http = pointer to the http structure to hold elements.
331 * Returns : JB_ERR_OK on success
332 * JB_ERR_MEMORY on out of memory
333 * JB_ERR_PARSE on malformed command/URL
334 * or >100 domains deep.
336 *********************************************************************/
337 jb_err init_domain_components(struct http_request *http)
339 char *vec[BUFFER_SIZE];
343 http->dbuffer = strdup(http->host);
344 if (NULL == http->dbuffer)
346 return JB_ERR_MEMORY;
349 /* map to lower case */
350 for (p = http->dbuffer; *p ; p++)
352 *p = (char)tolower((int)(unsigned char)*p);
355 /* split the domain name into components */
356 http->dcount = ssplit(http->dbuffer, ".", vec, SZ(vec), 1, 1);
358 if (http->dcount <= 0)
361 * Error: More than SZ(vec) components in domain
362 * or: no components in domain
364 log_error(LOG_LEVEL_ERROR, "More than SZ(vec) components in domain or none at all.");
368 /* save a copy of the pointers in dvec */
369 size = (size_t)http->dcount * sizeof(*http->dvec);
371 http->dvec = (char **)malloc(size);
372 if (NULL == http->dvec)
374 return JB_ERR_MEMORY;
377 memcpy(http->dvec, vec, size);
383 /*********************************************************************
385 * Function : parse_http_url
387 * Description : Parse out the host and port from the URL. Find the
388 * hostname & path, port (if ':'), and/or password (if '@')
391 * 1 : url = URL (or is it URI?) to break down
392 * 2 : http = pointer to the http structure to hold elements.
393 * Must be initialized with valid values (like NULLs).
394 * 3 : require_protocol = Whether or not URLs without
395 * protocol are acceptable.
397 * Returns : JB_ERR_OK on success
398 * JB_ERR_MEMORY on out of memory
399 * JB_ERR_PARSE on malformed command/URL
400 * or >100 domains deep.
402 *********************************************************************/
403 jb_err parse_http_url(const char *url, struct http_request *http, int require_protocol)
405 int host_available = 1; /* A proxy can dream. */
408 * Save our initial URL
410 http->url = strdup(url);
411 if (http->url == NULL)
413 return JB_ERR_MEMORY;
418 * Check for * URI. If found, we're done.
420 if (*http->url == '*')
422 if ( NULL == (http->path = strdup("*"))
423 || NULL == (http->hostport = strdup("")) )
425 return JB_ERR_MEMORY;
427 if (http->url[1] != '\0')
436 * Split URL into protocol,hostport,path.
446 return JB_ERR_MEMORY;
449 /* Find the start of the URL in our scratch space */
451 if (strncmpic(url_noproto, "http://", 7) == 0)
455 else if (strncmpic(url_noproto, "https://", 8) == 0)
458 * Should only happen when called from cgi_show_url_info().
463 else if (*url_noproto == '/')
466 * Short request line without protocol and host.
467 * Most likely because the client's request
468 * was intercepted and redirected into Privoxy.
473 else if (require_protocol)
479 url_path = strchr(url_noproto, '/');
480 if (url_path != NULL)
485 * NOTE: The following line ignores the path for HTTPS URLS.
486 * This means that you get consistent behaviour if you type a
487 * https URL in and it's parsed by the function. (When the
488 * URL is actually retrieved, SSL hides the path part).
490 http->path = strdup(http->ssl ? "/" : url_path);
492 http->hostport = strdup(url_noproto);
497 * Repair broken HTTP requests that don't contain a path,
498 * or CONNECT requests
500 http->path = strdup("/");
501 http->hostport = strdup(url_noproto);
506 if ( (http->path == NULL)
507 || (http->hostport == NULL))
509 return JB_ERR_MEMORY;
515 /* Without host, there is nothing left to do here */
520 * Split hostport into user/password (ignored), host, port.
527 buf = strdup(http->hostport);
530 return JB_ERR_MEMORY;
533 /* check if url contains username and/or password */
534 host = strchr(buf, '@');
537 /* Contains username/password, skip it and the @ sign. */
542 /* No username or password. */
546 /* Move after hostname before port number */
549 /* Numeric IPv6 address delimited by brackets */
551 port = strchr(host, ']');
555 /* Missing closing bracket */
566 else if (*port != ':')
568 /* Garbage after closing bracket */
575 /* Plain non-escaped hostname */
576 port = strchr(host, ':');
579 /* check if url contains port */
583 /* Terminate hostname and point to start of port string */
585 http->port = atoi(port);
589 /* No port specified. */
590 http->port = (http->ssl ? 443 : 80);
593 http->host = strdup(host);
597 if (http->host == NULL)
599 return JB_ERR_MEMORY;
604 * Split domain name so we can compare it against wildcards
606 return init_domain_components(http);
611 /*********************************************************************
613 * Function : unknown_method
615 * Description : Checks whether a method is unknown.
618 * 1 : method = points to a http method
620 * Returns : TRUE if it's unknown, FALSE otherwise.
622 *********************************************************************/
623 static int unknown_method(const char *method)
625 static const char *known_http_methods[] = {
626 /* Basic HTTP request type */
627 "GET", "HEAD", "POST", "PUT", "DELETE", "OPTIONS", "TRACE", "CONNECT",
628 /* webDAV extensions (RFC2518) */
629 "PROPFIND", "PROPPATCH", "MOVE", "COPY", "MKCOL", "LOCK", "UNLOCK",
631 * Microsoft webDAV extension for Exchange 2000. See:
632 * http://lists.w3.org/Archives/Public/w3c-dist-auth/2002JanMar/0001.html
633 * http://msdn.microsoft.com/library/en-us/wss/wss/_webdav_methods.asp
635 "BCOPY", "BMOVE", "BDELETE", "BPROPFIND", "BPROPPATCH",
637 * Another Microsoft webDAV extension for Exchange 2000. See:
638 * http://systems.cs.colorado.edu/grunwald/MobileComputing/Papers/draft-cohen-gena-p-base-00.txt
639 * http://lists.w3.org/Archives/Public/w3c-dist-auth/2002JanMar/0001.html
640 * http://msdn.microsoft.com/library/en-us/wss/wss/_webdav_methods.asp
642 "SUBSCRIBE", "UNSUBSCRIBE", "NOTIFY", "POLL",
644 * Yet another WebDAV extension, this time for
645 * Web Distributed Authoring and Versioning (RFC3253)
647 "VERSION-CONTROL", "REPORT", "CHECKOUT", "CHECKIN", "UNCHECKOUT",
648 "MKWORKSPACE", "UPDATE", "LABEL", "MERGE", "BASELINE-CONTROL", "MKACTIVITY",
652 for (i = 0; i < SZ(known_http_methods); i++)
654 if (0 == strcmpic(method, known_http_methods[i]))
665 /*********************************************************************
667 * Function : parse_http_request
669 * Description : Parse out the host and port from the URL. Find the
670 * hostname & path, port (if ':'), and/or password (if '@')
673 * 1 : req = HTTP request line to break down
674 * 2 : http = pointer to the http structure to hold elements
676 * Returns : JB_ERR_OK on success
677 * JB_ERR_MEMORY on out of memory
678 * JB_ERR_CGI_PARAMS on malformed command/URL
679 * or >100 domains deep.
681 *********************************************************************/
682 jb_err parse_http_request(const char *req, struct http_request *http)
685 char *v[10]; /* XXX: Why 10? We should only need three. */
689 memset(http, '\0', sizeof(*http));
694 return JB_ERR_MEMORY;
697 n = ssplit(buf, " \r\n", v, SZ(v), 1, 1);
705 * Fail in case of unknown methods
706 * which we might not handle correctly.
708 * XXX: There should be a config option
709 * to forward requests with unknown methods
710 * anyway. Most of them don't need special
713 if (unknown_method(v[0]))
715 log_error(LOG_LEVEL_ERROR, "Unknown HTTP method detected: %s", v[0]);
720 if (strcmpic(v[2], "HTTP/1.1") && strcmpic(v[2], "HTTP/1.0"))
722 log_error(LOG_LEVEL_ERROR, "The only supported HTTP "
723 "versions are 1.0 and 1.1. This rules out: %s", v[2]);
728 http->ssl = !strcmpic(v[0], "CONNECT");
730 err = parse_http_url(v[1], http, !http->ssl);
738 * Copy the details into the structure
740 http->cmd = strdup(req);
741 http->gpc = strdup(v[0]);
742 http->ver = strdup(v[2]);
746 if ( (http->cmd == NULL)
747 || (http->gpc == NULL)
748 || (http->ver == NULL) )
750 return JB_ERR_MEMORY;
758 /*********************************************************************
760 * Function : compile_pattern
762 * Description : Compiles a host, domain or TAG pattern.
765 * 1 : pattern = The pattern to compile.
766 * 2 : anchoring = How the regex should be anchored.
767 * Can be either one of NO_ANCHORING,
768 * LEFT_ANCHORED or RIGHT_ANCHORED.
769 * 3 : url = In case of failures, the spec member is
770 * logged and the structure freed.
771 * 4 : regex = Where the compiled regex should be stored.
773 * Returns : JB_ERR_OK - Success
774 * JB_ERR_MEMORY - Out of memory
775 * JB_ERR_PARSE - Cannot parse regex
777 *********************************************************************/
778 static jb_err compile_pattern(const char *pattern, enum regex_anchoring anchoring,
779 struct url_spec *url, regex_t **regex)
782 char rebuf[BUFFER_SIZE];
783 const char *fmt = NULL;
786 assert(strlen(pattern) < sizeof(rebuf) - 2);
788 if (pattern[0] == '\0')
806 log_error(LOG_LEVEL_FATAL,
807 "Invalid anchoring in compile_pattern %d", anchoring);
810 *regex = zalloc(sizeof(**regex));
814 return JB_ERR_MEMORY;
817 snprintf(rebuf, sizeof(rebuf), fmt, pattern);
819 errcode = regcomp(*regex, rebuf, (REG_EXTENDED|REG_NOSUB|REG_ICASE));
823 size_t errlen = regerror(errcode, *regex, rebuf, sizeof(rebuf));
824 if (errlen > (sizeof(rebuf) - (size_t)1))
826 errlen = sizeof(rebuf) - (size_t)1;
828 rebuf[errlen] = '\0';
829 log_error(LOG_LEVEL_ERROR, "error compiling %s from %s: %s",
830 pattern, url->spec, rebuf);
841 /*********************************************************************
843 * Function : compile_url_pattern
845 * Description : Compiles the three parts of an URL pattern.
848 * 1 : url = Target url_spec to be filled in.
849 * 2 : buf = The url pattern to compile. Will be messed up.
851 * Returns : JB_ERR_OK - Success
852 * JB_ERR_MEMORY - Out of memory
853 * JB_ERR_PARSE - Cannot parse regex
855 *********************************************************************/
856 static jb_err compile_url_pattern(struct url_spec *url, char *buf)
860 p = strchr(buf, '/');
864 * Only compile the regex if it consists of more than
865 * a single slash, otherwise it wouldn't affect the result.
870 * XXX: does it make sense to compile the slash at the beginning?
872 jb_err err = compile_pattern(p, LEFT_ANCHORED, url, &url->preg);
874 if (JB_ERR_OK != err)
883 * IPv6 numeric hostnames can contain colons, thus we need
884 * to delimit the hostname before the real port separator.
885 * As brackets are already used in the hostname pattern,
886 * we use angle brackets ('<', '>') instead.
888 if ((buf[0] == '<') && (NULL != (p = strchr(buf + 1, '>'))))
895 /* IPv6 address without port number */
900 /* Garbage after address delimiter */
906 p = strchr(buf, ':');
912 url->port_list = strdup(p);
913 if (NULL == url->port_list)
915 return JB_ERR_MEMORY;
920 url->port_list = NULL;
925 return compile_host_pattern(url, buf);
933 #ifdef FEATURE_EXTENDED_HOST_PATTERNS
934 /*********************************************************************
936 * Function : compile_host_pattern
938 * Description : Parses and compiles a host pattern..
941 * 1 : url = Target url_spec to be filled in.
942 * 2 : host_pattern = Host pattern to compile.
944 * Returns : JB_ERR_OK - Success
945 * JB_ERR_MEMORY - Out of memory
946 * JB_ERR_PARSE - Cannot parse regex
948 *********************************************************************/
949 static jb_err compile_host_pattern(struct url_spec *url, const char *host_pattern)
951 return compile_pattern(host_pattern, RIGHT_ANCHORED, url, &url->host_regex);
956 /*********************************************************************
958 * Function : compile_host_pattern
960 * Description : Parses and "compiles" an old-school host pattern.
963 * 1 : url = Target url_spec to be filled in.
964 * 2 : host_pattern = Host pattern to parse.
966 * Returns : JB_ERR_OK - Success
967 * JB_ERR_MEMORY - Out of memory
968 * JB_ERR_PARSE - Cannot parse regex
970 *********************************************************************/
971 static jb_err compile_host_pattern(struct url_spec *url, const char *host_pattern)
980 if (host_pattern[strlen(host_pattern) - 1] == '.')
982 url->unanchored |= ANCHOR_RIGHT;
984 if (host_pattern[0] == '.')
986 url->unanchored |= ANCHOR_LEFT;
990 * Split domain into components
992 url->dbuffer = strdup(host_pattern);
993 if (NULL == url->dbuffer)
996 return JB_ERR_MEMORY;
1002 for (p = url->dbuffer; *p ; p++)
1004 *p = (char)tolower((int)(unsigned char)*p);
1008 * Split the domain name into components
1010 url->dcount = ssplit(url->dbuffer, ".", v, SZ(v), 1, 1);
1012 if (url->dcount < 0)
1015 return JB_ERR_MEMORY;
1017 else if (url->dcount != 0)
1020 * Save a copy of the pointers in dvec
1022 size = (size_t)url->dcount * sizeof(*url->dvec);
1024 url->dvec = (char **)malloc(size);
1025 if (NULL == url->dvec)
1028 return JB_ERR_MEMORY;
1031 memcpy(url->dvec, v, size);
1034 * else dcount == 0 in which case we needn't do anything,
1035 * since dvec will never be accessed and the pattern will
1036 * match all domains.
1042 /*********************************************************************
1044 * Function : simplematch
1046 * Description : String matching, with a (greedy) '*' wildcard that
1047 * stands for zero or more arbitrary characters and
1048 * character classes in [], which take both enumerations
1052 * 1 : pattern = pattern for matching
1053 * 2 : text = text to be matched
1055 * Returns : 0 if match, else nonzero
1057 *********************************************************************/
1058 static int simplematch(const char *pattern, const char *text)
1060 const unsigned char *pat = (const unsigned char *)pattern;
1061 const unsigned char *txt = (const unsigned char *)text;
1062 const unsigned char *fallback = pat;
1065 unsigned char lastchar = 'a';
1067 unsigned char charmap[32];
1072 /* EOF pattern but !EOF text? */
1085 /* '*' in the pattern? */
1089 /* The pattern ends afterwards? Speed up the return. */
1095 /* Else, set wildcard mode and remember position after '*' */
1100 /* Character range specification? */
1103 memset(charmap, '\0', sizeof(charmap));
1105 while (*++pat != ']')
1111 else if (*pat == '-')
1113 if ((*++pat == ']') || *pat == '\0')
1117 for (i = lastchar; i <= *pat; i++)
1119 charmap[i / 8] |= (unsigned char)(1 << (i % 8));
1124 charmap[*pat / 8] |= (unsigned char)(1 << (*pat % 8));
1128 } /* -END- if Character range specification */
1132 * Char match, or char range match?
1136 || ((*pat == ']') && (charmap[*txt / 8] & (1 << (*txt % 8)))) )
1146 * No match && no wildcard: No luck
1150 else if (pat != fallback)
1153 * Increment text pointer if in char range matching
1160 * Wildcard mode && nonmatch beyond fallback: Rewind pattern
1164 * Restart matching from current text pointer
1171 /* Cut off extra '*'s */
1172 if(*pat == '*') pat++;
1174 /* If this is the pattern's end, fine! */
1180 /*********************************************************************
1182 * Function : simple_domaincmp
1184 * Description : Domain-wise Compare fqdn's. The comparison is
1185 * both left- and right-anchored. The individual
1186 * domain names are compared with simplematch().
1187 * This is only used by domain_match.
1190 * 1 : pv = array of patterns to compare
1191 * 2 : fv = array of domain components to compare
1192 * 3 : len = length of the arrays (both arrays are the
1193 * same length - if they weren't, it couldn't
1194 * possibly be a match).
1196 * Returns : 0 => domains are equivalent, else no match.
1198 *********************************************************************/
1199 static int simple_domaincmp(char **pv, char **fv, int len)
1203 for (n = 0; n < len; n++)
1205 if (simplematch(pv[n], fv[n]))
1216 /*********************************************************************
1218 * Function : domain_match
1220 * Description : Domain-wise Compare fqdn's. Governed by the bimap in
1221 * pattern->unachored, the comparison is un-, left-,
1222 * right-anchored, or both.
1223 * The individual domain names are compared with
1227 * 1 : pattern = a domain that may contain a '*' as a wildcard.
1228 * 2 : fqdn = domain name against which the patterns are compared.
1230 * Returns : 0 => domains are equivalent, else no match.
1232 *********************************************************************/
1233 static int domain_match(const struct url_spec *pattern, const struct http_request *fqdn)
1235 char **pv, **fv; /* vectors */
1237 int unanchored = pattern->unanchored & (ANCHOR_RIGHT | ANCHOR_LEFT);
1239 plen = pattern->dcount;
1240 flen = fqdn->dcount;
1244 /* fqdn is too short to match this pattern */
1251 if (unanchored == ANCHOR_LEFT)
1256 * Convert this into a fully anchored pattern with
1257 * the fqdn and pattern the same length
1259 fv += (flen - plen); /* flen - plen >= 0 due to check above */
1260 return simple_domaincmp(pv, fv, plen);
1262 else if (unanchored == 0)
1264 /* Fully anchored, check length */
1269 return simple_domaincmp(pv, fv, plen);
1271 else if (unanchored == ANCHOR_RIGHT)
1273 /* Left anchored, ignore all extra in fqdn */
1274 return simple_domaincmp(pv, fv, plen);
1280 int maxn = flen - plen;
1281 for (n = 0; n <= maxn; n++)
1283 if (!simple_domaincmp(pv, fv, plen))
1288 * Doesn't match from start of fqdn
1289 * Try skipping first part of fqdn
1297 #endif /* def FEATURE_EXTENDED_HOST_PATTERNS */
1300 /*********************************************************************
1302 * Function : create_url_spec
1304 * Description : Creates a "url_spec" structure from a string.
1305 * When finished, free with free_url_spec().
1308 * 1 : url = Target url_spec to be filled in. Will be
1309 * zeroed before use.
1310 * 2 : buf = Source pattern, null terminated. NOTE: The
1311 * contents of this buffer are destroyed by this
1312 * function. If this function succeeds, the
1313 * buffer is copied to url->spec. If this
1314 * function fails, the contents of the buffer
1317 * Returns : JB_ERR_OK - Success
1318 * JB_ERR_MEMORY - Out of memory
1319 * JB_ERR_PARSE - Cannot parse regex (Detailed message
1320 * written to system log)
1322 *********************************************************************/
1323 jb_err create_url_spec(struct url_spec *url, char *buf)
1328 memset(url, '\0', sizeof(*url));
1330 /* Remember the original specification for the CGI pages. */
1331 url->spec = strdup(buf);
1332 if (NULL == url->spec)
1334 return JB_ERR_MEMORY;
1337 /* Is it tag pattern? */
1338 if (0 == strncmpic("TAG:", url->spec, 4))
1340 /* The pattern starts with the first character after "TAG:" */
1341 const char *tag_pattern = buf + 4;
1342 return compile_pattern(tag_pattern, NO_ANCHORING, url, &url->tag_regex);
1345 /* If it isn't a tag pattern it must be a URL pattern. */
1346 return compile_url_pattern(url, buf);
1350 /*********************************************************************
1352 * Function : free_url_spec
1354 * Description : Called from the "unloaders". Freez the url
1355 * structure elements.
1358 * 1 : url = pointer to a url_spec structure.
1362 *********************************************************************/
1363 void free_url_spec(struct url_spec *url)
1365 if (url == NULL) return;
1368 #ifdef FEATURE_EXTENDED_HOST_PATTERNS
1369 if (url->host_regex)
1371 regfree(url->host_regex);
1372 freez(url->host_regex);
1375 freez(url->dbuffer);
1378 #endif /* ndef FEATURE_EXTENDED_HOST_PATTERNS */
1379 freez(url->port_list);
1387 regfree(url->tag_regex);
1388 freez(url->tag_regex);
1393 /*********************************************************************
1395 * Function : url_match
1397 * Description : Compare a URL against a URL pattern.
1400 * 1 : pattern = a URL pattern
1401 * 2 : url = URL to match
1403 * Returns : Nonzero if the URL matches the pattern, else 0.
1405 *********************************************************************/
1406 int url_match(const struct url_spec *pattern,
1407 const struct http_request *http)
1409 /* XXX: these should probably be functions. */
1410 #define PORT_MATCHES ((NULL == pattern->port_list) || match_portlist(pattern->port_list, http->port))
1411 #ifdef FEATURE_EXTENDED_HOST_PATTERNS
1412 #define DOMAIN_MATCHES ((NULL == pattern->host_regex) || (0 == regexec(pattern->host_regex, http->host, 0, NULL, 0)))
1414 #define DOMAIN_MATCHES ((NULL == pattern->dbuffer) || (0 == domain_match(pattern, http)))
1416 #define PATH_MATCHES ((NULL == pattern->preg) || (0 == regexec(pattern->preg, http->path, 0, NULL, 0)))
1418 if (pattern->tag_regex != NULL)
1420 /* It's a tag pattern and shouldn't be matched against URLs */
1424 return (PORT_MATCHES && DOMAIN_MATCHES && PATH_MATCHES);
1429 /*********************************************************************
1431 * Function : match_portlist
1433 * Description : Check if a given number is covered by a comma
1434 * separated list of numbers and ranges (a,b-c,d,..)
1437 * 1 : portlist = String with list
1438 * 2 : port = port to check
1440 * Returns : 0 => no match
1443 *********************************************************************/
1444 int match_portlist(const char *portlist, int port)
1446 char *min, *max, *next, *portlist_copy;
1448 min = next = portlist_copy = strdup(portlist);
1451 * Zero-terminate first item and remember offset for next
1453 if (NULL != (next = strchr(portlist_copy, (int) ',')))
1459 * Loop through all items, checking for match
1463 if (NULL == (max = strchr(min, (int) '-')))
1466 * No dash, check for equality
1468 if (port == atoi(min))
1470 freez(portlist_copy);
1477 * This is a range, so check if between min and max,
1478 * or, if max was omitted, between min and 65K
1481 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
1483 freez(portlist_copy);
1495 * Zero-terminate next item and remember offset for n+1
1497 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
1503 freez(portlist_copy);