1 const char urlmatch_rcs[] = "$Id: urlmatch.c,v 1.47 2009/03/02 19:18:10 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 1.47 2009/03/02 19:18:10 fabiankeil
37 * Streamline parse_http_request()'s prototype. As
38 * cparser pointed out it doesn't actually use csp.
40 * Revision 1.46 2009/02/11 19:31:32 fabiankeil
41 * Reject request lines that end with neither HTTP/1.0 nor HTTP/1.1.
43 * Revision 1.45 2008/06/21 21:19:18 fabiankeil
44 * Silence bogus compiler warning.
46 * Revision 1.44 2008/05/04 16:18:32 fabiankeil
47 * Provide parse_http_url() with a third parameter to specify
48 * whether or not URLs without protocol are acceptable.
50 * Revision 1.43 2008/05/04 13:30:55 fabiankeil
51 * Streamline parse_http_url()'s prototype.
53 * Revision 1.42 2008/05/04 13:24:16 fabiankeil
54 * If the method isn't CONNECT, reject URLs without protocol.
56 * Revision 1.41 2008/05/02 09:51:34 fabiankeil
57 * In parse_http_url(), don't muck around with values
58 * that are none of its business: require an initialized
59 * http structure and never unset http->ssl.
61 * Revision 1.40 2008/04/23 16:12:28 fabiankeil
64 * Revision 1.39 2008/04/22 16:27:42 fabiankeil
65 * In parse_http_request(), remove a pointless
66 * temporary variable and free the buffer earlier.
68 * Revision 1.38 2008/04/18 05:17:18 fabiankeil
69 * Mark simplematch()'s parameters as immutable.
71 * Revision 1.37 2008/04/17 14:53:29 fabiankeil
72 * Move simplematch() into urlmatch.c as it's only
73 * used to match (old-school) domain patterns.
75 * Revision 1.36 2008/04/14 18:19:48 fabiankeil
76 * Remove now-pointless cast in create_url_spec().
78 * Revision 1.35 2008/04/14 18:11:21 fabiankeil
79 * The compiler might not notice it, but the buffer passed to
80 * create_url_spec() is modified later on and thus shouldn't
81 * be declared immutable.
83 * Revision 1.34 2008/04/13 13:32:07 fabiankeil
84 * Factor URL pattern compilation out of create_url_spec().
86 * Revision 1.33 2008/04/12 14:03:13 fabiankeil
87 * Remove an obvious comment and improve another one.
89 * Revision 1.32 2008/04/12 12:38:06 fabiankeil
90 * Factor out duplicated code to compile host, path and tag patterns.
92 * Revision 1.31 2008/04/10 14:41:04 fabiankeil
93 * Ditch url_spec's path member now that it's no longer used.
95 * Revision 1.30 2008/04/10 04:24:24 fabiankeil
96 * Stop duplicating the plain text representation of the path regex
97 * (and keeping the copy around). Once the regex is compiled it's no
100 * Revision 1.29 2008/04/10 04:17:56 fabiankeil
101 * In url_match(), check the right member for NULL when determining
102 * whether there's a path regex to execute. Looking for a plain-text
103 * representation works as well, but it looks "interesting" and that
104 * member will be removed soonish anyway.
106 * Revision 1.28 2008/04/08 16:07:39 fabiankeil
107 * Make it harder to mistake url_match()'s
108 * second parameter for an url_spec.
110 * Revision 1.27 2008/04/08 15:44:33 fabiankeil
111 * Save a bit of memory (and a few cpu cycles) by not bothering to
112 * compile slash-only path regexes that don't affect the result.
114 * Revision 1.26 2008/04/07 16:57:18 fabiankeil
115 * - Use free_url_spec() more consistently.
116 * - Let it reset url->dcount just in case.
118 * Revision 1.25 2008/04/06 15:18:38 fabiankeil
119 * Oh well, rename the --enable-pcre-host-patterns option to
120 * --enable-extended-host-patterns as it's not really PCRE syntax.
122 * Revision 1.24 2008/04/06 14:54:26 fabiankeil
123 * Use PCRE syntax in host patterns when configured
124 * with --enable-pcre-host-patterns.
126 * Revision 1.23 2008/04/05 12:19:20 fabiankeil
127 * Factor compile_host_pattern() out of create_url_spec().
129 * Revision 1.22 2008/03/30 15:02:32 fabiankeil
130 * SZitify unknown_method().
132 * Revision 1.21 2007/12/24 16:34:23 fabiankeil
133 * Band-aid (and micro-optimization) that makes it less likely to run out of
134 * stack space with overly-complex path patterns. Probably masks the problem
135 * reported by Lee in #1856679. Hohoho.
137 * Revision 1.20 2007/09/02 15:31:20 fabiankeil
138 * Move match_portlist() from filter.c to urlmatch.c.
139 * It's used for url matching, not for filtering.
141 * Revision 1.19 2007/09/02 13:42:11 fabiankeil
142 * - Allow port lists in url patterns.
143 * - Ditch unused url_spec member pathlen.
145 * Revision 1.18 2007/07/30 16:42:21 fabiankeil
146 * Move the method check into unknown_method()
147 * and loop through the known methods instead
148 * of using a screen-long OR chain.
150 * Revision 1.17 2007/04/15 16:39:21 fabiankeil
151 * Introduce tags as alternative way to specify which
152 * actions apply to a request. At the moment tags can be
153 * created based on client and server headers.
155 * Revision 1.16 2007/02/13 13:59:24 fabiankeil
156 * Remove redundant log message.
158 * Revision 1.15 2007/01/28 16:11:23 fabiankeil
159 * Accept WebDAV methods for subversion
160 * in parse_http_request(). Closes FR 1581425.
162 * Revision 1.14 2007/01/06 14:23:56 fabiankeil
163 * Fix gcc43 warnings. Mark *csp as immutable
164 * for parse_http_url() and url_match().
165 * Replace a sprintf call with snprintf.
167 * Revision 1.13 2006/12/06 19:50:54 fabiankeil
168 * parse_http_url() now handles intercepted
169 * HTTP request lines as well. Moved parts
170 * of parse_http_url()'s code into
171 * init_domain_components() so that it can
172 * be reused in chat().
174 * Revision 1.12 2006/07/18 14:48:47 david__schmidt
175 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
176 * with what was really the latest development (the v_3_0_branch branch)
178 * Revision 1.10.2.7 2003/05/17 15:57:24 oes
179 * - parse_http_url now checks memory allocation failure for
180 * duplication of "*" URL and rejects "*something" URLs
182 * - Added a comment to what might look like a bug in
183 * create_url_spec (see !bug #736931)
184 * - Comment cosmetics
186 * Revision 1.10.2.6 2003/05/07 12:39:48 oes
187 * Fix typo: Default port for https URLs is 443, not 143.
188 * Thanks to Scott Tregear for spotting this one.
190 * Revision 1.10.2.5 2003/02/28 13:09:29 oes
191 * Fixed a rare double free condition as per Bug #694713
193 * Revision 1.10.2.4 2003/02/28 12:57:44 oes
194 * Moved freeing of http request structure to its owner
195 * as per Dan Price's observations in Bug #694713
197 * Revision 1.10.2.3 2002/11/12 16:50:40 oes
198 * Fixed memory leak in parse_http_request() reported by Oliver Stoeneberg. Fixes bug #637073
200 * Revision 1.10.2.2 2002/09/25 14:53:15 oes
201 * Added basic support for OPTIONS and TRACE HTTP methods:
202 * parse_http_url now recognizes the "*" URI as well as
203 * the OPTIONS and TRACE method keywords.
205 * Revision 1.10.2.1 2002/06/06 19:06:44 jongfoster
206 * Adding support for proprietary Microsoft WebDAV extensions
208 * Revision 1.10 2002/05/12 21:40:37 jongfoster
209 * - Removing some unused code
211 * Revision 1.9 2002/04/04 00:36:36 gliptak
212 * always use pcre for matching
214 * Revision 1.8 2002/04/03 23:32:47 jongfoster
215 * Fixing memory leak on error
217 * Revision 1.7 2002/03/26 22:29:55 swa
218 * we have a new homepage!
220 * Revision 1.6 2002/03/24 13:25:43 swa
221 * name change related issues
223 * Revision 1.5 2002/03/13 00:27:05 jongfoster
226 * Revision 1.4 2002/03/07 03:46:17 oes
227 * Fixed compiler warnings
229 * Revision 1.3 2002/03/03 14:51:11 oes
230 * Fixed CLF logging: Added ocmd member for client's request to struct http_request
232 * Revision 1.2 2002/01/21 00:14:09 jongfoster
233 * Correcting comment style
234 * Fixing an uninitialized memory bug in create_url_spec()
236 * Revision 1.1 2002/01/17 20:53:46 jongfoster
237 * Moving all our URL and URL pattern parsing code to the same file - it
238 * was scattered around in filters.c, loaders.c and parsers.c.
240 * Providing a single, simple url_match(pattern,url) function - rather than
241 * the 3-line match routine which was repeated all over the place.
243 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
245 * Providing parse_http_url() so that URLs can be parsed without faking a
246 * HTTP request line for parse_http_request() or repeating the parsing
247 * code (both of which were techniques that were actually in use).
249 * Standardizing that struct http_request is used to represent a URL, and
250 * struct url_spec is used to represent a URL pattern. (Before, URLs were
251 * represented as seperate variables and a partially-filled-in url_spec).
254 *********************************************************************/
261 #include <sys/types.h>
269 #if !defined(_WIN32) && !defined(__OS2__)
274 #include "urlmatch.h"
276 #include "miscutil.h"
279 const char urlmatch_h_rcs[] = URLMATCH_H_VERSION;
281 enum regex_anchoring {NO_ANCHORING, LEFT_ANCHORED, RIGHT_ANCHORED};
282 static jb_err compile_host_pattern(struct url_spec *url, const char *host_pattern);
284 /*********************************************************************
286 * Function : free_http_request
288 * Description : Freez a http_request structure
291 * 1 : http = points to a http_request structure to free
295 *********************************************************************/
296 void free_http_request(struct http_request *http)
305 freez(http->hostport);
308 freez(http->host_ip_addr_str);
309 freez(http->dbuffer);
315 /*********************************************************************
317 * Function : init_domain_components
319 * Description : Splits the domain name so we can compare it
320 * against wildcards. It used to be part of
321 * parse_http_url, but was separated because the
322 * same code is required in chat in case of
323 * intercepted requests.
326 * 1 : http = pointer to the http structure to hold elements.
328 * Returns : JB_ERR_OK on success
329 * JB_ERR_MEMORY on out of memory
330 * JB_ERR_PARSE on malformed command/URL
331 * or >100 domains deep.
333 *********************************************************************/
334 jb_err init_domain_components(struct http_request *http)
336 char *vec[BUFFER_SIZE];
340 http->dbuffer = strdup(http->host);
341 if (NULL == http->dbuffer)
343 return JB_ERR_MEMORY;
346 /* map to lower case */
347 for (p = http->dbuffer; *p ; p++)
349 *p = (char)tolower((int)(unsigned char)*p);
352 /* split the domain name into components */
353 http->dcount = ssplit(http->dbuffer, ".", vec, SZ(vec), 1, 1);
355 if (http->dcount <= 0)
358 * Error: More than SZ(vec) components in domain
359 * or: no components in domain
361 log_error(LOG_LEVEL_ERROR, "More than SZ(vec) components in domain or none at all.");
365 /* save a copy of the pointers in dvec */
366 size = (size_t)http->dcount * sizeof(*http->dvec);
368 http->dvec = (char **)malloc(size);
369 if (NULL == http->dvec)
371 return JB_ERR_MEMORY;
374 memcpy(http->dvec, vec, size);
380 /*********************************************************************
382 * Function : parse_http_url
384 * Description : Parse out the host and port from the URL. Find the
385 * hostname & path, port (if ':'), and/or password (if '@')
388 * 1 : url = URL (or is it URI?) to break down
389 * 2 : http = pointer to the http structure to hold elements.
390 * Must be initialized with valid values (like NULLs).
391 * 3 : require_protocol = Whether or not URLs without
392 * protocol are acceptable.
394 * Returns : JB_ERR_OK on success
395 * JB_ERR_MEMORY on out of memory
396 * JB_ERR_PARSE on malformed command/URL
397 * or >100 domains deep.
399 *********************************************************************/
400 jb_err parse_http_url(const char *url, struct http_request *http, int require_protocol)
402 int host_available = 1; /* A proxy can dream. */
405 * Save our initial URL
407 http->url = strdup(url);
408 if (http->url == NULL)
410 return JB_ERR_MEMORY;
415 * Check for * URI. If found, we're done.
417 if (*http->url == '*')
419 if ( NULL == (http->path = strdup("*"))
420 || NULL == (http->hostport = strdup("")) )
422 return JB_ERR_MEMORY;
424 if (http->url[1] != '\0')
433 * Split URL into protocol,hostport,path.
443 return JB_ERR_MEMORY;
446 /* Find the start of the URL in our scratch space */
448 if (strncmpic(url_noproto, "http://", 7) == 0)
452 else if (strncmpic(url_noproto, "https://", 8) == 0)
455 * Should only happen when called from cgi_show_url_info().
460 else if (*url_noproto == '/')
463 * Short request line without protocol and host.
464 * Most likely because the client's request
465 * was intercepted and redirected into Privoxy.
470 else if (require_protocol)
476 url_path = strchr(url_noproto, '/');
477 if (url_path != NULL)
482 * NOTE: The following line ignores the path for HTTPS URLS.
483 * This means that you get consistent behaviour if you type a
484 * https URL in and it's parsed by the function. (When the
485 * URL is actually retrieved, SSL hides the path part).
487 http->path = strdup(http->ssl ? "/" : url_path);
489 http->hostport = strdup(url_noproto);
494 * Repair broken HTTP requests that don't contain a path,
495 * or CONNECT requests
497 http->path = strdup("/");
498 http->hostport = strdup(url_noproto);
503 if ( (http->path == NULL)
504 || (http->hostport == NULL))
506 return JB_ERR_MEMORY;
512 /* Without host, there is nothing left to do here */
517 * Split hostport into user/password (ignored), host, port.
524 buf = strdup(http->hostport);
527 return JB_ERR_MEMORY;
530 /* check if url contains username and/or password */
531 host = strchr(buf, '@');
534 /* Contains username/password, skip it and the @ sign. */
539 /* No username or password. */
543 /* Move after hostname before port number */
546 /* Numeric IPv6 address delimited by brackets */
548 port = strchr(host, ']');
552 /* Missing closing bracket */
563 else if (*port != ':')
565 /* Garbage after closing bracket */
572 /* Plain non-escaped hostname */
573 port = strchr(host, ':');
576 /* check if url contains port */
580 /* Terminate hostname and point to start of port string */
582 http->port = atoi(port);
586 /* No port specified. */
587 http->port = (http->ssl ? 443 : 80);
590 http->host = strdup(host);
594 if (http->host == NULL)
596 return JB_ERR_MEMORY;
601 * Split domain name so we can compare it against wildcards
603 return init_domain_components(http);
608 /*********************************************************************
610 * Function : unknown_method
612 * Description : Checks whether a method is unknown.
615 * 1 : method = points to a http method
617 * Returns : TRUE if it's unknown, FALSE otherwise.
619 *********************************************************************/
620 static int unknown_method(const char *method)
622 static const char *known_http_methods[] = {
623 /* Basic HTTP request type */
624 "GET", "HEAD", "POST", "PUT", "DELETE", "OPTIONS", "TRACE", "CONNECT",
625 /* webDAV extensions (RFC2518) */
626 "PROPFIND", "PROPPATCH", "MOVE", "COPY", "MKCOL", "LOCK", "UNLOCK",
628 * Microsoft webDAV extension for Exchange 2000. See:
629 * http://lists.w3.org/Archives/Public/w3c-dist-auth/2002JanMar/0001.html
630 * http://msdn.microsoft.com/library/en-us/wss/wss/_webdav_methods.asp
632 "BCOPY", "BMOVE", "BDELETE", "BPROPFIND", "BPROPPATCH",
634 * Another Microsoft webDAV extension for Exchange 2000. See:
635 * http://systems.cs.colorado.edu/grunwald/MobileComputing/Papers/draft-cohen-gena-p-base-00.txt
636 * http://lists.w3.org/Archives/Public/w3c-dist-auth/2002JanMar/0001.html
637 * http://msdn.microsoft.com/library/en-us/wss/wss/_webdav_methods.asp
639 "SUBSCRIBE", "UNSUBSCRIBE", "NOTIFY", "POLL",
641 * Yet another WebDAV extension, this time for
642 * Web Distributed Authoring and Versioning (RFC3253)
644 "VERSION-CONTROL", "REPORT", "CHECKOUT", "CHECKIN", "UNCHECKOUT",
645 "MKWORKSPACE", "UPDATE", "LABEL", "MERGE", "BASELINE-CONTROL", "MKACTIVITY",
649 for (i = 0; i < SZ(known_http_methods); i++)
651 if (0 == strcmpic(method, known_http_methods[i]))
662 /*********************************************************************
664 * Function : parse_http_request
666 * Description : Parse out the host and port from the URL. Find the
667 * hostname & path, port (if ':'), and/or password (if '@')
670 * 1 : req = HTTP request line to break down
671 * 2 : http = pointer to the http structure to hold elements
673 * Returns : JB_ERR_OK on success
674 * JB_ERR_MEMORY on out of memory
675 * JB_ERR_CGI_PARAMS on malformed command/URL
676 * or >100 domains deep.
678 *********************************************************************/
679 jb_err parse_http_request(const char *req, struct http_request *http)
682 char *v[10]; /* XXX: Why 10? We should only need three. */
686 memset(http, '\0', sizeof(*http));
691 return JB_ERR_MEMORY;
694 n = ssplit(buf, " \r\n", v, SZ(v), 1, 1);
702 * Fail in case of unknown methods
703 * which we might not handle correctly.
705 * XXX: There should be a config option
706 * to forward requests with unknown methods
707 * anyway. Most of them don't need special
710 if (unknown_method(v[0]))
712 log_error(LOG_LEVEL_ERROR, "Unknown HTTP method detected: %s", v[0]);
717 if (strcmpic(v[2], "HTTP/1.1") && strcmpic(v[2], "HTTP/1.0"))
719 log_error(LOG_LEVEL_ERROR, "The only supported HTTP "
720 "versions are 1.0 and 1.1. This rules out: %s", v[2]);
725 http->ssl = !strcmpic(v[0], "CONNECT");
727 err = parse_http_url(v[1], http, !http->ssl);
735 * Copy the details into the structure
737 http->cmd = strdup(req);
738 http->gpc = strdup(v[0]);
739 http->ver = strdup(v[2]);
743 if ( (http->cmd == NULL)
744 || (http->gpc == NULL)
745 || (http->ver == NULL) )
747 return JB_ERR_MEMORY;
755 /*********************************************************************
757 * Function : compile_pattern
759 * Description : Compiles a host, domain or TAG pattern.
762 * 1 : pattern = The pattern to compile.
763 * 2 : anchoring = How the regex should be anchored.
764 * Can be either one of NO_ANCHORING,
765 * LEFT_ANCHORED or RIGHT_ANCHORED.
766 * 3 : url = In case of failures, the spec member is
767 * logged and the structure freed.
768 * 4 : regex = Where the compiled regex should be stored.
770 * Returns : JB_ERR_OK - Success
771 * JB_ERR_MEMORY - Out of memory
772 * JB_ERR_PARSE - Cannot parse regex
774 *********************************************************************/
775 static jb_err compile_pattern(const char *pattern, enum regex_anchoring anchoring,
776 struct url_spec *url, regex_t **regex)
779 char rebuf[BUFFER_SIZE];
780 const char *fmt = NULL;
783 assert(strlen(pattern) < sizeof(rebuf) - 2);
785 if (pattern[0] == '\0')
803 log_error(LOG_LEVEL_FATAL,
804 "Invalid anchoring in compile_pattern %d", anchoring);
807 *regex = zalloc(sizeof(**regex));
811 return JB_ERR_MEMORY;
814 snprintf(rebuf, sizeof(rebuf), fmt, pattern);
816 errcode = regcomp(*regex, rebuf, (REG_EXTENDED|REG_NOSUB|REG_ICASE));
820 size_t errlen = regerror(errcode, *regex, rebuf, sizeof(rebuf));
821 if (errlen > (sizeof(rebuf) - (size_t)1))
823 errlen = sizeof(rebuf) - (size_t)1;
825 rebuf[errlen] = '\0';
826 log_error(LOG_LEVEL_ERROR, "error compiling %s from %s: %s",
827 pattern, url->spec, rebuf);
838 /*********************************************************************
840 * Function : compile_url_pattern
842 * Description : Compiles the three parts of an URL pattern.
845 * 1 : url = Target url_spec to be filled in.
846 * 2 : buf = The url pattern to compile. Will be messed up.
848 * Returns : JB_ERR_OK - Success
849 * JB_ERR_MEMORY - Out of memory
850 * JB_ERR_PARSE - Cannot parse regex
852 *********************************************************************/
853 static jb_err compile_url_pattern(struct url_spec *url, char *buf)
857 p = strchr(buf, '/');
861 * Only compile the regex if it consists of more than
862 * a single slash, otherwise it wouldn't affect the result.
867 * XXX: does it make sense to compile the slash at the beginning?
869 jb_err err = compile_pattern(p, LEFT_ANCHORED, url, &url->preg);
871 if (JB_ERR_OK != err)
879 /* XXX: IPv6 numeric hostname contains colons, thus we need to delimit the
880 * hostname before real port separator. Because brackets are used in
881 * hostname matching on lower layer, we can't use it. I decided to use
882 * angle brackets '<' '>' instead. */
883 if (buf[0] == '<' && NULL != (p = strchr(buf + 1, '>')))
890 /* Only IPv6 address without port number */
895 /* Garbage after address delimiter */
901 p = strchr(buf, ':');
907 url->port_list = strdup(p);
908 if (NULL == url->port_list)
910 return JB_ERR_MEMORY;
915 url->port_list = NULL;
920 return compile_host_pattern(url, buf);
928 #ifdef FEATURE_EXTENDED_HOST_PATTERNS
929 /*********************************************************************
931 * Function : compile_host_pattern
933 * Description : Parses and compiles a host pattern..
936 * 1 : url = Target url_spec to be filled in.
937 * 2 : host_pattern = Host pattern to compile.
939 * Returns : JB_ERR_OK - Success
940 * JB_ERR_MEMORY - Out of memory
941 * JB_ERR_PARSE - Cannot parse regex
943 *********************************************************************/
944 static jb_err compile_host_pattern(struct url_spec *url, const char *host_pattern)
946 return compile_pattern(host_pattern, RIGHT_ANCHORED, url, &url->host_regex);
951 /*********************************************************************
953 * Function : compile_host_pattern
955 * Description : Parses and "compiles" an old-school host pattern.
958 * 1 : url = Target url_spec to be filled in.
959 * 2 : host_pattern = Host pattern to parse.
961 * Returns : JB_ERR_OK - Success
962 * JB_ERR_MEMORY - Out of memory
963 * JB_ERR_PARSE - Cannot parse regex
965 *********************************************************************/
966 static jb_err compile_host_pattern(struct url_spec *url, const char *host_pattern)
975 if (host_pattern[strlen(host_pattern) - 1] == '.')
977 url->unanchored |= ANCHOR_RIGHT;
979 if (host_pattern[0] == '.')
981 url->unanchored |= ANCHOR_LEFT;
985 * Split domain into components
987 url->dbuffer = strdup(host_pattern);
988 if (NULL == url->dbuffer)
991 return JB_ERR_MEMORY;
997 for (p = url->dbuffer; *p ; p++)
999 *p = (char)tolower((int)(unsigned char)*p);
1003 * Split the domain name into components
1005 url->dcount = ssplit(url->dbuffer, ".", v, SZ(v), 1, 1);
1007 if (url->dcount < 0)
1010 return JB_ERR_MEMORY;
1012 else if (url->dcount != 0)
1015 * Save a copy of the pointers in dvec
1017 size = (size_t)url->dcount * sizeof(*url->dvec);
1019 url->dvec = (char **)malloc(size);
1020 if (NULL == url->dvec)
1023 return JB_ERR_MEMORY;
1026 memcpy(url->dvec, v, size);
1029 * else dcount == 0 in which case we needn't do anything,
1030 * since dvec will never be accessed and the pattern will
1031 * match all domains.
1037 /*********************************************************************
1039 * Function : simplematch
1041 * Description : String matching, with a (greedy) '*' wildcard that
1042 * stands for zero or more arbitrary characters and
1043 * character classes in [], which take both enumerations
1047 * 1 : pattern = pattern for matching
1048 * 2 : text = text to be matched
1050 * Returns : 0 if match, else nonzero
1052 *********************************************************************/
1053 static int simplematch(const char *pattern, const char *text)
1055 const unsigned char *pat = (const unsigned char *)pattern;
1056 const unsigned char *txt = (const unsigned char *)text;
1057 const unsigned char *fallback = pat;
1060 unsigned char lastchar = 'a';
1062 unsigned char charmap[32];
1067 /* EOF pattern but !EOF text? */
1080 /* '*' in the pattern? */
1084 /* The pattern ends afterwards? Speed up the return. */
1090 /* Else, set wildcard mode and remember position after '*' */
1095 /* Character range specification? */
1098 memset(charmap, '\0', sizeof(charmap));
1100 while (*++pat != ']')
1106 else if (*pat == '-')
1108 if ((*++pat == ']') || *pat == '\0')
1112 for (i = lastchar; i <= *pat; i++)
1114 charmap[i / 8] |= (unsigned char)(1 << (i % 8));
1119 charmap[*pat / 8] |= (unsigned char)(1 << (*pat % 8));
1123 } /* -END- if Character range specification */
1127 * Char match, or char range match?
1131 || ((*pat == ']') && (charmap[*txt / 8] & (1 << (*txt % 8)))) )
1141 * No match && no wildcard: No luck
1145 else if (pat != fallback)
1148 * Increment text pointer if in char range matching
1155 * Wildcard mode && nonmatch beyond fallback: Rewind pattern
1159 * Restart matching from current text pointer
1166 /* Cut off extra '*'s */
1167 if(*pat == '*') pat++;
1169 /* If this is the pattern's end, fine! */
1175 /*********************************************************************
1177 * Function : simple_domaincmp
1179 * Description : Domain-wise Compare fqdn's. The comparison is
1180 * both left- and right-anchored. The individual
1181 * domain names are compared with simplematch().
1182 * This is only used by domain_match.
1185 * 1 : pv = array of patterns to compare
1186 * 2 : fv = array of domain components to compare
1187 * 3 : len = length of the arrays (both arrays are the
1188 * same length - if they weren't, it couldn't
1189 * possibly be a match).
1191 * Returns : 0 => domains are equivalent, else no match.
1193 *********************************************************************/
1194 static int simple_domaincmp(char **pv, char **fv, int len)
1198 for (n = 0; n < len; n++)
1200 if (simplematch(pv[n], fv[n]))
1211 /*********************************************************************
1213 * Function : domain_match
1215 * Description : Domain-wise Compare fqdn's. Governed by the bimap in
1216 * pattern->unachored, the comparison is un-, left-,
1217 * right-anchored, or both.
1218 * The individual domain names are compared with
1222 * 1 : pattern = a domain that may contain a '*' as a wildcard.
1223 * 2 : fqdn = domain name against which the patterns are compared.
1225 * Returns : 0 => domains are equivalent, else no match.
1227 *********************************************************************/
1228 static int domain_match(const struct url_spec *pattern, const struct http_request *fqdn)
1230 char **pv, **fv; /* vectors */
1232 int unanchored = pattern->unanchored & (ANCHOR_RIGHT | ANCHOR_LEFT);
1234 plen = pattern->dcount;
1235 flen = fqdn->dcount;
1239 /* fqdn is too short to match this pattern */
1246 if (unanchored == ANCHOR_LEFT)
1251 * Convert this into a fully anchored pattern with
1252 * the fqdn and pattern the same length
1254 fv += (flen - plen); /* flen - plen >= 0 due to check above */
1255 return simple_domaincmp(pv, fv, plen);
1257 else if (unanchored == 0)
1259 /* Fully anchored, check length */
1264 return simple_domaincmp(pv, fv, plen);
1266 else if (unanchored == ANCHOR_RIGHT)
1268 /* Left anchored, ignore all extra in fqdn */
1269 return simple_domaincmp(pv, fv, plen);
1275 int maxn = flen - plen;
1276 for (n = 0; n <= maxn; n++)
1278 if (!simple_domaincmp(pv, fv, plen))
1283 * Doesn't match from start of fqdn
1284 * Try skipping first part of fqdn
1292 #endif /* def FEATURE_EXTENDED_HOST_PATTERNS */
1295 /*********************************************************************
1297 * Function : create_url_spec
1299 * Description : Creates a "url_spec" structure from a string.
1300 * When finished, free with free_url_spec().
1303 * 1 : url = Target url_spec to be filled in. Will be
1304 * zeroed before use.
1305 * 2 : buf = Source pattern, null terminated. NOTE: The
1306 * contents of this buffer are destroyed by this
1307 * function. If this function succeeds, the
1308 * buffer is copied to url->spec. If this
1309 * function fails, the contents of the buffer
1312 * Returns : JB_ERR_OK - Success
1313 * JB_ERR_MEMORY - Out of memory
1314 * JB_ERR_PARSE - Cannot parse regex (Detailed message
1315 * written to system log)
1317 *********************************************************************/
1318 jb_err create_url_spec(struct url_spec *url, char *buf)
1323 memset(url, '\0', sizeof(*url));
1325 /* Remember the original specification for the CGI pages. */
1326 url->spec = strdup(buf);
1327 if (NULL == url->spec)
1329 return JB_ERR_MEMORY;
1332 /* Is it tag pattern? */
1333 if (0 == strncmpic("TAG:", url->spec, 4))
1335 /* The pattern starts with the first character after "TAG:" */
1336 const char *tag_pattern = buf + 4;
1337 return compile_pattern(tag_pattern, NO_ANCHORING, url, &url->tag_regex);
1340 /* If it isn't a tag pattern it must be a URL pattern. */
1341 return compile_url_pattern(url, buf);
1345 /*********************************************************************
1347 * Function : free_url_spec
1349 * Description : Called from the "unloaders". Freez the url
1350 * structure elements.
1353 * 1 : url = pointer to a url_spec structure.
1357 *********************************************************************/
1358 void free_url_spec(struct url_spec *url)
1360 if (url == NULL) return;
1363 #ifdef FEATURE_EXTENDED_HOST_PATTERNS
1364 if (url->host_regex)
1366 regfree(url->host_regex);
1367 freez(url->host_regex);
1370 freez(url->dbuffer);
1373 #endif /* ndef FEATURE_EXTENDED_HOST_PATTERNS */
1374 freez(url->port_list);
1382 regfree(url->tag_regex);
1383 freez(url->tag_regex);
1388 /*********************************************************************
1390 * Function : url_match
1392 * Description : Compare a URL against a URL pattern.
1395 * 1 : pattern = a URL pattern
1396 * 2 : url = URL to match
1398 * Returns : Nonzero if the URL matches the pattern, else 0.
1400 *********************************************************************/
1401 int url_match(const struct url_spec *pattern,
1402 const struct http_request *http)
1404 /* XXX: these should probably be functions. */
1405 #define PORT_MATCHES ((NULL == pattern->port_list) || match_portlist(pattern->port_list, http->port))
1406 #ifdef FEATURE_EXTENDED_HOST_PATTERNS
1407 #define DOMAIN_MATCHES ((NULL == pattern->host_regex) || (0 == regexec(pattern->host_regex, http->host, 0, NULL, 0)))
1409 #define DOMAIN_MATCHES ((NULL == pattern->dbuffer) || (0 == domain_match(pattern, http)))
1411 #define PATH_MATCHES ((NULL == pattern->preg) || (0 == regexec(pattern->preg, http->path, 0, NULL, 0)))
1413 if (pattern->tag_regex != NULL)
1415 /* It's a tag pattern and shouldn't be matched against URLs */
1419 return (PORT_MATCHES && DOMAIN_MATCHES && PATH_MATCHES);
1424 /*********************************************************************
1426 * Function : match_portlist
1428 * Description : Check if a given number is covered by a comma
1429 * separated list of numbers and ranges (a,b-c,d,..)
1432 * 1 : portlist = String with list
1433 * 2 : port = port to check
1435 * Returns : 0 => no match
1438 *********************************************************************/
1439 int match_portlist(const char *portlist, int port)
1441 char *min, *max, *next, *portlist_copy;
1443 min = next = portlist_copy = strdup(portlist);
1446 * Zero-terminate first item and remember offset for next
1448 if (NULL != (next = strchr(portlist_copy, (int) ',')))
1454 * Loop through all items, checking for match
1458 if (NULL == (max = strchr(min, (int) '-')))
1461 * No dash, check for equality
1463 if (port == atoi(min))
1465 freez(portlist_copy);
1472 * This is a range, so check if between min and max,
1473 * or, if max was omitted, between min and 65K
1476 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
1478 freez(portlist_copy);
1490 * Zero-terminate next item and remember offset for n+1
1492 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
1498 freez(portlist_copy);
1509 vim:softtabstop=3 shiftwidth=3