X-Git-Url: http://www.privoxy.org/gitweb/?p=privoxy.git;a=blobdiff_plain;f=urlmatch.c;h=ce12ecc81ccc6d217de86f29b1588b455415ccd6;hp=1b2aee86f89733797c03cb35383e3daeceee1291;hb=bf18b40dd30e0c16392285408cad379c2ead11d7;hpb=d0c1f6261a9773bd2271117cdde20d734d320a93 diff --git a/urlmatch.c b/urlmatch.c index 1b2aee86..ce12ecc8 100644 --- a/urlmatch.c +++ b/urlmatch.c @@ -1,4 +1,4 @@ -const char urlmatch_rcs[] = "$Id: urlmatch.c,v 1.9 2002/04/04 00:36:36 gliptak Exp $"; +const char urlmatch_rcs[] = "$Id: urlmatch.c,v 1.46 2009/02/11 19:31:32 fabiankeil Exp $"; /********************************************************************* * * File : $Source: /cvsroot/ijbswa/current/urlmatch.c,v $ @@ -6,8 +6,8 @@ const char urlmatch_rcs[] = "$Id: urlmatch.c,v 1.9 2002/04/04 00:36:36 gliptak E * Purpose : Declares functions to match URLs against URL * patterns. * - * Copyright : Written by and Copyright (C) 2001 the SourceForge - * Privoxy team. http://www.privoxy.org/ + * Copyright : Written by and Copyright (C) 2001-2009 + * the Privoxy team. http://www.privoxy.org/ * * Based on the Internet Junkbuster originally written * by and Copyright (C) 1997 Anonymous Coders and @@ -33,6 +33,177 @@ const char urlmatch_rcs[] = "$Id: urlmatch.c,v 1.9 2002/04/04 00:36:36 gliptak E * * Revisions : * $Log: urlmatch.c,v $ + * Revision 1.46 2009/02/11 19:31:32 fabiankeil + * Reject request lines that end with neither HTTP/1.0 nor HTTP/1.1. + * + * Revision 1.45 2008/06/21 21:19:18 fabiankeil + * Silence bogus compiler warning. + * + * Revision 1.44 2008/05/04 16:18:32 fabiankeil + * Provide parse_http_url() with a third parameter to specify + * whether or not URLs without protocol are acceptable. + * + * Revision 1.43 2008/05/04 13:30:55 fabiankeil + * Streamline parse_http_url()'s prototype. + * + * Revision 1.42 2008/05/04 13:24:16 fabiankeil + * If the method isn't CONNECT, reject URLs without protocol. + * + * Revision 1.41 2008/05/02 09:51:34 fabiankeil + * In parse_http_url(), don't muck around with values + * that are none of its business: require an initialized + * http structure and never unset http->ssl. + * + * Revision 1.40 2008/04/23 16:12:28 fabiankeil + * Free with freez(). + * + * Revision 1.39 2008/04/22 16:27:42 fabiankeil + * In parse_http_request(), remove a pointless + * temporary variable and free the buffer earlier. + * + * Revision 1.38 2008/04/18 05:17:18 fabiankeil + * Mark simplematch()'s parameters as immutable. + * + * Revision 1.37 2008/04/17 14:53:29 fabiankeil + * Move simplematch() into urlmatch.c as it's only + * used to match (old-school) domain patterns. + * + * Revision 1.36 2008/04/14 18:19:48 fabiankeil + * Remove now-pointless cast in create_url_spec(). + * + * Revision 1.35 2008/04/14 18:11:21 fabiankeil + * The compiler might not notice it, but the buffer passed to + * create_url_spec() is modified later on and thus shouldn't + * be declared immutable. + * + * Revision 1.34 2008/04/13 13:32:07 fabiankeil + * Factor URL pattern compilation out of create_url_spec(). + * + * Revision 1.33 2008/04/12 14:03:13 fabiankeil + * Remove an obvious comment and improve another one. + * + * Revision 1.32 2008/04/12 12:38:06 fabiankeil + * Factor out duplicated code to compile host, path and tag patterns. + * + * Revision 1.31 2008/04/10 14:41:04 fabiankeil + * Ditch url_spec's path member now that it's no longer used. + * + * Revision 1.30 2008/04/10 04:24:24 fabiankeil + * Stop duplicating the plain text representation of the path regex + * (and keeping the copy around). Once the regex is compiled it's no + * longer useful. + * + * Revision 1.29 2008/04/10 04:17:56 fabiankeil + * In url_match(), check the right member for NULL when determining + * whether there's a path regex to execute. Looking for a plain-text + * representation works as well, but it looks "interesting" and that + * member will be removed soonish anyway. + * + * Revision 1.28 2008/04/08 16:07:39 fabiankeil + * Make it harder to mistake url_match()'s + * second parameter for an url_spec. + * + * Revision 1.27 2008/04/08 15:44:33 fabiankeil + * Save a bit of memory (and a few cpu cycles) by not bothering to + * compile slash-only path regexes that don't affect the result. + * + * Revision 1.26 2008/04/07 16:57:18 fabiankeil + * - Use free_url_spec() more consistently. + * - Let it reset url->dcount just in case. + * + * Revision 1.25 2008/04/06 15:18:38 fabiankeil + * Oh well, rename the --enable-pcre-host-patterns option to + * --enable-extended-host-patterns as it's not really PCRE syntax. + * + * Revision 1.24 2008/04/06 14:54:26 fabiankeil + * Use PCRE syntax in host patterns when configured + * with --enable-pcre-host-patterns. + * + * Revision 1.23 2008/04/05 12:19:20 fabiankeil + * Factor compile_host_pattern() out of create_url_spec(). + * + * Revision 1.22 2008/03/30 15:02:32 fabiankeil + * SZitify unknown_method(). + * + * Revision 1.21 2007/12/24 16:34:23 fabiankeil + * Band-aid (and micro-optimization) that makes it less likely to run out of + * stack space with overly-complex path patterns. Probably masks the problem + * reported by Lee in #1856679. Hohoho. + * + * Revision 1.20 2007/09/02 15:31:20 fabiankeil + * Move match_portlist() from filter.c to urlmatch.c. + * It's used for url matching, not for filtering. + * + * Revision 1.19 2007/09/02 13:42:11 fabiankeil + * - Allow port lists in url patterns. + * - Ditch unused url_spec member pathlen. + * + * Revision 1.18 2007/07/30 16:42:21 fabiankeil + * Move the method check into unknown_method() + * and loop through the known methods instead + * of using a screen-long OR chain. + * + * Revision 1.17 2007/04/15 16:39:21 fabiankeil + * Introduce tags as alternative way to specify which + * actions apply to a request. At the moment tags can be + * created based on client and server headers. + * + * Revision 1.16 2007/02/13 13:59:24 fabiankeil + * Remove redundant log message. + * + * Revision 1.15 2007/01/28 16:11:23 fabiankeil + * Accept WebDAV methods for subversion + * in parse_http_request(). Closes FR 1581425. + * + * Revision 1.14 2007/01/06 14:23:56 fabiankeil + * Fix gcc43 warnings. Mark *csp as immutable + * for parse_http_url() and url_match(). + * Replace a sprintf call with snprintf. + * + * Revision 1.13 2006/12/06 19:50:54 fabiankeil + * parse_http_url() now handles intercepted + * HTTP request lines as well. Moved parts + * of parse_http_url()'s code into + * init_domain_components() so that it can + * be reused in chat(). + * + * Revision 1.12 2006/07/18 14:48:47 david__schmidt + * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch) + * with what was really the latest development (the v_3_0_branch branch) + * + * Revision 1.10.2.7 2003/05/17 15:57:24 oes + * - parse_http_url now checks memory allocation failure for + * duplication of "*" URL and rejects "*something" URLs + * Closes bug #736344 + * - Added a comment to what might look like a bug in + * create_url_spec (see !bug #736931) + * - Comment cosmetics + * + * Revision 1.10.2.6 2003/05/07 12:39:48 oes + * Fix typo: Default port for https URLs is 443, not 143. + * Thanks to Scott Tregear for spotting this one. + * + * Revision 1.10.2.5 2003/02/28 13:09:29 oes + * Fixed a rare double free condition as per Bug #694713 + * + * Revision 1.10.2.4 2003/02/28 12:57:44 oes + * Moved freeing of http request structure to its owner + * as per Dan Price's observations in Bug #694713 + * + * Revision 1.10.2.3 2002/11/12 16:50:40 oes + * Fixed memory leak in parse_http_request() reported by Oliver Stoeneberg. Fixes bug #637073 + * + * Revision 1.10.2.2 2002/09/25 14:53:15 oes + * Added basic support for OPTIONS and TRACE HTTP methods: + * parse_http_url now recognizes the "*" URI as well as + * the OPTIONS and TRACE method keywords. + * + * Revision 1.10.2.1 2002/06/06 19:06:44 jongfoster + * Adding support for proprietary Microsoft WebDAV extensions + * + * Revision 1.10 2002/05/12 21:40:37 jongfoster + * - Removing some unused code + * * Revision 1.9 2002/04/04 00:36:36 gliptak * always use pcre for matching * @@ -103,6 +274,8 @@ const char urlmatch_rcs[] = "$Id: urlmatch.c,v 1.9 2002/04/04 00:36:36 gliptak E const char urlmatch_h_rcs[] = URLMATCH_H_VERSION; +enum regex_anchoring {NO_ANCHORING, LEFT_ANCHORED, RIGHT_ANCHORED}; +static jb_err compile_host_pattern(struct url_spec *url, const char *host_pattern); /********************************************************************* * @@ -135,6 +308,71 @@ void free_http_request(struct http_request *http) } +/********************************************************************* + * + * Function : init_domain_components + * + * Description : Splits the domain name so we can compare it + * against wildcards. It used to be part of + * parse_http_url, but was separated because the + * same code is required in chat in case of + * intercepted requests. + * + * Parameters : + * 1 : http = pointer to the http structure to hold elements. + * + * Returns : JB_ERR_OK on success + * JB_ERR_MEMORY on out of memory + * JB_ERR_PARSE on malformed command/URL + * or >100 domains deep. + * + *********************************************************************/ +jb_err init_domain_components(struct http_request *http) +{ + char *vec[BUFFER_SIZE]; + size_t size; + char *p; + + http->dbuffer = strdup(http->host); + if (NULL == http->dbuffer) + { + return JB_ERR_MEMORY; + } + + /* map to lower case */ + for (p = http->dbuffer; *p ; p++) + { + *p = (char)tolower((int)(unsigned char)*p); + } + + /* split the domain name into components */ + http->dcount = ssplit(http->dbuffer, ".", vec, SZ(vec), 1, 1); + + if (http->dcount <= 0) + { + /* + * Error: More than SZ(vec) components in domain + * or: no components in domain + */ + log_error(LOG_LEVEL_ERROR, "More than SZ(vec) components in domain or none at all."); + return JB_ERR_PARSE; + } + + /* save a copy of the pointers in dvec */ + size = (size_t)http->dcount * sizeof(*http->dvec); + + http->dvec = (char **)malloc(size); + if (NULL == http->dvec) + { + return JB_ERR_MEMORY; + } + + memcpy(http->dvec, vec, size); + + return JB_ERR_OK; +} + + /********************************************************************* * * Function : parse_http_url @@ -145,26 +383,19 @@ void free_http_request(struct http_request *http) * Parameters : * 1 : url = URL (or is it URI?) to break down * 2 : http = pointer to the http structure to hold elements. - * Will be zeroed before use. Note that this - * function sets the http->gpc and http->ver - * members to NULL. - * 3 : csp = Current client state (buffers, headers, etc...) + * Must be initialized with valid values (like NULLs). + * 3 : require_protocol = Whether or not URLs without + * protocol are acceptable. * * Returns : JB_ERR_OK on success * JB_ERR_MEMORY on out of memory - * JB_ERR_CGI_PARAMS on malformed command/URL - * or >100 domains deep. + * JB_ERR_PARSE on malformed command/URL + * or >100 domains deep. * *********************************************************************/ -jb_err parse_http_url(const char * url, - struct http_request *http, - struct client_state *csp) +jb_err parse_http_url(const char *url, struct http_request *http, int require_protocol) { - /* - * Zero out the results structure - */ - memset(http, '\0', sizeof(*http)); - + int host_available = 1; /* A proxy can dream. */ /* * Save our initial URL @@ -176,6 +407,24 @@ jb_err parse_http_url(const char * url, } + /* + * Check for * URI. If found, we're done. + */ + if (*http->url == '*') + { + if ( NULL == (http->path = strdup("*")) + || NULL == (http->hostport = strdup("")) ) + { + return JB_ERR_MEMORY; + } + if (http->url[1] != '\0') + { + return JB_ERR_PARSE; + } + return JB_ERR_OK; + } + + /* * Split URL into protocol,hostport,path. */ @@ -195,16 +444,29 @@ jb_err parse_http_url(const char * url, if (strncmpic(url_noproto, "http://", 7) == 0) { url_noproto += 7; - http->ssl = 0; } else if (strncmpic(url_noproto, "https://", 8) == 0) { + /* + * Should only happen when called from cgi_show_url_info(). + */ url_noproto += 8; http->ssl = 1; } - else + else if (*url_noproto == '/') { - http->ssl = 0; + /* + * Short request line without protocol and host. + * Most likely because the client's request + * was intercepted and redirected into Privoxy. + */ + http->host = NULL; + host_available = 0; + } + else if (require_protocol) + { + freez(buf); + return JB_ERR_PARSE; } url_path = strchr(url_noproto, '/'); @@ -232,17 +494,20 @@ jb_err parse_http_url(const char * url, http->hostport = strdup(url_noproto); } - free(buf); + freez(buf); if ( (http->path == NULL) || (http->hostport == NULL)) { - free(buf); - free_http_request(http); return JB_ERR_MEMORY; } } + if (!host_available) + { + /* Without host, there is nothing left to do here */ + return JB_ERR_OK; + } /* * Split hostport into user/password (ignored), host, port. @@ -255,7 +520,6 @@ jb_err parse_http_url(const char * url, buf = strdup(http->hostport); if (buf == NULL) { - free_http_request(http); return JB_ERR_MEMORY; } @@ -284,70 +548,78 @@ jb_err parse_http_url(const char * url, else { /* No port specified. */ - http->port = (http->ssl ? 143 : 80); + http->port = (http->ssl ? 443 : 80); } http->host = strdup(host); - free(buf); + freez(buf); if (http->host == NULL) { - free_http_request(http); return JB_ERR_MEMORY; } } - /* * Split domain name so we can compare it against wildcards */ - { - char *vec[BUFFER_SIZE]; - size_t size; - char *p; - - http->dbuffer = strdup(http->host); - if (NULL == http->dbuffer) - { - free_http_request(http); - return JB_ERR_MEMORY; - } + return init_domain_components(http); - /* map to lower case */ - for (p = http->dbuffer; *p ; p++) - { - *p = tolower((int)(unsigned char)*p); - } - - /* split the domain name into components */ - http->dcount = ssplit(http->dbuffer, ".", vec, SZ(vec), 1, 1); +} - if (http->dcount <= 0) - { - /* - * Error: More than SZ(vec) components in domain - * or: no components in domain - */ - free_http_request(http); - return JB_ERR_PARSE; - } - /* save a copy of the pointers in dvec */ - size = http->dcount * sizeof(*http->dvec); +/********************************************************************* + * + * Function : unknown_method + * + * Description : Checks whether a method is unknown. + * + * Parameters : + * 1 : method = points to a http method + * + * Returns : TRUE if it's unknown, FALSE otherwise. + * + *********************************************************************/ +static int unknown_method(const char *method) +{ + static const char *known_http_methods[] = { + /* Basic HTTP request type */ + "GET", "HEAD", "POST", "PUT", "DELETE", "OPTIONS", "TRACE", "CONNECT", + /* webDAV extensions (RFC2518) */ + "PROPFIND", "PROPPATCH", "MOVE", "COPY", "MKCOL", "LOCK", "UNLOCK", + /* + * Microsoft webDAV extension for Exchange 2000. See: + * http://lists.w3.org/Archives/Public/w3c-dist-auth/2002JanMar/0001.html + * http://msdn.microsoft.com/library/en-us/wss/wss/_webdav_methods.asp + */ + "BCOPY", "BMOVE", "BDELETE", "BPROPFIND", "BPROPPATCH", + /* + * Another Microsoft webDAV extension for Exchange 2000. See: + * http://systems.cs.colorado.edu/grunwald/MobileComputing/Papers/draft-cohen-gena-p-base-00.txt + * http://lists.w3.org/Archives/Public/w3c-dist-auth/2002JanMar/0001.html + * http://msdn.microsoft.com/library/en-us/wss/wss/_webdav_methods.asp + */ + "SUBSCRIBE", "UNSUBSCRIBE", "NOTIFY", "POLL", + /* + * Yet another WebDAV extension, this time for + * Web Distributed Authoring and Versioning (RFC3253) + */ + "VERSION-CONTROL", "REPORT", "CHECKOUT", "CHECKIN", "UNCHECKOUT", + "MKWORKSPACE", "UPDATE", "LABEL", "MERGE", "BASELINE-CONTROL", "MKACTIVITY", + }; + int i; - http->dvec = (char **)malloc(size); - if (NULL == http->dvec) + for (i = 0; i < SZ(known_http_methods); i++) + { + if (0 == strcmpic(method, known_http_methods[i])) { - free_http_request(http); - return JB_ERR_MEMORY; + return FALSE; } - - memcpy(http->dvec, vec, size); } + return TRUE; - return JB_ERR_OK; } @@ -361,7 +633,6 @@ jb_err parse_http_url(const char * url, * Parameters : * 1 : req = HTTP request line to break down * 2 : http = pointer to the http structure to hold elements - * 3 : csp = Current client state (buffers, headers, etc...) * * Returns : JB_ERR_OK on success * JB_ERR_MEMORY on out of memory @@ -369,15 +640,12 @@ jb_err parse_http_url(const char * url, * or >100 domains deep. * *********************************************************************/ -jb_err parse_http_request(const char *req, - struct http_request *http, - struct client_state *csp) +jb_err parse_http_request(const char *req, struct http_request *http) { char *buf; - char *v[10]; + char *v[10]; /* XXX: Why 10? We should only need three. */ int n; jb_err err; - int is_connect = 0; memset(http, '\0', sizeof(*http)); @@ -390,68 +658,457 @@ jb_err parse_http_request(const char *req, n = ssplit(buf, " \r\n", v, SZ(v), 1, 1); if (n != 3) { - free(buf); + freez(buf); return JB_ERR_PARSE; } - /* this could be a CONNECT request */ - if (strcmpic(v[0], "connect") == 0) - { - /* Secure */ - is_connect = 1; - } - /* or it could be any other basic HTTP request type */ - else if ((0 == strcmpic(v[0], "get")) - || (0 == strcmpic(v[0], "head")) - || (0 == strcmpic(v[0], "post")) - || (0 == strcmpic(v[0], "put")) - || (0 == strcmpic(v[0], "delete")) - - /* or a webDAV extension (RFC2518) */ - || (0 == strcmpic(v[0], "propfind")) - || (0 == strcmpic(v[0], "proppatch")) - || (0 == strcmpic(v[0], "move")) - || (0 == strcmpic(v[0], "copy")) - || (0 == strcmpic(v[0], "mkcol")) - || (0 == strcmpic(v[0], "lock")) - || (0 == strcmpic(v[0], "unlock")) - ) - { - /* Normal */ - is_connect = 0; + /* + * Fail in case of unknown methods + * which we might not handle correctly. + * + * XXX: There should be a config option + * to forward requests with unknown methods + * anyway. Most of them don't need special + * steps. + */ + if (unknown_method(v[0])) + { + log_error(LOG_LEVEL_ERROR, "Unknown HTTP method detected: %s", v[0]); + freez(buf); + return JB_ERR_PARSE; } - else + + if (strcmpic(v[2], "HTTP/1.1") && strcmpic(v[2], "HTTP/1.0")) { - /* Unknown HTTP method */ - free(buf); + log_error(LOG_LEVEL_ERROR, "The only supported HTTP " + "versions are 1.0 and 1.1. This rules out: %s", v[2]); + freez(buf); return JB_ERR_PARSE; } - err = parse_http_url(v[1], http, csp); + http->ssl = !strcmpic(v[0], "CONNECT"); + + err = parse_http_url(v[1], http, !http->ssl); if (err) { - free(buf); + freez(buf); return err; } /* * Copy the details into the structure */ - http->ssl = is_connect; http->cmd = strdup(req); http->gpc = strdup(v[0]); http->ver = strdup(v[2]); + freez(buf); + if ( (http->cmd == NULL) || (http->gpc == NULL) || (http->ver == NULL) ) { - free(buf); - free_http_request(http); return JB_ERR_MEMORY; } return JB_ERR_OK; + +} + + +/********************************************************************* + * + * Function : compile_pattern + * + * Description : Compiles a host, domain or TAG pattern. + * + * Parameters : + * 1 : pattern = The pattern to compile. + * 2 : anchoring = How the regex should be anchored. + * Can be either one of NO_ANCHORING, + * LEFT_ANCHORED or RIGHT_ANCHORED. + * 3 : url = In case of failures, the spec member is + * logged and the structure freed. + * 4 : regex = Where the compiled regex should be stored. + * + * Returns : JB_ERR_OK - Success + * JB_ERR_MEMORY - Out of memory + * JB_ERR_PARSE - Cannot parse regex + * + *********************************************************************/ +static jb_err compile_pattern(const char *pattern, enum regex_anchoring anchoring, + struct url_spec *url, regex_t **regex) +{ + int errcode; + char rebuf[BUFFER_SIZE]; + const char *fmt = NULL; + + assert(pattern); + assert(strlen(pattern) < sizeof(rebuf) - 2); + + if (pattern[0] == '\0') + { + *regex = NULL; + return JB_ERR_OK; + } + + switch (anchoring) + { + case NO_ANCHORING: + fmt = "%s"; + break; + case RIGHT_ANCHORED: + fmt = "%s$"; + break; + case LEFT_ANCHORED: + fmt = "^%s"; + break; + default: + log_error(LOG_LEVEL_FATAL, + "Invalid anchoring in compile_pattern %d", anchoring); + } + + *regex = zalloc(sizeof(**regex)); + if (NULL == *regex) + { + free_url_spec(url); + return JB_ERR_MEMORY; + } + + snprintf(rebuf, sizeof(rebuf), fmt, pattern); + + errcode = regcomp(*regex, rebuf, (REG_EXTENDED|REG_NOSUB|REG_ICASE)); + + if (errcode) + { + size_t errlen = regerror(errcode, *regex, rebuf, sizeof(rebuf)); + if (errlen > (sizeof(rebuf) - (size_t)1)) + { + errlen = sizeof(rebuf) - (size_t)1; + } + rebuf[errlen] = '\0'; + log_error(LOG_LEVEL_ERROR, "error compiling %s from %s: %s", + pattern, url->spec, rebuf); + free_url_spec(url); + + return JB_ERR_PARSE; + } + + return JB_ERR_OK; + +} + + +/********************************************************************* + * + * Function : compile_url_pattern + * + * Description : Compiles the three parts of an URL pattern. + * + * Parameters : + * 1 : url = Target url_spec to be filled in. + * 2 : buf = The url pattern to compile. Will be messed up. + * + * Returns : JB_ERR_OK - Success + * JB_ERR_MEMORY - Out of memory + * JB_ERR_PARSE - Cannot parse regex + * + *********************************************************************/ +static jb_err compile_url_pattern(struct url_spec *url, char *buf) +{ + char *p; + + p = strchr(buf, '/'); + if (NULL != p) + { + /* + * Only compile the regex if it consists of more than + * a single slash, otherwise it wouldn't affect the result. + */ + if (p[1] != '\0') + { + /* + * XXX: does it make sense to compile the slash at the beginning? + */ + jb_err err = compile_pattern(p, LEFT_ANCHORED, url, &url->preg); + + if (JB_ERR_OK != err) + { + return err; + } + } + *p = '\0'; + } + + p = strchr(buf, ':'); + if (NULL != p) + { + *p++ = '\0'; + url->port_list = strdup(p); + if (NULL == url->port_list) + { + return JB_ERR_MEMORY; + } + } + else + { + url->port_list = NULL; + } + + if (buf[0] != '\0') + { + return compile_host_pattern(url, buf); + } + + return JB_ERR_OK; + +} + + +#ifdef FEATURE_EXTENDED_HOST_PATTERNS +/********************************************************************* + * + * Function : compile_host_pattern + * + * Description : Parses and compiles a host pattern.. + * + * Parameters : + * 1 : url = Target url_spec to be filled in. + * 2 : host_pattern = Host pattern to compile. + * + * Returns : JB_ERR_OK - Success + * JB_ERR_MEMORY - Out of memory + * JB_ERR_PARSE - Cannot parse regex + * + *********************************************************************/ +static jb_err compile_host_pattern(struct url_spec *url, const char *host_pattern) +{ + return compile_pattern(host_pattern, RIGHT_ANCHORED, url, &url->host_regex); +} + +#else + +/********************************************************************* + * + * Function : compile_host_pattern + * + * Description : Parses and "compiles" an old-school host pattern. + * + * Parameters : + * 1 : url = Target url_spec to be filled in. + * 2 : host_pattern = Host pattern to parse. + * + * Returns : JB_ERR_OK - Success + * JB_ERR_MEMORY - Out of memory + * JB_ERR_PARSE - Cannot parse regex + * + *********************************************************************/ +static jb_err compile_host_pattern(struct url_spec *url, const char *host_pattern) +{ + char *v[150]; + size_t size; + char *p; + + /* + * Parse domain part + */ + if (host_pattern[strlen(host_pattern) - 1] == '.') + { + url->unanchored |= ANCHOR_RIGHT; + } + if (host_pattern[0] == '.') + { + url->unanchored |= ANCHOR_LEFT; + } + + /* + * Split domain into components + */ + url->dbuffer = strdup(host_pattern); + if (NULL == url->dbuffer) + { + free_url_spec(url); + return JB_ERR_MEMORY; + } + + /* + * Map to lower case + */ + for (p = url->dbuffer; *p ; p++) + { + *p = (char)tolower((int)(unsigned char)*p); + } + + /* + * Split the domain name into components + */ + url->dcount = ssplit(url->dbuffer, ".", v, SZ(v), 1, 1); + + if (url->dcount < 0) + { + free_url_spec(url); + return JB_ERR_MEMORY; + } + else if (url->dcount != 0) + { + /* + * Save a copy of the pointers in dvec + */ + size = (size_t)url->dcount * sizeof(*url->dvec); + + url->dvec = (char **)malloc(size); + if (NULL == url->dvec) + { + free_url_spec(url); + return JB_ERR_MEMORY; + } + + memcpy(url->dvec, v, size); + } + /* + * else dcount == 0 in which case we needn't do anything, + * since dvec will never be accessed and the pattern will + * match all domains. + */ + return JB_ERR_OK; +} + + +/********************************************************************* + * + * Function : simplematch + * + * Description : String matching, with a (greedy) '*' wildcard that + * stands for zero or more arbitrary characters and + * character classes in [], which take both enumerations + * and ranges. + * + * Parameters : + * 1 : pattern = pattern for matching + * 2 : text = text to be matched + * + * Returns : 0 if match, else nonzero + * + *********************************************************************/ +static int simplematch(const char *pattern, const char *text) +{ + const unsigned char *pat = (const unsigned char *)pattern; + const unsigned char *txt = (const unsigned char *)text; + const unsigned char *fallback = pat; + int wildcard = 0; + + unsigned char lastchar = 'a'; + unsigned i; + unsigned char charmap[32]; + + while (*txt) + { + + /* EOF pattern but !EOF text? */ + if (*pat == '\0') + { + if (wildcard) + { + pat = fallback; + } + else + { + return 1; + } + } + + /* '*' in the pattern? */ + if (*pat == '*') + { + + /* The pattern ends afterwards? Speed up the return. */ + if (*++pat == '\0') + { + return 0; + } + + /* Else, set wildcard mode and remember position after '*' */ + wildcard = 1; + fallback = pat; + } + + /* Character range specification? */ + if (*pat == '[') + { + memset(charmap, '\0', sizeof(charmap)); + + while (*++pat != ']') + { + if (!*pat) + { + return 1; + } + else if (*pat == '-') + { + if ((*++pat == ']') || *pat == '\0') + { + return(1); + } + for (i = lastchar; i <= *pat; i++) + { + charmap[i / 8] |= (unsigned char)(1 << (i % 8)); + } + } + else + { + charmap[*pat / 8] |= (unsigned char)(1 << (*pat % 8)); + lastchar = *pat; + } + } + } /* -END- if Character range specification */ + + + /* + * Char match, or char range match? + */ + if ( (*pat == *txt) + || (*pat == '?') + || ((*pat == ']') && (charmap[*txt / 8] & (1 << (*txt % 8)))) ) + { + /* + * Sucess: Go ahead + */ + pat++; + } + else if (!wildcard) + { + /* + * No match && no wildcard: No luck + */ + return 1; + } + else if (pat != fallback) + { + /* + * Increment text pointer if in char range matching + */ + if (*pat == ']') + { + txt++; + } + /* + * Wildcard mode && nonmatch beyond fallback: Rewind pattern + */ + pat = fallback; + /* + * Restart matching from current text pointer + */ + continue; + } + txt++; + } + + /* Cut off extra '*'s */ + if(*pat == '*') pat++; + + /* If this is the pattern's end, fine! */ + return(*pat); + } @@ -572,6 +1229,7 @@ static int domain_match(const struct url_spec *pattern, const struct http_reques } } +#endif /* def FEATURE_EXTENDED_HOST_PATTERNS */ /********************************************************************* @@ -579,7 +1237,7 @@ static int domain_match(const struct url_spec *pattern, const struct http_reques * Function : create_url_spec * * Description : Creates a "url_spec" structure from a string. - * When finished, free with unload_url(). + * When finished, free with free_url_spec(). * * Parameters : * 1 : url = Target url_spec to be filled in. Will be @@ -597,155 +1255,30 @@ static int domain_match(const struct url_spec *pattern, const struct http_reques * written to system log) * *********************************************************************/ -jb_err create_url_spec(struct url_spec * url, const char * buf) +jb_err create_url_spec(struct url_spec *url, char *buf) { - char *p; - assert(url); assert(buf); - /* Zero memory */ memset(url, '\0', sizeof(*url)); - /* save a copy of the orignal specification */ - if ((url->spec = strdup(buf)) == NULL) + /* Remember the original specification for the CGI pages. */ + url->spec = strdup(buf); + if (NULL == url->spec) { return JB_ERR_MEMORY; } - if ((p = strchr(buf, '/')) != NULL) - { - if (NULL == (url->path = strdup(p))) - { - freez(url->spec); - return JB_ERR_MEMORY; - } - url->pathlen = strlen(url->path); - *p = '\0'; - } - else - { - url->path = NULL; - url->pathlen = 0; - } - if (url->path) - { - int errcode; - char rebuf[BUFFER_SIZE]; - - if (NULL == (url->preg = zalloc(sizeof(*url->preg)))) - { - freez(url->spec); - freez(url->path); - return JB_ERR_MEMORY; - } - - sprintf(rebuf, "^(%s)", url->path); - - errcode = regcomp(url->preg, rebuf, - (REG_EXTENDED|REG_NOSUB|REG_ICASE)); - if (errcode) - { - size_t errlen = regerror(errcode, - url->preg, rebuf, sizeof(rebuf)); - - if (errlen > (sizeof(rebuf) - (size_t)1)) - { - errlen = sizeof(rebuf) - (size_t)1; - } - rebuf[errlen] = '\0'; - - log_error(LOG_LEVEL_ERROR, "error compiling %s: %s", - url->spec, rebuf); - - freez(url->spec); - freez(url->path); - regfree(url->preg); - freez(url->preg); - - return JB_ERR_PARSE; - } - } - if ((p = strchr(buf, ':')) == NULL) - { - url->port = 0; - } - else - { - *p++ = '\0'; - url->port = atoi(p); - } - - if (buf[0] != '\0') + /* Is it tag pattern? */ + if (0 == strncmpic("TAG:", url->spec, 4)) { - char *v[150]; - size_t size; - - /* Parse domain part */ - if (buf[strlen(buf) - 1] == '.') - { - url->unanchored |= ANCHOR_RIGHT; - } - if (buf[0] == '.') - { - url->unanchored |= ANCHOR_LEFT; - } - - /* split domain into components */ - - url->dbuffer = strdup(buf); - if (NULL == url->dbuffer) - { - freez(url->spec); - freez(url->path); - regfree(url->preg); - freez(url->preg); - return JB_ERR_MEMORY; - } - - /* map to lower case */ - for (p = url->dbuffer; *p ; p++) - { - *p = tolower((int)(unsigned char)*p); - } - - /* split the domain name into components */ - url->dcount = ssplit(url->dbuffer, ".", v, SZ(v), 1, 1); - - if (url->dcount < 0) - { - freez(url->spec); - freez(url->path); - regfree(url->preg); - freez(url->preg); - freez(url->dbuffer); - url->dcount = 0; - return JB_ERR_MEMORY; - } - else if (url->dcount != 0) - { - - /* save a copy of the pointers in dvec */ - size = url->dcount * sizeof(*url->dvec); - - url->dvec = (char **)malloc(size); - if (NULL == url->dvec) - { - freez(url->spec); - freez(url->path); - regfree(url->preg); - freez(url->preg); - freez(url->dbuffer); - url->dcount = 0; - return JB_ERR_MEMORY; - } - - memcpy(url->dvec, v, size); - } + /* The pattern starts with the first character after "TAG:" */ + const char *tag_pattern = buf + 4; + return compile_pattern(tag_pattern, NO_ANCHORING, url, &url->tag_regex); } - return JB_ERR_OK; - + /* If it isn't a tag pattern it must be a URL pattern. */ + return compile_url_pattern(url, buf); } @@ -767,14 +1300,28 @@ void free_url_spec(struct url_spec *url) if (url == NULL) return; freez(url->spec); +#ifdef FEATURE_EXTENDED_HOST_PATTERNS + if (url->host_regex) + { + regfree(url->host_regex); + freez(url->host_regex); + } +#else freez(url->dbuffer); freez(url->dvec); - freez(url->path); + url->dcount = 0; +#endif /* ndef FEATURE_EXTENDED_HOST_PATTERNS */ + freez(url->port_list); if (url->preg) { regfree(url->preg); freez(url->preg); } + if (url->tag_regex) + { + regfree(url->tag_regex); + freez(url->tag_regex); + } } @@ -788,17 +1335,109 @@ void free_url_spec(struct url_spec *url) * 1 : pattern = a URL pattern * 2 : url = URL to match * - * Returns : 0 iff the URL matches the pattern, else nonzero. + * Returns : Nonzero if the URL matches the pattern, else 0. * *********************************************************************/ int url_match(const struct url_spec *pattern, - const struct http_request *url) + const struct http_request *http) { - return ((pattern->port == 0) || (pattern->port == url->port)) - && ((pattern->dbuffer == NULL) || (domain_match(pattern, url) == 0)) - && ((pattern->path == NULL) || - (regexec(pattern->preg, url->path, 0, NULL, 0) == 0) - ); + /* XXX: these should probably be functions. */ +#define PORT_MATCHES ((NULL == pattern->port_list) || match_portlist(pattern->port_list, http->port)) +#ifdef FEATURE_EXTENDED_HOST_PATTERNS +#define DOMAIN_MATCHES ((NULL == pattern->host_regex) || (0 == regexec(pattern->host_regex, http->host, 0, NULL, 0))) +#else +#define DOMAIN_MATCHES ((NULL == pattern->dbuffer) || (0 == domain_match(pattern, http))) +#endif +#define PATH_MATCHES ((NULL == pattern->preg) || (0 == regexec(pattern->preg, http->path, 0, NULL, 0))) + + if (pattern->tag_regex != NULL) + { + /* It's a tag pattern and shouldn't be matched against URLs */ + return 0; + } + + return (PORT_MATCHES && DOMAIN_MATCHES && PATH_MATCHES); + +} + + +/********************************************************************* + * + * Function : match_portlist + * + * Description : Check if a given number is covered by a comma + * separated list of numbers and ranges (a,b-c,d,..) + * + * Parameters : + * 1 : portlist = String with list + * 2 : port = port to check + * + * Returns : 0 => no match + * 1 => match + * + *********************************************************************/ +int match_portlist(const char *portlist, int port) +{ + char *min, *max, *next, *portlist_copy; + + min = next = portlist_copy = strdup(portlist); + + /* + * Zero-terminate first item and remember offset for next + */ + if (NULL != (next = strchr(portlist_copy, (int) ','))) + { + *next++ = '\0'; + } + + /* + * Loop through all items, checking for match + */ + while(min) + { + if (NULL == (max = strchr(min, (int) '-'))) + { + /* + * No dash, check for equality + */ + if (port == atoi(min)) + { + freez(portlist_copy); + return(1); + } + } + else + { + /* + * This is a range, so check if between min and max, + * or, if max was omitted, between min and 65K + */ + *max++ = '\0'; + if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535)) + { + freez(portlist_copy); + return(1); + } + + } + + /* + * Jump to next item + */ + min = next; + + /* + * Zero-terminate next item and remember offset for n+1 + */ + if ((NULL != next) && (NULL != (next = strchr(next, (int) ',')))) + { + *next++ = '\0'; + } + } + + freez(portlist_copy); + return 0; + }