const char filters_rcs[] = "$Id: filters.c,v 1.40 2001/10/26 17:37:55 oes Exp $"; /********************************************************************* * * File : $Source: /cvsroot/ijbswa/current/filters.c,v $ * * Purpose : Declares functions to parse/crunch headers and pages. * Functions declared include: * `acl_addr', `add_stats', `block_acl', `block_imageurl', * `block_url', `url_actions', `domaincmp', `dsplit', * `filter_popups', `forward_url', 'redirect_url', * `ij_untrusted_url', `intercept_url', `pcrs_filter_respose', * `show_proxy_args', 'ijb_send_banner', and `trust_url' * * Copyright : Written by and Copyright (C) 2001 the SourceForge * IJBSWA team. http://ijbswa.sourceforge.net * * Based on the Internet Junkbuster originally written * by and Copyright (C) 1997 Anonymous Coders and * Junkbusters Corporation. http://www.junkbusters.com * * This program is free software; you can redistribute it * and/or modify it under the terms of the GNU General * Public License as published by the Free Software * Foundation; either version 2 of the License, or (at * your option) any later version. * * This program is distributed in the hope that it will * be useful, but WITHOUT ANY WARRANTY; without even the * implied warranty of MERCHANTABILITY or FITNESS FOR A * PARTICULAR PURPOSE. See the GNU General Public * License for more details. * * The GNU General Public License should be included with * this file. If not, you can view it at * http://www.gnu.org/copyleft/gpl.html * or write to the Free Software Foundation, Inc., 59 * Temple Place - Suite 330, Boston, MA 02111-1307, USA. * * Revisions : * $Log: filters.c,v $ * Revision 1.40 2001/10/26 17:37:55 oes * - Re-enabled Netscape 200/404 bug workaround in block_url(): * - Removed OS/2 special case * - Made block_url() independant from sed() having been run * - Made trust_url independant from sed() having been run * - Made is_imageurl independant from sed() having been run. * It now checks User-Agent: and Accept: by itself. * * * Revision 1.39 2001/10/25 03:40:48 david__schmidt * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple * threads to call select() simultaneously. So, it's time to do a real, live, * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__ * (native). Both versions will work, but using __OS2__ offers multi-threading. * * Revision 1.38 2001/10/23 21:32:33 jongfoster * Adding error-checking to selected functions * * Revision 1.37 2001/10/22 15:33:56 david__schmidt * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in * filters.c. Added a FIXME in front of the offending code. I'll gladly * put in a better/more robust fix for all parties if one is presented... * It seems that just returning 200 instead of 404 would pretty much fix * it for everyone, but I don't know all the history of the problem. * * Revision 1.36 2001/10/10 16:44:16 oes * Added match_portlist function * * Revision 1.35 2001/10/07 15:41:23 oes * Replaced 6 boolean members of csp with one bitmap (csp->flags) * * New function remove_chunked_transfer_coding that strips chunked * transfer coding to plain and is called by pcrs_filter_response * and gif_deanimate_response if neccessary * * Improved handling of zero-change re_filter runs * * pcrs_filter_response and gif_deanimate_response now remove * chunked transfer codeing before processing the body. * * Revision 1.34 2001/09/20 15:49:36 steudten * * Fix BUG: Change int size to size_t size in pcrs_filter_response(). * See cgi.c fill_template(). * * Revision 1.33 2001/09/16 17:05:14 jongfoster * Removing unused #include showarg.h * * Revision 1.32 2001/09/16 13:21:27 jongfoster * Changes to use new list functions. * * Revision 1.31 2001/09/16 11:38:02 jongfoster * Splitting fill_template() into 2 functions: * template_load() loads the file * template_fill() performs the PCRS regexps. * This is because the CGI edit interface has a "table row" * template which is used many times in the page - this * change means it's only loaded from disk once. * * Revision 1.30 2001/09/16 11:00:10 jongfoster * New function alloc_http_response, for symmetry with free_http_response * * Revision 1.29 2001/09/13 23:32:40 jongfoster * Moving image data to cgi.c rather than cgi.h * Fixing a GPF under Win32 (and any other OS that protects global * constants from being written to). * * Revision 1.28 2001/09/10 10:18:51 oes * Silenced compiler warnings * * Revision 1.27 2001/08/05 16:06:20 jongfoster * Modifiying "struct map" so that there are now separate header and * "map_entry" structures. This means that functions which modify a * map no longer need to return a pointer to the modified map. * Also, it no longer reverses the order of the entries (which may be * important with some advanced template substitutions). * * Revision 1.26 2001/07/30 22:08:36 jongfoster * Tidying up #defines: * - All feature #defines are now of the form FEATURE_xxx * - Permanently turned off WIN_GUI_EDIT * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS * * Revision 1.25 2001/07/26 10:09:46 oes * Made browser detection a little less naive * * Revision 1.24 2001/07/25 17:22:51 oes * Added workaround for Netscape bug that prevents display of page when loading a component fails. * * Revision 1.23 2001/07/23 13:40:12 oes * Fixed bug that caused document body to be dropped when pcrs joblist was empty. * * Revision 1.22 2001/07/18 12:29:34 oes * - Made gif_deanimate_response respect * csp->action->string[ACTION_STRING_DEANIMATE] * - Logging cosmetics * * Revision 1.21 2001/07/13 13:59:53 oes * - Introduced gif_deanimate_response which shares the * generic content modification interface of pcrs_filter_response * and acts as a wrapper to deanimate.c:gif_deanimate() * - Renamed re_process_buffer to pcrs_filter_response * - pcrs_filter_response now returns NULL on failiure * - Removed all #ifdef PCRS * * Revision 1.20 2001/07/01 17:01:04 oes * Added comments and missing return statement in is_untrusted_url() * * Revision 1.19 2001/06/29 21:45:41 oes * Indentation, CRLF->LF, Tab-> Space * * Revision 1.18 2001/06/29 13:27:38 oes * - Cleaned up, renamed and reorderd functions * and improved comments * * - block_url: * - Ported to CGI platform. Now delivers * http_response or NULL * - Unified HTML and GIF generation (moved image detection * and GIF generation here from jcc.c:chat()) * - Fixed HTTP status to: * - 403 (Forbidden) for the "blocked" HTML message * - 200 (OK) for GIF answers * - 302 (Redirect) for redirect to GIF * * - trust_url: * - Ported to CGI platform. Now delivers * http_response or NULL * - Separated detection of untrusted URL into * (bool)is_untrusted_url * - Added enforcement of untrusted requests * * - Moved redirect_url() from cgi.c to here * and ported it to the CGI platform * * - Removed logentry from cancelled commit * * Revision 1.17 2001/06/09 10:55:28 jongfoster * Changing BUFSIZ ==> BUFFER_SIZE * * Revision 1.16 2001/06/07 23:10:26 jongfoster * Allowing unanchored domain patterns to back off and retry * if they partially match. Optimized right-anchored patterns. * Moving ACL and forward files into config file. * Replacing struct gateway with struct forward_spec * * Revision 1.15 2001/06/03 19:12:00 oes * extracted-CGI relevant stuff * * Revision 1.14 2001/06/01 10:30:55 oes * Added optional left-anchoring to domaincmp * * Revision 1.13 2001/05/31 21:21:30 jongfoster * Permissionsfile / actions file changes: * - Changed "permission" to "action" throughout * - changes to file format to allow string parameters * - Moved helper functions to actions.c * * Revision 1.12 2001/05/31 17:35:20 oes * * - Enhanced domain part globbing with infix and prefix asterisk * matching and optional unanchored operation * * Revision 1.11 2001/05/29 11:53:23 oes * "See why" link added to "blocked" page * * Revision 1.10 2001/05/29 09:50:24 jongfoster * Unified blocklist/imagelist/permissionslist. * File format is still under discussion, but the internal changes * are (mostly) done. * * Also modified interceptor behaviour: * - We now intercept all URLs beginning with one of the following * prefixes (and *only* these prefixes): * * http://i.j.b/ * * http://ijbswa.sf.net/config/ * * http://ijbswa.sourceforge.net/config/ * - New interceptors "home page" - go to http://i.j.b/ to see it. * - Internal changes so that intercepted and fast redirect pages * are not replaced with an image. * - Interceptors now have the option to send a binary page direct * to the client. (i.e. ijb-send-banner uses this) * - Implemented show-url-info interceptor. (Which is why I needed * the above interceptors changes - a typical URL is * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif". * The previous mechanism would not have intercepted that, and * if it had been intercepted then it then it would have replaced * it with an image.) * * Revision 1.9 2001/05/27 22:17:04 oes * * - re_process_buffer no longer writes the modified buffer * to the client, which was very ugly. It now returns the * buffer, which it is then written by chat. * * - content_length now adjusts the Content-Length: header * for modified documents rather than crunch()ing it. * (Length info in csp->content_length, which is 0 for * unmodified documents) * * - For this to work, sed() is called twice when filtering. * * Revision 1.8 2001/05/26 17:13:28 jongfoster * Filled in a function comment. * * Revision 1.7 2001/05/26 15:26:15 jongfoster * ACL feature now provides more security by immediately dropping * connections from untrusted hosts. * * Revision 1.6 2001/05/26 00:28:36 jongfoster * Automatic reloading of config file. * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32). * Most of the global variables have been moved to a new * struct configuration_spec, accessed through csp->config->globalname * Most of the globals remaining are used by the Win32 GUI. * * Revision 1.5 2001/05/25 22:34:30 jongfoster * Hard tabs->Spaces * * Revision 1.4 2001/05/22 18:46:04 oes * * - Enabled filtering banners by size rather than URL * by adding patterns that replace all standard banner * sizes with the "Junkbuster" gif to the re_filterfile * * - Enabled filtering WebBugs by providing a pattern * which kills all 1x1 images * * - Added support for PCRE_UNGREEDY behaviour to pcrs, * which is selected by the (nonstandard and therefore * capital) letter 'U' in the option string. * It causes the quantifiers to be ungreedy by default. * Appending a ? turns back to greedy (!). * * - Added a new interceptor ijb-send-banner, which * sends back the "Junkbuster" gif. Without imagelist or * MSIE detection support, or if tinygif = 1, or the * URL isn't recognized as an imageurl, a lame HTML * explanation is sent instead. * * - Added new feature, which permits blocking remote * script redirects and firing back a local redirect * to the browser. * The feature is conditionally compiled, i.e. it * can be disabled with --disable-fast-redirects, * plus it must be activated by a "fast-redirects" * line in the config file, has its own log level * and of course wants to be displayed by show-proxy-args * Note: Boy, all the #ifdefs in 1001 locations and * all the fumbling with configure.in and acconfig.h * were *way* more work than the feature itself :-( * * - Because a generic redirect template was needed for * this, tinygif = 3 now uses the same. * * - Moved GIFs, and other static HTTP response templates * to project.h * * - Some minor fixes * * - Removed some >400 CRs again (Jon, you really worked * a lot! ;-) * * Revision 1.3 2001/05/20 16:44:47 jongfoster * Removing last hardcoded JunkBusters.com URLs. * * Revision 1.2 2001/05/20 01:21:20 jongfoster * Version 2.9.4 checkin. * - Merged popupfile and cookiefile, and added control over PCRS * filtering, in new "permissionsfile". * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration * file error you now get a message box (in the Win32 GUI) rather * than the program exiting with no explanation. * - Made killpopup use the PCRS MIME-type checking and HTTP-header * skipping. * - Removed tabs from "config" * - Moved duplicated url parsing code in "loaders.c" to a new funcition. * - Bumped up version number. * * Revision 1.1.1.1 2001/05/15 13:58:52 oes * Initial import of version 2.9.3 source tree * * *********************************************************************/ #include "config.h" #include #include #include #include #include #include #ifndef _WIN32 #ifndef __OS2__ #include #endif /* ndef __OS2__ */ #include #else #include #endif /* ndef _WIN32 */ #ifdef __OS2__ #include #endif /* def __OS2__ */ #include "project.h" #include "filters.h" #include "encode.h" #include "parsers.h" #include "ssplit.h" #include "errlog.h" #include "jbsockets.h" #include "miscutil.h" #include "actions.h" #include "cgi.h" #include "list.h" #include "deanimate.h" #ifdef _WIN32 #include "win32.h" #endif const char filters_h_rcs[] = FILTERS_H_VERSION; /* Fix a problem with Solaris. There should be no effect on other * platforms. * Solaris's isspace() is a macro which uses it's argument directly * as an array index. Therefore we need to make sure that high-bit * characters generate +ve values, and ideally we also want to make * the argument match the declared parameter type of "int". */ #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X)) #ifdef FEATURE_ACL /********************************************************************* * * Function : block_acl * * Description : Block this request? * Decide yes or no based on ACL file. * * Parameters : * 1 : dst = The proxy or gateway address this is going to. * Or NULL to check all possible targets. * 2 : csp = Current client state (buffers, headers, etc...) * Also includes the client IP address. * * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block) * *********************************************************************/ int block_acl(struct access_control_addr *dst, struct client_state *csp) { struct access_control_list *acl = csp->config->acl; /* if not using an access control list, then permit the connection */ if (acl == NULL) { return(0); } /* search the list */ while (acl != NULL) { if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr) { if (dst == NULL) { /* Just want to check if they have any access */ if (acl->action == ACL_PERMIT) { return(0); } } else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr) && ((dst->port == acl->dst->port) || (acl->dst->port == 0))) { if (acl->action == ACL_PERMIT) { return(0); } else { return(1); } } } acl = acl->next; } return(1); } /********************************************************************* * * Function : acl_addr * * Description : Called from `load_aclfile' to parse an ACL address. * * Parameters : * 1 : aspec = String specifying ACL address. * 2 : aca = struct access_control_addr to fill in. * * Returns : 0 => Ok, everything else is an error. * *********************************************************************/ int acl_addr(char *aspec, struct access_control_addr *aca) { int i, masklength, port; char *p; masklength = 32; port = 0; if ((p = strchr(aspec, '/'))) { *p++ = '\0'; if (ijb_isdigit(*p) == 0) { return(-1); } masklength = atoi(p); } if ((masklength < 0) || (masklength > 32)) { return(-1); } if ((p = strchr(aspec, ':'))) { *p++ = '\0'; if (ijb_isdigit(*p) == 0) { return(-1); } port = atoi(p); } aca->port = port; aca->addr = ntohl(resolve_hostname_to_ip(aspec)); if (aca->addr == -1) { log_error(LOG_LEVEL_ERROR, "can't resolve address for %s", aspec); return(-1); } /* build the netmask */ aca->mask = 0; for (i=1; i <= masklength ; i++) { aca->mask |= (1 << (32 - i)); } /* now mask off the host portion of the ip address * (i.e. save on the network portion of the address). */ aca->addr = aca->addr & aca->mask; return(0); } #endif /* def FEATURE_ACL */ /********************************************************************* * * 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)) { free(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)) { free(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'; } } free(portlist_copy); return 0; } /********************************************************************* * * Function : block_url * * Description : Called from `chat'. Check to see if we need to block this. * * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) * * Returns : NULL => unblocked, else HTTP block response * *********************************************************************/ struct http_response *block_url(struct client_state *csp) { #ifdef FEATURE_IMAGE_BLOCKING char *p; #endif /* def FEATURE_IMAGE_BLOCKING */ struct http_response *rsp; /* * If it's not blocked, don't block it ;-) */ if ((csp->action->flags & ACTION_BLOCK) == 0) { return NULL; } /* * Else, prepare a response */ if (NULL == (rsp = alloc_http_response())) { return cgi_error_memory(); } /* * If it's an image-url, send back an image or redirect * as specified by the relevant +image action */ #ifdef FEATURE_IMAGE_BLOCKING if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0) && is_imageurl(csp)) { /* determine HOW images should be blocked */ p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER]; /* and handle accordingly: */ if ((p == NULL) || (0 == strcmpic(p, "logo"))) { rsp->body = bindup(image_junkbuster_gif_data, image_junkbuster_gif_length); if (rsp->body == NULL) { free_http_response(rsp); return cgi_error_memory(); } rsp->content_length = image_junkbuster_gif_length; if (enlist_unique_header(rsp->headers, "Content-Type", "image/gif")) { free_http_response(rsp); return cgi_error_memory(); } } else if (0 == strcmpic(p, "blank")) { rsp->body = bindup(image_blank_gif_data, image_blank_gif_length); if (rsp->body == NULL) { free_http_response(rsp); return cgi_error_memory(); } rsp->content_length = image_blank_gif_length; if (enlist_unique_header(rsp->headers, "Content-Type", "image/gif")) { free_http_response(rsp); return cgi_error_memory(); } } else { rsp->status = strdup("302 Local Redirect from Junkbuster"); if (rsp->status == NULL) { free_http_response(rsp); return cgi_error_memory(); } if (enlist_unique_header(rsp->headers, "Location", p)) { free_http_response(rsp); return cgi_error_memory(); } } } else #endif /* def FEATURE_IMAGE_BLOCKING */ /* * Else, generate an HTML "blocked" message: */ { jb_err err; struct map * exports; /* * Workaround for stupid Netscape bug which prevents * pages from being displayed if loading a referenced * JavaScript or style sheet fails. So make it appear * as if it succeeded. */ if ( NULL != (p = get_header_value(csp->headers, "User-Agent:")) && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */ && !strstr(p, "Gecko") /* save Mozilla, */ && !strstr(p, "compatible") /* MSIE */ && !strstr(p, "Opera")) /* and Opera. */ { rsp->status = strdup("200 Request for blocked URL"); } else { rsp->status = strdup("404 Request for blocked URL"); } if (rsp->status == NULL) { free_http_response(rsp); return cgi_error_memory(); } exports = default_exports(csp, NULL); if (exports == NULL) { free_http_response(rsp); return cgi_error_memory(); } #ifdef FEATURE_FORCE_LOAD err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1); #else /* ifndef FEATURE_FORCE_LOAD */ err = map_block_killer(exports, "force-support"); #endif /* ndef FEATURE_FORCE_LOAD */ if (!err) err = map(exports, "hostport", 1, csp->http->hostport, 1); if (!err) err = map(exports, "hostport-html", 1, html_encode(csp->http->hostport), 0); if (!err) err = map(exports, "path", 1, csp->http->path, 1); if (!err) err = map(exports, "path-html", 1, html_encode(csp->http->path), 0); if (err) { free_map(exports); free_http_response(rsp); return cgi_error_memory(); } err = template_fill_for_cgi(csp, "blocked", exports, rsp); if (err) { free_http_response(rsp); return cgi_error_memory(); } } return finish_http_response(rsp); } #ifdef FEATURE_TRUST /********************************************************************* * * Function : trust_url FIXME: I should be called distrust_url * * Description : Calls is_untrusted_url to determine if the URL is trusted * and if not, returns a HTTP 304 response with a reject message. * * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) * * Returns : NULL => trusted, else http_response. * *********************************************************************/ struct http_response *trust_url(struct client_state *csp) { struct http_response *rsp; struct map * exports; char buf[BUFFER_SIZE]; char *p; struct url_spec **tl; struct url_spec *t; jb_err err; /* * Don't bother to work on trusted URLs */ if (!is_untrusted_url(csp)) { return NULL; } /* * Else, prepare a response: */ if (NULL == (rsp = alloc_http_response())) { return cgi_error_memory(); } exports = default_exports(csp, NULL); if (exports == NULL) { free_http_response(rsp); return cgi_error_memory(); } /* * Export the host, port, and referrer information */ err = map(exports, "hostport", 1, csp->http->hostport, 1); if (!err) err = map(exports, "path", 1, csp->http->path, 1); if (!err) err = map(exports, "hostport-html", 1, html_encode(csp->http->hostport), 0); if (!err) err = map(exports, "path-html", 1, html_encode(csp->http->path), 0); if (NULL != (p = get_header_value(csp->headers, "Referer:"))) { if (!err) err = map(exports, "referrer", 1, p, 1); if (!err) err = map(exports, "referrer-html", 1, html_encode(p), 0); } else { if (!err) err = map(exports, "referrer", 1, "unknown", 1); if (!err) err = map(exports, "referrer-html", 1, "unknown", 1); } if (err) { free_map(exports); free_http_response(rsp); return cgi_error_memory(); } /* * Export the trust list */ p = strdup(""); for (tl = csp->config->trust_list; (t = *tl) ; tl++) { sprintf(buf, "
  • %s
  • \n", t->spec); string_append(&p, buf); } err = map(exports, "trusted-referrers", 1, p, 0); if (err) { free_map(exports); free_http_response(rsp); return cgi_error_memory(); } /* * Export the trust info, if available */ if (csp->config->trust_info->first) { struct list_entry *l; p = strdup(""); for (l = csp->config->trust_info->first; l ; l = l->next) { sprintf(buf, "
  • %s
    \n",l->str, l->str); string_append(&p, buf); } err = map(exports, "trust-info", 1, p, 0); } else { err = map_block_killer(exports, "have-trust-info"); } if (err) { free_map(exports); free_http_response(rsp); return cgi_error_memory(); } /* * Export the force prefix or the force conditional block killer */ #ifdef FEATURE_FORCE_LOAD err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1); #else /* ifndef FEATURE_FORCE_LOAD */ err = map_block_killer(exports, "force-support"); #endif /* ndef FEATURE_FORCE_LOAD */ if (err) { free_map(exports); free_http_response(rsp); return cgi_error_memory(); } /* * Build the response */ err = template_fill_for_cgi(csp, "untrusted", exports, rsp); if (err) { free_http_response(rsp); return cgi_error_memory(); } return finish_http_response(rsp); } #endif /* def FEATURE_TRUST */ #ifdef FEATURE_FAST_REDIRECTS /********************************************************************* * * Function : redirect_url * * Description : Checks for redirection URLs and returns a HTTP redirect * to the destination URL, if necessary * * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) * * Returns : NULL if URL was clean, HTTP redirect otherwise. * *********************************************************************/ struct http_response *redirect_url(struct client_state *csp) { char *p, *q; struct http_response *rsp; p = q = csp->http->path; log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p); /* * find the last URL encoded in the request */ while ((p = strstr(p, "http://"))) { q = p++; } /* * if there was any, generate and return a HTTP redirect */ if (q != csp->http->path) { log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q); if (NULL == (rsp = alloc_http_response())) { return cgi_error_memory(); } if ( enlist_unique_header(rsp->headers, "Location", q) || (NULL == (rsp->status = strdup("302 Local Redirect from Junkbuster"))) ) { free_http_response(rsp); return cgi_error_memory(); } return finish_http_response(rsp); } else { return NULL; } } #endif /* def FEATURE_FAST_REDIRECTS */ #ifdef FEATURE_IMAGE_BLOCKING /********************************************************************* * * Function : is_imageurl * * Description : Given a URL, decide whether it is an image or not, * using either the info from a previous +image action * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, the info from * the browser's accept header. * * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) * * Returns : True (nonzero) if URL is an image, false (0) * otherwise * *********************************************************************/ int is_imageurl(struct client_state *csp) { #ifdef FEATURE_IMAGE_DETECT_MSIE char *tmp; tmp = get_header_value(csp->headers, "User-Agent:"); if (tmp && strstr(tmp, "MSIE")) { tmp = get_header_value(csp->headers, "Accept:"); if (tmp && strstr(tmp, "image/gif")) { /* Client will accept HTML. If this seems counterintuitive, * blame Microsoft. */ return(0); } else { return(1); } } #endif /* def FEATURE_IMAGE_DETECT_MSIE */ return ((csp->action->flags & ACTION_IMAGE) != 0); } #endif /* def FEATURE_IMAGE_BLOCKING */ #ifdef FEATURE_COOKIE_JAR /********************************************************************* * * Function : is_untrusted_url * * Description : Should we "distrust" this URL (and block it)? * * Yes if it matches a line in the trustfile, or if the * referrer matches a line starting with "+" in the * trustfile. * No otherwise. * * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) * * Returns : 0 => trusted, 1 => untrusted * *********************************************************************/ int is_untrusted_url(struct client_state *csp) { struct file_list *fl; struct block_spec *b; struct url_spec url[1], **tl, *t; struct http_request rhttp[1]; char *p, *h; /* * If we don't have a trustlist, we trust everybody */ if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL)) { return(0); } /* * Do we trust the request URL itself? */ *url = dsplit(csp->http->host); /* if splitting the domain fails, punt */ if (url->dbuf == NULL) return(0); memset(rhttp, '\0', sizeof(*rhttp)); for (b = b->next; b ; b = b->next) { if ((b->url->port == 0) || (b->url->port == csp->http->port)) { if ((b->url->domain[0] == '\0') || (domaincmp(b->url, url) == 0)) { if ((b->url->path == NULL) || #ifdef REGEX (regexec(b->url->preg, csp->http->path, 0, NULL, 0) == 0) #else (strncmp(b->url->path, csp->http->path, b->url->pathlen) == 0) #endif ) { freez(url->dbuf); freez(url->dvec); if (b->reject == 0) return(0); return(1); } } } } freez(url->dbuf); freez(url->dvec); if (NULL == (h = get_header_value(csp->headers, "Referer:"))) { /* no referrer was supplied */ return(1); } /* forge a URL from the referrer so we can use * convert_url() to parse it into its components. */ p = NULL; p = strsav(p, "GET "); p = strsav(p, h); p = strsav(p, " HTTP/1.0"); parse_http_request(p, rhttp, csp); freez(p); if (rhttp->cmd == NULL) { return(1); } /* * If not, do we maybe trust its referrer? */ *url = dsplit(rhttp->host); /* if splitting the domain fails, punt */ if (url->dbuf == NULL) return(1); for (tl = csp->config->trust_list; (t = *tl) ; tl++) { if ((t->port == 0) || (t->port == rhttp->port)) { if ((t->domain[0] == '\0') || domaincmp(t, url) == 0) { if ((t->path == NULL) || #ifdef REGEX (regexec(t->preg, rhttp->path, 0, NULL, 0) == 0) #else (strncmp(t->path, rhttp->path, t->pathlen) == 0) #endif ) { /* if the URL's referrer is from a trusted referrer, then * add the target spec to the trustfile as an unblocked * domain and return NULL (which means it's OK). */ FILE *fp; freez(url->dbuf); freez(url->dvec); if ((fp = fopen(csp->config->trustfile, "a"))) { h = NULL; h = strsav(h, "~"); h = strsav(h, csp->http->hostport); p = csp->http->path; if ((*p++ == '/') && (*p++ == '~')) { /* since this path points into a user's home space * be sure to include this spec in the trustfile. */ if ((p = strchr(p, '/'))) { *p = '\0'; h = strsav(h, csp->http->path); /* FIXME: p?! */ h = strsav(h, "/"); } } fprintf(fp, "%s\n", h); freez(h); fclose(fp); } return(0); } } } } return(1); } #endif /* def FEATURE_COOKIE_JAR */ /********************************************************************* * * Function : pcrs_filter_response * * Description : Apply all the pcrs jobs from the joblist (re_filterfile) * to the text buffer that's been accumulated in * csp->iob->buf and set csp->content_length to the modified * size and raise the CSP_FLAG_MODIFIED flag if appropriate. * * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) * * Returns : a pointer to the (newly allocated) modified buffer. * or NULL in case something went wrong * *********************************************************************/ char *pcrs_filter_response(struct client_state *csp) { int hits=0; size_t size; char *old = csp->iob->cur, *new = NULL; pcrs_job *job; struct file_list *fl; struct re_filterfile_spec *b; /* Sanity first */ if (csp->iob->cur >= csp->iob->eod) { return(NULL); } size = csp->iob->eod - csp->iob->cur; /* * If the body has a "chunked" transfer-encoding, * get rid of it first, adjusting size and iob->eod */ if (csp->flags & CSP_FLAG_CHUNKED) { log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first"); if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size))) { return(NULL); } csp->iob->eod = csp->iob->cur + size; csp->flags |= CSP_FLAG_MODIFIED; } if ( ( NULL == (fl = csp->rlist) ) || ( NULL == (b = fl->f) ) ) { log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering."); return(NULL); } if ( NULL == b->joblist ) { log_error(LOG_LEVEL_RE_FILTER, "Empty joblist. Nothing to do."); return(NULL); } log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) ...", csp->http->hostport, csp->http->path, size); /* Apply all jobs from the joblist */ for (job = b->joblist; NULL != job; job = job->next) { hits += pcrs_execute(job, old, size, &new, &size); if (old != csp->iob->cur) free(old); old=new; } log_error(LOG_LEVEL_RE_FILTER, " produced %d hits (new size %d).", hits, size); /* * If there were no hits, destroy our copy and let * chat() use the original in csp->iob */ if (!hits) { free(new); return(NULL); } csp->flags |= CSP_FLAG_MODIFIED; csp->content_length = size; IOB_RESET(csp); return(new); } /********************************************************************* * * Function : gif_deanimate_response * * Description : Deanimate the GIF image that has been accumulated in * csp->iob->buf, set csp->content_length to the modified * size and raise the CSP_FLAG_MODIFIED flag. * * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) * * Returns : a pointer to the (newly allocated) modified buffer. * or NULL in case something went wrong. * *********************************************************************/ char *gif_deanimate_response(struct client_state *csp) { struct binbuffer *in, *out; char *p; int size = csp->iob->eod - csp->iob->cur; /* * If the body has a "chunked" transfer-encoding, * get rid of it first, adjusting size and iob->eod */ if (csp->flags & CSP_FLAG_CHUNKED) { log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first"); if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size))) { return(NULL); } csp->iob->eod = csp->iob->cur + size; csp->flags |= CSP_FLAG_MODIFIED; } if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in ))) || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) ) { log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)"); return NULL; } in->buffer = csp->iob->cur; in->size = size; if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4))) { log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)"); free(in); buf_free(out); return(NULL); } else { log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset); csp->content_length = out->offset; csp->flags |= CSP_FLAG_MODIFIED; p = out->buffer; free(in); free(out); return(p); } } /********************************************************************* * * Function : remove_chunked_transfer_coding * * Description : In-situ remove the "chunked" transfer coding as defined * in rfc2616 from a buffer. * * Parameters : * 1 : buffer = Pointer to the text buffer * 2 : size = Number of bytes to be processed * * Returns : The new size, i.e. the number of bytes from buffer which * are occupied by the stripped body, or 0 in case something * went wrong * *********************************************************************/ int remove_chunked_transfer_coding(char *buffer, const size_t size) { size_t newsize = 0; unsigned int chunksize = 0; char *from_p, *to_p; assert(buffer); from_p = to_p = buffer; if (sscanf(buffer, "%x", &chunksize) != 1) { log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding"); return(0); } while (chunksize > 0) { if (NULL == (from_p = strstr(from_p, "\r\n"))) { log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding"); return(0); } newsize += chunksize; from_p += 2; memmove(to_p, from_p, (size_t) chunksize); to_p = buffer + newsize; from_p += chunksize + 2; if (sscanf(from_p, "%x", &chunksize) != 1) { log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding"); return(0); } } /* FIXME: Should this get its own loglevel? */ log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize); return(newsize); } /********************************************************************* * * Function : url_actions * * Description : Gets the actions for this URL. * * Parameters : * 1 : http = http_request request for blocked URLs * 2 : csp = Current client state (buffers, headers, etc...) * * Returns : N/A * *********************************************************************/ void url_actions(struct http_request *http, struct client_state *csp) { struct file_list *fl; struct url_actions *b; init_current_action(csp->action); if (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL)) { return; } apply_url_actions(csp->action, http, b); } /********************************************************************* * * Function : apply_url_actions * * Description : Applies a list of URL actions. * * Parameters : * 1 : action = Destination. * 2 : http = Current URL * 3 : b = list of URL actions to apply * * Returns : N/A * *********************************************************************/ void apply_url_actions(struct current_action_spec *action, struct http_request *http, struct url_actions *b) { struct url_spec url[1]; if (b == NULL) { /* Should never happen */ return; } *url = dsplit(http->host); /* if splitting the domain fails, punt */ if (url->dbuf == NULL) { return; } for (b = b->next; NULL != b; b = b->next) { if ((b->url->port == 0) || (b->url->port == http->port)) { if ((b->url->domain[0] == '\0') || (domaincmp(b->url, url) == 0)) { if ((b->url->path == NULL) || #ifdef REGEX (regexec(b->url->preg, http->path, 0, NULL, 0) == 0) #else (strncmp(b->url->path, http->path, b->url->pathlen) == 0) #endif ) { merge_current_action(action, b->action); } } } } freez(url->dbuf); freez(url->dvec); } /********************************************************************* * * Function : forward_url * * Description : Should we forward this to another proxy? * * Parameters : * 1 : http = http_request request for current URL * 2 : csp = Current client state (buffers, headers, etc...) * * Returns : Pointer to forwarding information. * *********************************************************************/ const struct forward_spec * forward_url(struct http_request *http, struct client_state *csp) { static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER }; struct forward_spec *fwd = csp->config->forward; struct url_spec url[1]; if (fwd == NULL) { return(fwd_default); } *url = dsplit(http->host); /* if splitting the domain fails, punt */ if (url->dbuf == NULL) { return(fwd_default); } while (fwd != NULL) { if ((fwd->url->port == 0) || (fwd->url->port == http->port)) { if ((fwd->url->domain[0] == '\0') || (domaincmp(fwd->url, url) == 0)) { if ((fwd->url->path == NULL) || #ifdef REGEX (regexec(fwd->url->preg, http->path, 0, NULL, 0) == 0) #else (strncmp(fwd->url->path, http->path, fwd->url->pathlen) == 0) #endif ) { freez(url->dbuf); freez(url->dvec); return(fwd); } } } fwd = fwd->next; } freez(url->dbuf); freez(url->dvec); return(fwd_default); } /********************************************************************* * * Function : dsplit * * Description : Takes a domain and returns a pointer to a url_spec * structure populated with dbuf, dcnt and dvec. The * other fields in the structure that is returned are zero. * * Parameters : * 1 : domain = a URL address * * Returns : url_spec structure populated with dbuf, dcnt and dvec. * On error, the dbuf field will be set to NULL. (As * will all the others, but you don't need to check * them). * * FIXME: Returning a structure is horribly inefficient, please can * this structure take a (struct url_spec * dest) * pointer instead? * *********************************************************************/ struct url_spec dsplit(char *domain) { struct url_spec ret[1]; char *v[BUFFER_SIZE]; int size; char *p; memset(ret, '\0', sizeof(*ret)); if (domain[strlen(domain) - 1] == '.') { ret->unanchored |= ANCHOR_RIGHT; } if (domain[0] == '.') { ret->unanchored |= ANCHOR_LEFT; } ret->dbuf = strdup(domain); if (NULL == ret->dbuf) { return *ret; } /* map to lower case */ for (p = ret->dbuf; *p ; p++) { *p = tolower((int)(unsigned char)*p); } /* split the domain name into components */ ret->dcnt = ssplit(ret->dbuf, ".", v, SZ(v), 1, 1); if (ret->dcnt < 0) { free(ret->dbuf); memset(ret, '\0', sizeof(ret)); return *ret; } else if (ret->dcnt == 0) { return *ret; } /* save a copy of the pointers in dvec */ size = ret->dcnt * sizeof(*ret->dvec); ret->dvec = (char **)malloc(size); if (NULL == ret->dvec) { free(ret->dbuf); memset(ret, '\0', sizeof(ret)); return *ret; } memcpy(ret->dvec, v, size); return *ret; } /********************************************************************* * * Function : simple_domaincmp * * Description : Domain-wise Compare fqdn's. The comparison is * both left- and right-anchored. The individual * domain names are compared with simplematch(). * This is only used by domaincmp. * * Parameters : * 1 : pv = array of patterns to compare * 2 : fv = array of domain components to compare * 3 : len = length of the arrays (both arrays are the * same length - if they weren't, it couldn't * possibly be a match). * * Returns : 0 => domains are equivalent, else no match. * *********************************************************************/ static int simple_domaincmp(char **pv, char **fv, int len) { int n; for (n = 0; n < len; n++) { if (simplematch(pv[n], fv[n])) { return 1; } } return 0; } /********************************************************************* * * Function : domaincmp * * Description : Domain-wise Compare fqdn's. Governed by the bimap in * pattern->unachored, the comparison is un-, left-, * right-anchored, or both. * The individual domain names are compared with * simplematch(). * * Parameters : * 1 : pattern = a domain that may contain a '*' as a wildcard. * 2 : fqdn = domain name against which the patterns are compared. * * Returns : 0 => domains are equivalent, else no match. * *********************************************************************/ int domaincmp(struct url_spec *pattern, struct url_spec *fqdn) { char **pv, **fv; /* vectors */ int plen, flen; int unanchored = pattern->unanchored & (ANCHOR_RIGHT | ANCHOR_LEFT); plen = pattern->dcnt; flen = fqdn->dcnt; if (flen < plen) { /* fqdn is too short to match this pattern */ return 1; } pv = pattern->dvec; fv = fqdn->dvec; if (unanchored == ANCHOR_LEFT) { /* * Right anchored. * * Convert this into a fully anchored pattern with * the fqdn and pattern the same length */ fv += (flen - plen); /* flen - plen >= 0 due to check above */ return simple_domaincmp(pv, fv, plen); } else if (unanchored == 0) { /* Fully anchored, check length */ if (flen != plen) { return 1; } return simple_domaincmp(pv, fv, plen); } else if (unanchored == ANCHOR_RIGHT) { /* Left anchored, ignore all extra in fqdn */ return simple_domaincmp(pv, fv, plen); } else { /* Unanchored */ int n; int maxn = flen - plen; for (n = 0; n <= maxn; n++) { if (!simple_domaincmp(pv, fv, plen)) { return 0; } /* * Doesn't match from start of fqdn * Try skipping first part of fqdn */ fv++; } return 1; } } /* Local Variables: tab-width: 3 end: */