X-Git-Url: http://www.privoxy.org/gitweb/?p=privoxy.git;a=blobdiff_plain;f=filters.c;h=0831df75b5e8c0747bcfbe9e1856584cf4999494;hp=e43ad99afd8cf74af5d0bc679ec57ba12ad625db;hb=734a597393f1fe7367c1f78ab6ce07fb8db1e767;hpb=e30955dc472815457c4e10ed744fa7aa29ac4800 diff --git a/filters.c b/filters.c index e43ad99a..0831df75 100644 --- a/filters.c +++ b/filters.c @@ -1,4 +1,4 @@ -const char filters_rcs[] = "$Id: filters.c,v 1.15 2001/06/03 11:03:48 oes Exp $"; +const char filters_rcs[] = "$Id: filters.c,v 1.42 2002/01/17 21:00:32 jongfoster Exp $"; /********************************************************************* * * File : $Source: /cvsroot/ijbswa/current/filters.c,v $ @@ -6,19 +6,19 @@ const char filters_rcs[] = "$Id: filters.c,v 1.15 2001/06/03 11:03:48 oes Exp $" * Purpose : Declares functions to parse/crunch headers and pages. * Functions declared include: * `acl_addr', `add_stats', `block_acl', `block_imageurl', - * `block_url', `url_actions', `domaincmp', `dsplit', + * `block_url', `url_actions', `domain_split', * `filter_popups', `forward_url', 'redirect_url', - * `ij_untrusted_url', `intercept_url', `re_process_buffer', - * `show_proxy_args', 'ijb_send_banner', and `trust_url' + * `ij_untrusted_url', `intercept_url', `pcrs_filter_respose', + * 'ijb_send_banner', and `trust_url' * * Copyright : Written by and Copyright (C) 2001 the SourceForge * IJBSWA team. http://ijbswa.sourceforge.net * * Based on the Internet Junkbuster originally written - * by and Copyright (C) 1997 Anonymous Coders and + * by and Copyright (C) 1997 Anonymous Coders and * Junkbusters Corporation. http://www.junkbusters.com * - * This program is free software; you can redistribute it + * This program is free software; you can redistribute it * and/or modify it under the terms of the GNU General * Public License as published by the Free Software * Foundation; either version 2 of the License, or (at @@ -38,59 +38,169 @@ const char filters_rcs[] = "$Id: filters.c,v 1.15 2001/06/03 11:03:48 oes Exp $" * * Revisions : * $Log: filters.c,v $ - * Revision 1.15 2001/06/03 11:03:48 oes - * Makefile/in + * Revision 1.42 2002/01/17 21:00:32 jongfoster + * Moving all our URL and URL pattern parsing code to urlmatch.c. * - * introduced cgi.c + * Using a single, simple url_match(pattern,url) function - rather than + * the 3-line match routine which was repeated all over the place. * - * actions.c: + * Renaming free_url to free_url_spec, since it frees a struct url_spec. * - * adapted to new enlist_unique arg format + * Using parse_http_url() to parse URLs without faking a HTTP + * request line for parse_http_request(). * - * conf loadcfg.c + * Revision 1.41 2001/11/13 00:14:07 jongfoster + * Fixing stupid bug now I've figured out what || means. + * (It always returns 0 or 1, not one of it's paramaters.) * - * introduced confdir option + * 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. * - * filters.c filtrers.h * - * extracted-CGI relevant stuff + * 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. * - * jbsockets.c + * Revision 1.38 2001/10/23 21:32:33 jongfoster + * Adding error-checking to selected functions * - * filled comment + * 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. * - * jcc.c + * 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() * - * support for new cgi mechansim + * 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 * - * list.c list.h + * - 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 * - * functions for new list type: "map" - * extended enlist_unique + * - 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 * - * miscutil.c .h - * introduced bindup() - * - * parsers.c parsers.h - * - * deleted const struct interceptors - * - * pcrs.c - * added FIXME - * - * project.h - * - * added struct map - * added struct http_response - * changes struct interceptors to struct cgi_dispatcher - * moved HTML stuff to cgi.h - * - * re_filterfile: - * - * changed - * - * showargs.c - * NO TIME LEFT + * - 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 @@ -236,28 +346,34 @@ const char filters_rcs[] = "$Id: filters.c,v 1.15 2001/06/03 11:03:48 oes Exp $" #include #include #include +#include #ifndef _WIN32 +#ifndef __OS2__ #include +#endif /* ndef __OS2__ */ #include #else #include -#endif +#endif /* ndef _WIN32 */ + +#ifdef __OS2__ +#include +#endif /* def __OS2__ */ #include "project.h" #include "filters.h" #include "encode.h" -#include "jcc.h" -#include "showargs.h" #include "parsers.h" #include "ssplit.h" -#include "gateway.h" -#include "jbsockets.h" #include "errlog.h" #include "jbsockets.h" #include "miscutil.h" #include "actions.h" #include "cgi.h" +#include "list.h" +#include "deanimate.h" +#include "urlmatch.h" #ifdef _WIN32 #include "win32.h" @@ -275,7 +391,7 @@ const char filters_h_rcs[] = FILTERS_H_VERSION; #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X)) -#ifdef ACL_FILES +#ifdef FEATURE_ACL /********************************************************************* * * Function : block_acl @@ -292,36 +408,33 @@ const char filters_h_rcs[] = FILTERS_H_VERSION; * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block) * *********************************************************************/ -int block_acl(struct access_control_addr *dst, - struct client_state *csp) +int block_acl(struct access_control_addr *dst, struct client_state *csp) { - struct file_list *fl; - struct access_control_list *a, *acl; + struct access_control_list *acl = csp->config->acl; /* if not using an access control list, then permit the connection */ - if (((fl = csp->alist) == NULL) || - ((acl = (struct access_control_list *) fl->f) == NULL)) + if (acl == NULL) { return(0); } /* search the list */ - for (a = acl->next ; a ; a = a->next) + while (acl != NULL) { - if ((csp->ip_addr_long & a->src->mask) == a->src->addr) + if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr) { if (dst == NULL) { /* Just want to check if they have any access */ - if (a->action == ACL_PERMIT) + if (acl->action == ACL_PERMIT) { return(0); } } - else if ( ((dst->addr & a->dst->mask) == a->dst->addr) - && ((dst->port == a->dst->port) || (a->dst->port == 0))) + else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr) + && ((dst->port == acl->dst->port) || (acl->dst->port == 0))) { - if (a->action == ACL_PERMIT) + if (acl->action == ACL_PERMIT) { return(0); } @@ -331,6 +444,7 @@ int block_acl(struct access_control_addr *dst, } } } + acl = acl->next; } return(1); @@ -411,628 +525,956 @@ int acl_addr(char *aspec, struct access_control_addr *aca) return(0); } -#endif /* def ACL_FILES */ +#endif /* def FEATURE_ACL */ /********************************************************************* * - * Function : block_url + * Function : match_portlist * - * Description : Called from `chat'. Check to see if we need to block this. + * Description : Check if a given number is covered by a comma + * separated list of numbers and ranges (a,b-c,d,..) * * Parameters : - * 1 : http = http_request request to "check" for blocked - * 2 : csp = Current client state (buffers, headers, etc...) + * 1 : portlist = String with list + * 2 : port = port to check * - * Returns : NULL => unblocked, else string to HTML block description. + * Returns : 0 => no match + * 1 => match * *********************************************************************/ -char *block_url(struct http_request *http, struct client_state *csp) +int match_portlist(const char *portlist, int port) { - char *p; - int n; - int factor = 2; + char *min, *max, *next, *portlist_copy; - if ((csp->action->flags & ACTION_BLOCK) == 0) + min = next = portlist_copy = strdup(portlist); + + /* + * Zero-terminate first item and remember offset for next + */ + if (NULL != (next = strchr(portlist_copy, (int) ','))) { - return(NULL); + *next++ = '\0'; } - else - { -#ifdef FORCE_LOAD - factor++; -#endif /* def FORCE_LOAD */ - n = strlen(CBLOCK); - n += factor * strlen(http->hostport); - n += factor * strlen(http->path); + /* + * 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); + } - p = (char *)malloc(n); + } -#ifdef FORCE_LOAD - sprintf(p, CBLOCK, http->hostport, http->path, http->hostport, http->path, - http->hostport, http->path); -#else - sprintf(p, CBLOCK, http->hostport, http->path, http->hostport, http->path); -#endif /* def FORCE_LOAD */ + /* + * Jump to next item + */ + min = next; - return(p); + /* + * 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; + } -#ifdef IMAGE_BLOCKING /********************************************************************* * - * Function : block_imageurl + * Function : block_url * - * Description : Given a URL which is blocked, decide whether to - * send the "blocked" image or HTML. + * Description : Called from `chat'. Check to see if we need to block this. * * Parameters : - * 1 : http = URL to check. - * 2 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) * - * Returns : True (nonzero) if URL is in image list, false (0) - * otherwise + * Returns : NULL => unblocked, else HTTP block response * *********************************************************************/ -int block_imageurl(struct http_request *http, struct client_state *csp) +struct http_response *block_url(struct client_state *csp) { -#ifdef DETECT_MSIE_IMAGES - if ((csp->accept_types - & (ACCEPT_TYPE_IS_MSIE|ACCEPT_TYPE_MSIE_IMAGE|ACCEPT_TYPE_MSIE_HTML)) - == (ACCEPT_TYPE_IS_MSIE|ACCEPT_TYPE_MSIE_IMAGE)) +#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 1; + return NULL; } - else if ((csp->accept_types - & (ACCEPT_TYPE_IS_MSIE|ACCEPT_TYPE_MSIE_IMAGE|ACCEPT_TYPE_MSIE_HTML)) - == (ACCEPT_TYPE_IS_MSIE|ACCEPT_TYPE_MSIE_HTML)) + + /* + * Else, prepare a response + */ + if (NULL == (rsp = alloc_http_response())) { - return 0; + return cgi_error_memory(); } -#endif - return ((csp->action->flags & ACTION_IMAGE) != 0); + /* + * 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, html_encode(csp->http->hostport), 0); + if (!err) err = map(exports, "path", 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); + } -#endif /* def IMAGE_BLOCKING */ -#ifdef TRUST_FILES +#ifdef FEATURE_TRUST /********************************************************************* * - * Function : trust_url + * Function : trust_url FIXME: I should be called distrust_url * - * Description : Should we "trust" this URL? See "trustfile" line in config. + * 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 : http = http_request request for requested URL - * 2 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) * - * Returns : NULL => trusted, else string to HTML "untrusted" description. + * Returns : NULL => trusted, else http_response. * *********************************************************************/ -char *trust_url(struct http_request *http, struct client_state *csp) +struct http_response *trust_url(struct client_state *csp) { - struct file_list *fl; - struct block_spec *b; - struct url_spec url[1], **tl, *t; - char *p, *h; - char *hostport, *path, *refer; - struct http_request rhttp[1]; - int n; + struct http_response *rsp; + struct map * exports; + char buf[BUFFER_SIZE]; + char *p; + struct url_spec **tl; + struct url_spec *t; + jb_err err; - if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL)) + /* + * Don't bother to work on trusted URLs + */ + if (!is_untrusted_url(csp)) { - return(NULL); + return NULL; } - *url = dsplit(http->host); + /* + * Else, prepare a response: + */ + if (NULL == (rsp = alloc_http_response())) + { + return cgi_error_memory(); + } - /* if splitting the domain fails, punt */ - if (url->dbuf == NULL) return(NULL); + exports = default_exports(csp, NULL); + if (exports == NULL) + { + free_http_response(rsp); + return cgi_error_memory(); + } - memset(rhttp, '\0', sizeof(*rhttp)); + /* + * 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); - for (b = b->next; b ; b = b->next) + if (NULL != (p = get_header_value(csp->headers, "Referer:"))) { - 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 - ) - { - freez(url->dbuf); - freez(url->dvec); + if (!err) err = map(exports, "referrer", 1, html_encode(p), 0); + } + else + { + if (!err) err = map(exports, "referrer", 1, "unknown", 1); + } - if (b->reject == 0) return(NULL); + if (err) + { + free_map(exports); + free_http_response(rsp); + return cgi_error_memory(); + } - hostport = url_encode(http->hostport); - path = url_encode(http->path); + /* + * 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 (csp->referrer) - { - refer = url_encode(csp->referrer); - } - else - { - refer = url_encode("undefined"); - } + if (err) + { + free_map(exports); + free_http_response(rsp); + return cgi_error_memory(); + } - n = strlen(CTRUST); - n += strlen(hostport); - n += strlen(path); - n += strlen(refer); + /* + * Export the trust info, if available + */ + if (csp->config->trust_info->first) + { + struct list_entry *l; - p = (char *)malloc(n); + 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(); + } - sprintf(p, CTRUST, hostport, path, refer); + /* + * 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 */ - freez(hostport); - freez(path); - freez(refer); + if (err) + { + free_map(exports); + free_http_response(rsp); + return cgi_error_memory(); + } - return(p); - } - } - } + /* + * Build the response + */ + err = template_fill_for_cgi(csp, "untrusted", exports, rsp); + if (err) + { + free_http_response(rsp); + return cgi_error_memory(); } - freez(url->dbuf); - freez(url->dvec); + 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); - if ((csp->referrer == NULL)|| (strlen(csp->referrer) <= 9)) + /* + * find the last URL encoded in the request + */ + while ((p = strstr(p, "http://"))) { - /* no referrer was supplied */ - goto trust_url_not_trusted; + q = p++; } - /* forge a URL from the referrer so we can use - * convert_url() to parse it into its components. + /* + * if there was any, generate and return a HTTP redirect */ + if (q != csp->http->path) + { + log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q); - p = NULL; - p = strsav(p, "GET "); - p = strsav(p, csp->referrer + 9); /* skip over "Referer: " */ - p = strsav(p, " HTTP/1.0"); + if (NULL == (rsp = alloc_http_response())) + { + return cgi_error_memory(); + } - parse_http_request(p, rhttp, csp); + 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(); + } - if (rhttp->cmd == NULL) + return finish_http_response(rsp); + } + else { - freez(p); - goto trust_url_not_trusted; + return NULL; } - freez(p); +} +#endif /* def FEATURE_FAST_REDIRECTS */ - *url = dsplit(rhttp->host); - /* if splitting the domain fails, punt */ - if (url->dbuf == NULL) goto trust_url_not_trusted; +#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; - for (tl = csp->config->trust_list; (t = *tl) ; tl++) + tmp = get_header_value(csp->headers, "User-Agent:"); + if (tmp && strstr(tmp, "MSIE")) { - if ((t->port == 0) || (t->port == rhttp->port)) + tmp = get_header_value(csp->headers, "Accept:"); + if (tmp && strstr(tmp, "image/gif")) { - 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). - */ + /* Client will accept HTML. If this seems counterintuitive, + * blame Microsoft. + */ + return(0); + } + else + { + return(1); + } + } +#endif /* def FEATURE_IMAGE_DETECT_MSIE */ - FILE *fp; + return ((csp->action->flags & ACTION_IMAGE) != 0); - freez(url->dbuf); - freez(url->dvec); +} +#endif /* def FEATURE_IMAGE_BLOCKING */ - if ((fp = fopen(csp->config->trustfile, "a"))) - { - h = NULL; - - h = strsav(h, "~"); - h = strsav(h, http->hostport); - - p = 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, http->path); - h = strsav(h, "/"); - } - } - - free_http_request(rhttp); - - fprintf(fp, "%s\n", h); - freez(h); - fclose(fp); - } - return(NULL); - } - } - } + +#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 **trusted_url; + struct http_request rhttp[1]; + const char * referer; + jb_err err; + + /* + * If we don't have a trustlist, we trust everybody + */ + if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL)) + { + return 0; } -trust_url_not_trusted: - free_http_request(rhttp); + memset(rhttp, '\0', sizeof(*rhttp)); - hostport = url_encode(http->hostport); - path = url_encode(http->path); + /* + * Do we trust the request URL itself? + */ + for (b = b->next; b ; b = b->next) + { + if (url_match(b->url, csp->http)) + { + return b->reject; + } + } - if (csp->referrer) + if (NULL == (referer = get_header_value(csp->headers, "Referer:"))) { - refer = url_encode(csp->referrer); + /* no referrer was supplied */ + return 1; } - else + + /* + * If not, do we maybe trust its referrer? + */ + + + /* + * Parse the URL from the referrer + */ + + err = parse_http_url(referer, rhttp, csp); + if (err) { - refer = url_encode("undefined"); + return 1; } - n = strlen(CTRUST); - n += strlen(hostport); - n += strlen(path); - n += strlen(refer); + for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++) + { + if (url_match(*trusted_url, rhttp)) + { + /* 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; - p = (char *)malloc(n); - sprintf(p, CTRUST, hostport, path, refer); + if ((fp = fopen(csp->config->trustfile, "a"))) + { + char * path; + char * path_end; + char * new_entry = strdup("~"); + + string_append(&new_entry, csp->http->hostport); - freez(hostport); - freez(path); - freez(refer); + path = csp->http->path; + if ( (path[0] == '/') + && (path[1] == '~') + && ((path_end = strchr(path + 2, '/')) != NULL)) + { + /* since this path points into a user's home space + * be sure to include this spec in the trustfile. + */ + int path_len = path_end - path; /* save offset */ + path = strdup(path); /* Copy string */ + if (path != NULL) + { + path_end = path + path_len; /* regenerate ptr to new buffer */ + *(path_end + 1) = '\0'; /* Truncate path after '/' */ + } + string_join(&new_entry, path); + } - return(p); + if (new_entry != NULL) + { + fprintf(fp, "%s\n", new_entry); + free(new_entry); + } + else + { + /* FIXME: No way to handle out-of memory, so mostly ignoring it */ + log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file"); + } + fclose(fp); + } + return 0; + } + } + return 1; } -#endif /* def TRUST_FILES */ +#endif /* def FEATURE_COOKIE_JAR */ -#ifdef PCRS /********************************************************************* * - * Function : re_process_buffer + * Function : pcrs_filter_response * - * Description : Apply all jobs from the joblist (aka. Perl regexp's) to - * the text buffer that's been accumulated in csp->iob->buf - * and set csp->content_length to the modified size. + * 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 *re_process_buffer(struct client_state *csp) +char *pcrs_filter_response(struct client_state *csp) { int hits=0; - int size = csp->iob->eod - csp->iob->cur; - char *old=csp->iob->cur, *new = NULL; - pcrs_job *job, *joblist; + 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 (size <= 0) + /* Sanity first */ + if (csp->iob->cur >= csp->iob->eod) { - return(strdup("")); + 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(strdup("")); + return(NULL); } - joblist = b->joblist; - + 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 = joblist; NULL != job; job = job->next) + for (job = b->joblist; NULL != job; job = job->next) { - hits += pcrs_exec_substitution(job, old, size, &new, &size); + 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); - csp->content_length = size; + /* + * If there were no hits, destroy our copy and let + * chat() use the original in csp->iob + */ + if (!hits) + { + free(new); + return(NULL); + } - /* fwiw, reset the iob */ + csp->flags |= CSP_FLAG_MODIFIED; + csp->content_length = size; IOB_RESET(csp); + return(new); } -#endif /* def PCRS */ /********************************************************************* * - * Function : url_actions + * Function : gif_deanimate_response * - * Description : Gets the actions for this URL. + * 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 : http = http_request request for blocked URLs - * 2 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) * - * Returns : N/A + * Returns : a pointer to the (newly allocated) modified buffer. + * or NULL in case something went wrong. * *********************************************************************/ -void url_actions(struct http_request *http, - struct client_state *csp) +char *gif_deanimate_response(struct client_state *csp) { - struct file_list *fl; - struct url_actions *b; + struct binbuffer *in, *out; + char *p; + int size = csp->iob->eod - csp->iob->cur; - init_current_action(csp->action); + /* + * 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 (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL)) + if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in ))) + || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) ) { - return; + 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); } - apply_url_actions(csp->action, http, b); } /********************************************************************* * - * Function : apply_url_actions + * Function : remove_chunked_transfer_coding * - * Description : Applies a list of URL actions. + * Description : In-situ remove the "chunked" transfer coding as defined + * in rfc2616 from a buffer. * * Parameters : - * 1 : action = Destination. - * 2 : http = Current URL - * 3 : b = list of URL actions to apply + * 1 : buffer = Pointer to the text buffer + * 2 : size = Number of bytes to be processed * - * Returns : N/A + * 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 * *********************************************************************/ -void apply_url_actions(struct current_action_spec *action, - struct http_request *http, - struct url_actions *b) +int remove_chunked_transfer_coding(char *buffer, const size_t size) { - struct url_spec url[1]; + size_t newsize = 0; + unsigned int chunksize = 0; + char *from_p, *to_p; - if (b == NULL) - { - /* Should never happen */ - return; - } + assert(buffer); + from_p = to_p = buffer; - *url = dsplit(http->host); - - /* if splitting the domain fails, punt */ - if (url->dbuf == NULL) + if (sscanf(buffer, "%x", &chunksize) != 1) { - return; + log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding"); + return(0); } - for (b = b->next; NULL != b; b = b->next) + while (chunksize > 0) { - if ((b->url->port == 0) || (b->url->port == http->port)) + if (NULL == (from_p = strstr(from_p, "\r\n"))) { - 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); - } - } + 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); } } - freez(url->dbuf); - freez(url->dvec); + /* 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 : forward_url + * Function : url_actions * - * Description : Should we forward this to another proxy? + * Description : Gets the actions for this URL. * * Parameters : - * 1 : http = http_request request for current URL + * 1 : http = http_request request for blocked URLs * 2 : csp = Current client state (buffers, headers, etc...) * - * Returns : Return gw_default for no forward match, - * else a gateway pointer to a specific forwarding proxy. + * Returns : N/A * *********************************************************************/ -const struct gateway *forward_url(struct http_request *http, struct client_state *csp) +void url_actions(struct http_request *http, + struct client_state *csp) { struct file_list *fl; - struct forward_spec *b; - struct url_spec url[1]; - - if (((fl = csp->flist) == NULL) || ((b = fl->f) == NULL)) - { - return(gw_default); - } - - *url = dsplit(http->host); + struct url_actions *b; - /* if splitting the domain fails, punt */ - if (url->dbuf == NULL) return(gw_default); + init_current_action(csp->action); - for (b = b->next; b ; b = b->next) + if (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL)) { - 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 - ) - { - freez(url->dbuf); - freez(url->dvec); - return(b->gw); - } - } - } + return; } - freez(url->dbuf); - freez(url->dvec); - return(gw_default); + apply_url_actions(csp->action, http, b); } /********************************************************************* * - * Function : dsplit + * Function : apply_url_actions * - * 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. + * Description : Applies a list of URL actions. * * Parameters : - * 1 : domain = a URL address + * 1 : action = Destination. + * 2 : http = Current URL + * 3 : b = list of URL actions to apply * - * Returns : url_spec structure populated with dbuf, dcnt and dvec. + * Returns : N/A * *********************************************************************/ -struct url_spec dsplit(char *domain) +void apply_url_actions(struct current_action_spec *action, + struct http_request *http, + struct url_actions *b) { - struct url_spec ret[1]; - char *v[BUFSIZ]; - 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); - - /* map to lower case */ - for (p = ret->dbuf; *p ; p++) *p = tolower(*p); - - /* split the domain name into components */ - ret->dcnt = ssplit(ret->dbuf, ".", v, SZ(v), 1, 1); - - if (ret->dcnt <= 0) + if (b == NULL) { - memset(ret, '\0', sizeof(ret)); - return(*ret); + /* Should never happen */ + return; } - /* save a copy of the pointers in dvec */ - size = ret->dcnt * sizeof(*ret->dvec); - - if ((ret->dvec = (char **)malloc(size))) + for (b = b->next; NULL != b; b = b->next) { - memcpy(ret->dvec, v, size); + if (url_match(b->url, http)) + { + merge_current_action(action, b->action); + } } - - - return(*ret); - } /********************************************************************* * - * Function : domaincmp + * Function : forward_url * - * 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 - * trivialmatch(). + * Description : Should we forward this to another proxy? * * Parameters : - * 1 : pattern = a domain that may contain a '*' as a wildcard. - * 2 : fqdn = domain name against which the patterns are compared. + * 1 : http = http_request request for current URL + * 2 : csp = Current client state (buffers, headers, etc...) * - * Returns : 0 => domains are equivalent, else no match. + * Returns : Pointer to forwarding information. * *********************************************************************/ -int domaincmp(struct url_spec *pattern, struct url_spec *fqdn) +const struct forward_spec * forward_url(struct http_request *http, + struct client_state *csp) { - char **pv, **fv; /* vectors */ - int pn, fn; /* counters */ - char *p, *f; /* chars */ - - pv = pattern->dvec; - fv = fqdn->dvec; - fn = pn = 0; + static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER }; + struct forward_spec *fwd = csp->config->forward; - while (fn < fqdn->dcnt && pn < pattern->dcnt) + if (fwd == NULL) { - p = pv[pn]; - f = fv[fn]; + return fwd_default; + } - if (simplematch(p, f)) - { - if(pn || !(pattern->unanchored & ANCHOR_LEFT)) - { - return 1; - } - } - else + while (fwd != NULL) + { + if (url_match(fwd->url, http)) { - pn++; + return fwd; } - fn++; + fwd = fwd->next; } - return ((pn < pattern->dcnt) || ((fn < fqdn->dcnt) && !(pattern->unanchored & ANCHOR_RIGHT))); - + return fwd_default; } - /* Local Variables: tab-width: 3