X-Git-Url: http://www.privoxy.org/gitweb/?p=privoxy.git;a=blobdiff_plain;f=filters.c;h=a08dcba1dd92a82781775bb7c99103d5bb488957;hp=9b456040a8c9d78ed046398c7c15fdaf1798525a;hb=5af5acaa9aa8c28ebcb6eb7be8a5d61b1a4fbf34;hpb=e5976382c471de052e5e2ad6bccbf7b9e1e4a643 diff --git a/filters.c b/filters.c index 9b456040..a08dcba1 100644 --- a/filters.c +++ b/filters.c @@ -1,4 +1,4 @@ -const char filters_rcs[] = "$Id: filters.c,v 1.34 2001/09/20 15:49:36 steudten Exp $"; +const char filters_rcs[] = "$Id: filters.c,v 1.127 2010/04/03 13:22:56 fabiankeil Exp $"; /********************************************************************* * * File : $Source: /cvsroot/ijbswa/current/filters.c,v $ @@ -6,19 +6,21 @@ const char filters_rcs[] = "$Id: filters.c,v 1.34 2001/09/20 15:49:36 steudten E * Purpose : Declares functions to parse/crunch headers and pages. * Functions declared include: * `acl_addr', `add_stats', `block_acl', `block_imageurl', - * `block_url', `url_actions', `domaincmp', `dsplit', + * `block_url', `url_actions', `domain_split', * `filter_popups', `forward_url', 'redirect_url', * `ij_untrusted_url', `intercept_url', `pcrs_filter_respose', - * `show_proxy_args', 'ijb_send_banner', and `trust_url' + * `ijb_send_banner', `trust_url', `gif_deanimate_response', + * `execute_single_pcrs_command', `rewrite_url', + * `get_last_url' * - * Copyright : Written by and Copyright (C) 2001 the SourceForge - * IJBSWA team. http://ijbswa.sourceforge.net + * Copyright : Written by and Copyright (C) 2001, 2004-2009 the + * Privoxy team. http://www.privoxy.org/ * * Based on the Internet Junkbuster originally written - * by and Copyright (C) 1997 Anonymous Coders and + * by and Copyright (C) 1997 Anonymous Coders and * Junkbusters Corporation. http://www.junkbusters.com * - * This program is free software; you can redistribute it + * This program is free software; you can redistribute it * and/or modify it under the terms of the GNU General * Public License as published by the Free Software * Foundation; either version 2 of the License, or (at @@ -36,253 +38,8 @@ const char filters_rcs[] = "$Id: filters.c,v 1.34 2001/09/20 15:49:36 steudten E * or write to the Free Software Foundation, Inc., 59 * Temple Place - Suite 330, Boston, MA 02111-1307, USA. * - * Revisions : - * $Log: filters.c,v $ - * Revision 1.34 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" @@ -293,12 +50,23 @@ const char filters_rcs[] = "$Id: filters.c,v 1.34 2001/09/20 15:49:36 steudten E #include #include +#ifdef HAVE_RFC2553 +#include +#include +#endif /* def HAVE_RFC2553 */ + #ifndef _WIN32 +#ifndef __OS2__ #include +#endif /* ndef __OS2__ */ #include #else #include -#endif +#endif /* ndef _WIN32 */ + +#ifdef __OS2__ +#include +#endif /* def __OS2__ */ #include "project.h" #include "filters.h" @@ -312,6 +80,8 @@ const char filters_rcs[] = "$Id: filters.c,v 1.34 2001/09/20 15:49:36 steudten E #include "cgi.h" #include "list.h" #include "deanimate.h" +#include "urlmatch.h" +#include "loaders.h" #ifdef _WIN32 #include "win32.h" @@ -328,8 +98,156 @@ const char filters_h_rcs[] = FILTERS_H_VERSION; */ #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X)) +static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size); +static jb_err prepare_for_filtering(struct client_state *csp); #ifdef FEATURE_ACL +#ifdef HAVE_RFC2553 +/********************************************************************* + * + * Function : sockaddr_storage_to_ip + * + * Description : Access internal structure of sockaddr_storage + * + * Parameters : + * 1 : addr = socket address + * 2 : ip = IP address as array of octets in network order + * (it points into addr) + * 3 : len = length of IP address in octets + * 4 : port = port number in network order; + * + * Returns : 0 = no errror; -1 otherwise. + * + *********************************************************************/ +static int sockaddr_storage_to_ip(const struct sockaddr_storage *addr, + uint8_t **ip, unsigned int *len, + in_port_t **port) +{ + if (NULL == addr) + { + return(-1); + } + + switch (addr->ss_family) + { + case AF_INET: + if (NULL != len) + { + *len = 4; + } + if (NULL != ip) + { + *ip = (uint8_t *) + &(((struct sockaddr_in *)addr)->sin_addr.s_addr); + } + if (NULL != port) + { + *port = &((struct sockaddr_in *)addr)->sin_port; + } + break; + + case AF_INET6: + if (NULL != len) + { + *len = 16; + } + if (NULL != ip) + { + *ip = ((struct sockaddr_in6 *)addr)->sin6_addr.s6_addr; + } + if (NULL != port) + { + *port = &((struct sockaddr_in6 *)addr)->sin6_port; + } + break; + + default: + /* Unsupported address family */ + return(-1); + } + + return(0); +} + + +/********************************************************************* + * + * Function : match_sockaddr + * + * Description : Check whether address matches network (IP address and port) + * + * Parameters : + * 1 : network = socket address of subnework + * 2 : netmask = network mask as socket address + * 3 : address = checked socket address against given network + * + * Returns : 0 = doesn't match; 1 = does match + * + *********************************************************************/ +static int match_sockaddr(const struct sockaddr_storage *network, + const struct sockaddr_storage *netmask, + const struct sockaddr_storage *address) +{ + uint8_t *network_addr, *netmask_addr, *address_addr; + unsigned int addr_len; + in_port_t *network_port, *netmask_port, *address_port; + int i; + + if (network->ss_family != netmask->ss_family) + { + /* This should never happen */ + log_error(LOG_LEVEL_ERROR, + "Internal error at %s:%llu: network and netmask differ in family", + __FILE__, __LINE__); + return 0; + } + + sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port); + sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port); + sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port); + + /* Check for family */ + if ((network->ss_family == AF_INET) && (address->ss_family == AF_INET6) + && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr)) + { + /* Map AF_INET6 V4MAPPED address into AF_INET */ + address_addr += 12; + addr_len = 4; + } + else if ((network->ss_family == AF_INET6) && (address->ss_family == AF_INET) + && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr)) + { + /* Map AF_INET6 V4MAPPED network into AF_INET */ + network_addr += 12; + netmask_addr += 12; + addr_len = 4; + } + else if (network->ss_family != address->ss_family) + { + return 0; + } + + /* XXX: Port check is signaled in netmask */ + if (*netmask_port && *network_port != *address_port) + { + return 0; + } + + /* TODO: Optimize by checking by words insted of octets */ + for (i = 0; (i < addr_len) && netmask_addr[i]; i++) + { + if ((network_addr[i] & netmask_addr[i]) != + (address_addr[i] & netmask_addr[i])) + { + return 0; + } + } + + return 1; +} +#endif /* def HAVE_RFC2553 */ + + /********************************************************************* * * Function : block_acl @@ -346,7 +264,7 @@ 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(const struct access_control_addr *dst, const struct client_state *csp) { struct access_control_list *acl = csp->config->acl; @@ -359,7 +277,13 @@ int block_acl(struct access_control_addr *dst, struct client_state *csp) /* search the list */ while (acl != NULL) { - if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr) + if ( +#ifdef HAVE_RFC2553 + match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr) +#else + (csp->ip_addr_long & acl->src->mask) == acl->src->addr +#endif + ) { if (dst == NULL) { @@ -369,8 +293,23 @@ int block_acl(struct access_control_addr *dst, struct client_state *csp) return(0); } } - else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr) - && ((dst->port == acl->dst->port) || (acl->dst->port == 0))) + else if ( +#ifdef HAVE_RFC2553 + /* + * XXX: An undefined acl->dst is full of zeros and should be + * considered a wildcard address. sockaddr_storage_to_ip() + * fails on such destinations because of unknown sa_familly + * (glibc only?). However this test is not portable. + * + * So, we signal the acl->dst is wildcard in wildcard_dst. + */ + acl->wildcard_dst || + match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr) +#else + ((dst->addr & acl->dst->mask) == acl->dst->addr) + && ((dst->port == acl->dst->port) || (acl->dst->port == 0)) +#endif + ) { if (acl->action == ACL_PERMIT) { @@ -394,7 +333,7 @@ int block_acl(struct access_control_addr *dst, struct client_state *csp) * * Function : acl_addr * - * Description : Called from `load_aclfile' to parse an ACL address. + * Description : Called from `load_config' to parse an ACL address. * * Parameters : * 1 : aspec = String specifying ACL address. @@ -403,62 +342,182 @@ int block_acl(struct access_control_addr *dst, struct client_state *csp) * Returns : 0 => Ok, everything else is an error. * *********************************************************************/ -int acl_addr(char *aspec, struct access_control_addr *aca) +int acl_addr(const char *aspec, struct access_control_addr *aca) { - int i, masklength, port; + int i, masklength; +#ifdef HAVE_RFC2553 + struct addrinfo hints, *result; + uint8_t *mask_data; + in_port_t *mask_port; + unsigned int addr_len; +#else + long port; +#endif /* def HAVE_RFC2553 */ char *p; + char *acl_spec = NULL; +#ifdef HAVE_RFC2553 + /* XXX: Depend on ai_family */ + masklength = 128; +#else masklength = 32; port = 0; +#endif - if ((p = strchr(aspec, '/'))) + /* + * Use a temporary acl spec copy so we can log + * the unmodified original in case of parse errors. + */ + acl_spec = strdup(aspec); + if (acl_spec == NULL) { - *p++ = '\0'; + /* XXX: This will be logged as parse error. */ + return(-1); + } + if ((p = strchr(acl_spec, '/')) != NULL) + { + *p++ = '\0'; if (ijb_isdigit(*p) == 0) { + freez(acl_spec); return(-1); } masklength = atoi(p); } - if ((masklength < 0) || (masklength > 32)) + if ((masklength < 0) || +#ifdef HAVE_RFC2553 + (masklength > 128) +#else + (masklength > 32) +#endif + ) + { + freez(acl_spec); + return(-1); + } + + if ((*acl_spec == '[') && (NULL != (p = strchr(acl_spec, ']')))) + { + *p = '\0'; + memmove(acl_spec, acl_spec + 1, (size_t)(p - acl_spec)); + + if (*++p != ':') + { + p = NULL; + } + } + else + { + p = strchr(acl_spec, ':'); + } + +#ifdef HAVE_RFC2553 + memset(&hints, 0, sizeof(struct addrinfo)); + hints.ai_family = AF_UNSPEC; + hints.ai_socktype = SOCK_STREAM; + + i = getaddrinfo(acl_spec, ((p) ? ++p : NULL), &hints, &result); + + if (i != 0) { + log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s", + acl_spec, p, gai_strerror(i)); + freez(acl_spec); return(-1); } + freez(acl_spec); - if ((p = strchr(aspec, ':'))) + /* TODO: Allow multihomed hostnames */ + memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen); + freeaddrinfo(result); +#else + if (p != NULL) { + char *endptr; + *p++ = '\0'; + port = strtol(p, &endptr, 10); - if (ijb_isdigit(*p) == 0) + if (port <= 0 || port > 65535 || *endptr != '\0') { + freez(acl_spec); return(-1); } - port = atoi(p); } - aca->port = port; + aca->port = (unsigned long)port; - aca->addr = ntohl(resolve_hostname_to_ip(aspec)); + aca->addr = ntohl(resolve_hostname_to_ip(acl_spec)); + freez(acl_spec); - if (aca->addr == -1) + if (aca->addr == INADDR_NONE) { - log_error(LOG_LEVEL_ERROR, "can't resolve address for %s", aspec); + /* XXX: This will be logged as parse error. */ return(-1); } +#endif /* def HAVE_RFC2553 */ /* build the netmask */ +#ifdef HAVE_RFC2553 + /* Clip masklength according to current family. */ + if ((aca->addr.ss_family == AF_INET) && (masklength > 32)) + { + masklength = 32; + } + + aca->mask.ss_family = aca->addr.ss_family; + if (sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port)) + { + return(-1); + } + + if (p) + { + /* ACL contains a port number, check ports in the future. */ + *mask_port = 1; + } + + /* + * XXX: This could be optimized to operate on whole words instead + * of octets (128-bit CPU could do it in one iteration). + */ + /* + * Octets after prefix can be ommitted because of + * previous initialization to zeros. + */ + for (i = 0; (i < addr_len) && masklength; i++) + { + if (masklength >= 8) + { + mask_data[i] = 0xFF; + masklength -= 8; + } + else + { + /* + * XXX: This assumes MSB of octet is on the left side. + * This should be true for all architectures or solved + * by the link layer. + */ + mask_data[i] = (uint8_t)~((1 << (8 - masklength)) - 1); + masklength = 0; + } + } + +#else aca->mask = 0; for (i=1; i <= masklength ; i++) { - aca->mask |= (1 << (32 - i)); + aca->mask |= (1U << (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; +#endif /* def HAVE_RFC2553 */ return(0); @@ -466,6 +525,28 @@ int acl_addr(char *aspec, struct access_control_addr *aca) #endif /* def FEATURE_ACL */ +/********************************************************************* + * + * Function : connect_port_is_forbidden + * + * Description : Check to see if CONNECT requests to the destination + * port of this request are forbidden. The check is + * independend of the actual request method. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * + * Returns : True if yes, false otherwise. + * + *********************************************************************/ +int connect_port_is_forbidden(const struct client_state *csp) +{ + return ((csp->action->flags & ACTION_LIMIT_CONNECT) && + !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT], + csp->http->port)); +} + + /********************************************************************* * * Function : block_url @@ -480,25 +561,26 @@ int acl_addr(char *aspec, struct access_control_addr *aca) *********************************************************************/ struct http_response *block_url(struct client_state *csp) { -#ifdef FEATURE_IMAGE_BLOCKING - char *p; -#endif /* def FEATURE_IMAGE_BLOCKING */ struct http_response *rsp; + const char *new_content_type = NULL; - /* + /* * If it's not blocked, don't block it ;-) */ if ((csp->action->flags & ACTION_BLOCK) == 0) { - return(NULL); + return NULL; } - - /* + if (csp->action->flags & ACTION_REDIRECT) + { + log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block."); + } + /* * Else, prepare a response */ if (NULL == (rsp = alloc_http_response())) { - return NULL; + return cgi_error_memory(); } /* @@ -509,74 +591,213 @@ struct http_response *block_url(struct client_state *csp) if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0) && is_imageurl(csp)) { + char *p; /* determine HOW images should be blocked */ p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER]; - /* and handle accordingly: */ - if ((p == NULL) || (0 == strcmpic(p, "logo"))) + if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT) { - rsp->body = bindup(image_junkbuster_gif_data, image_junkbuster_gif_length); - rsp->content_length = image_junkbuster_gif_length; - enlist_unique_header(rsp->headers, "Content-Type", "image/gif"); + log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image."); } - else if (0 == strcmpic(p, "blank")) + /* and handle accordingly: */ + if ((p == NULL) || (0 == strcmpic(p, "pattern"))) { - rsp->body = bindup(image_blank_gif_data, image_blank_gif_length); - rsp->content_length = image_blank_gif_length; - enlist_unique_header(rsp->headers, "Content-Type", "image/gif"); + rsp->status = strdup("403 Request blocked by Privoxy"); + if (rsp->status == NULL) + { + free_http_response(rsp); + return cgi_error_memory(); + } + rsp->body = bindup(image_pattern_data, image_pattern_length); + if (rsp->body == NULL) + { + free_http_response(rsp); + return cgi_error_memory(); + } + rsp->content_length = image_pattern_length; + + if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE)) + { + free_http_response(rsp); + return cgi_error_memory(); + } } + else if (0 == strcmpic(p, "blank")) + { + rsp->status = strdup("403 Request blocked by Privoxy"); + if (rsp->status == NULL) + { + free_http_response(rsp); + return cgi_error_memory(); + } + rsp->body = bindup(image_blank_data, image_blank_length); + if (rsp->body == NULL) + { + free_http_response(rsp); + return cgi_error_memory(); + } + rsp->content_length = image_blank_length; + if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE)) + { + free_http_response(rsp); + return cgi_error_memory(); + } + } else { - rsp->status = strdup("302 Local Redirect from Junkbuster"); - enlist_unique_header(rsp->headers, "Location", p); + rsp->status = strdup("302 Local Redirect from Privoxy"); + 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 */ + if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT) + { + /* + * Send empty document. + */ + new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE]; - /* + freez(rsp->body); + rsp->body = strdup(" "); + rsp->content_length = 1; + + if (csp->config->feature_flags & RUNTIME_FEATURE_EMPTY_DOC_RETURNS_OK) + { + /* + * Workaround for firefox bug 492459 + * https://bugzilla.mozilla.org/show_bug.cgi?id=492459 + * Return a 200 OK status for pages blocked with +handle-as-empty-document + * if the "handle-as-empty-doc-returns-ok" runtime config option is set. + */ + rsp->status = strdup("200 Request blocked by Privoxy"); + } + else + { + rsp->status = strdup("403 Request blocked by Privoxy"); + } + + if (rsp->status == NULL) + { + free_http_response(rsp); + return cgi_error_memory(); + } + if (new_content_type != 0) + { + log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type); + if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type)) + { + free_http_response(rsp); + return cgi_error_memory(); + } + } + } + else + + /* * Else, generate an HTML "blocked" message: */ { - struct map * exports = default_exports(csp, NULL); -#ifdef FEATURE_FORCE_LOAD - map(exports, "force-prefix", 1, FORCE_PREFIX, 1); -#else /* ifndef FEATURE_FORCE_LOAD */ - map_block_killer(exports, "force-support"); -#endif /* ndef FEATURE_FORCE_LOAD */ - - map(exports, "hostport", 1, csp->http->hostport, 1); - map(exports, "hostport-html", 1, html_encode(csp->http->hostport), 0); - map(exports, "path", 1, csp->http->path, 1); - map(exports, "path-html", 1, html_encode(csp->http->path), 0); + jb_err err; + struct map * exports; + char *p; - rsp->body = template_load(csp, "blocked"); - template_fill(&rsp->body, exports); - free_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 (csp->http->user_agent - && !strncmpic(csp->http->user_agent, "mozilla", 7) - && !strstr(csp->http->user_agent, "compatible") - && !strstr(csp->http->user_agent, "Opera")) + 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"); + rsp->status = strdup("200 Request for blocked URL"); } else { - rsp->status = strdup("404 Request for blocked URL"); + rsp->status = strdup("403 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); + /* + * Export the force conditional block killer if + * + * - Privoxy was compiled without FEATURE_FORCE_LOAD, or + * - Privoxy is configured to enforce blocks, or + * - it's a CONNECT request and enforcing wouldn't work anyway. + */ + if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS) + || (0 == strcmpic(csp->http->gpc, "connect"))) +#endif /* ndef FEATURE_FORCE_LOAD */ + { + err = map_block_killer(exports, "force-support"); + } + + if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1); + 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) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0); + if (!err) + { + const char *block_reason; + if (csp->action->string[ACTION_STRING_BLOCK] != NULL) + { + block_reason = csp->action->string[ACTION_STRING_BLOCK]; + } + else + { + assert(connect_port_is_forbidden(csp)); + block_reason = "Forbidden CONNECT port."; + } + err = map(exports, "block-reason", 1, html_encode(block_reason), 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(); + } } + rsp->reason = RSP_REASON_BLOCKED; - return(finish_http_response(rsp)); + return finish_http_response(csp, rsp); } @@ -587,7 +808,7 @@ struct http_response *block_url(struct client_state *csp) * 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. + * and if not, returns a HTTP 403 response with a reject message. * * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) @@ -599,8 +820,11 @@ struct http_response *trust_url(struct client_state *csp) { struct http_response *rsp; struct map * exports; - char buf[BUFFER_SIZE], *p = NULL; - struct url_spec **tl, *t; + 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 @@ -610,45 +834,62 @@ struct http_response *trust_url(struct client_state *csp) return NULL; } - /* + /* * Else, prepare a response: */ if (NULL == (rsp = alloc_http_response())) { - return NULL; + return cgi_error_memory(); } + rsp->status = strdup("403 Request blocked by Privoxy"); exports = default_exports(csp, NULL); + if (exports == NULL || rsp->status == NULL) + { + free_http_response(rsp); + return cgi_error_memory(); + } - /* - * Export the host, port, and referrer information + /* + * Export the protocol, host, port, and referrer information */ - map(exports, "hostport", 1, csp->http->hostport, 1); - map(exports, "path", 1, csp->http->path, 1); - map(exports, "hostport-html", 1, html_encode(csp->http->hostport), 0); - map(exports, "path-html", 1, html_encode(csp->http->path), 0); + err = map(exports, "hostport", 1, csp->http->hostport, 1); + if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1); + if (!err) err = map(exports, "path", 1, csp->http->path, 1); - if (csp->referrer && strlen(csp->referrer) > 9) + if (NULL != (p = get_header_value(csp->headers, "Referer:"))) { - map(exports, "referrer", 1, csp->referrer + 9, 1); - map(exports, "referrer-html", 1, html_encode(csp->referrer + 9), 0); + if (!err) err = map(exports, "referrer", 1, html_encode(p), 0); } else { - map(exports, "referrer", 1, "unknown", 1); - map(exports, "referrer-html", 1, "unknown", 1); + if (!err) err = map(exports, "referrer", 1, "none set", 1); + } + + if (err) + { + free_map(exports); + free_http_response(rsp); + return cgi_error_memory(); } /* * Export the trust list */ - for (tl = csp->config->trust_list; (t = *tl) ; tl++) + p = strdup(""); + for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++) { - sprintf(buf, "
  • %s
    • \n", t->spec); - p = strsav(p, buf); + snprintf(buf, sizeof(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(); } - map(exports, "trusted-referrers", 1, p, 0); - p = NULL; /* * Export the trust info, if available @@ -657,94 +898,414 @@ struct http_response *trust_url(struct client_state *csp) { 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); - p = strsav(p, buf); + snprintf(buf, sizeof(buf), "
  • %s
    \n", l->str, l->str); + string_append(&p, buf); } - map(exports, "trust-info", 1, p, 0); + err = map(exports, "trust-info", 1, p, 0); } else { - map_block_killer(exports, "have-trust-info"); + 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 + * Export the force conditional block killer if + * + * - Privoxy was compiled without FEATURE_FORCE_LOAD, or + * - Privoxy is configured to enforce blocks, or + * - it's a CONNECT request and enforcing wouldn't work anyway. */ #ifdef FEATURE_FORCE_LOAD - map(exports, "force-prefix", 1, FORCE_PREFIX, 1); + if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS) + || (0 == strcmpic(csp->http->gpc, "connect"))) + { + err = map_block_killer(exports, "force-support"); + } + else + { + err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1); + } #else /* ifndef FEATURE_FORCE_LOAD */ - map_block_killer(exports, "force-support"); + 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 */ - rsp->body = template_load(csp, "untrusted"); - template_fill(&rsp->body, exports); - free_map(exports); - - return(finish_http_response(rsp)); + err = template_fill_for_cgi(csp, "untrusted", exports, rsp); + if (err) + { + free_http_response(rsp); + return cgi_error_memory(); + } + rsp->reason = RSP_REASON_UNTRUSTED; + return finish_http_response(csp, rsp); } #endif /* def FEATURE_TRUST */ -#ifdef FEATURE_FAST_REDIRECTS /********************************************************************* * - * Function : redirect_url + * Function : compile_dynamic_pcrs_job_list * - * Description : Checks for redirection URLs and returns a HTTP redirect - * to the destination URL, if necessary + * Description : Compiles a dynamic pcrs job list (one with variables + * resolved at request time) * * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : b = The filter list to compile * - * Returns : NULL if URL was clean, HTTP redirect otherwise. + * Returns : NULL in case of errors, otherwise the + * pcrs job list. * *********************************************************************/ -struct http_response *redirect_url(struct client_state *csp) +pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b) { - char *p, *q; - struct http_response *rsp; + struct list_entry *pattern; + pcrs_job *job_list = NULL; + pcrs_job *dummy = NULL; + pcrs_job *lastjob = NULL; + int error = 0; - p = q = csp->http->path; - log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p); + const struct pcrs_variable variables[] = + { + {"url", csp->http->url, 1}, + {"path", csp->http->path, 1}, + {"host", csp->http->host, 1}, + {"origin", csp->ip_addr_str, 1}, + {NULL, NULL, 1} + }; + + for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next) + { + assert(pattern->str != NULL); - /* - * find the last URL encoded in the request + dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error); + if (NULL == dummy) + { + assert(error < 0); + log_error(LOG_LEVEL_ERROR, + "Adding filter job \'%s\' to dynamic filter %s failed: %s", + pattern->str, b->name, pcrs_strerror(error)); + continue; + } + else + { + if (error == PCRS_WARN_TRUNCATION) + { + log_error(LOG_LEVEL_ERROR, + "At least one of the variables in \'%s\' had to " + "be truncated before compilation", pattern->str); + } + if (job_list == NULL) + { + job_list = dummy; + } + else + { + lastjob->next = dummy; + } + lastjob = dummy; + } + } + + return job_list; +} + + +/********************************************************************* + * + * Function : rewrite_url + * + * Description : Rewrites a URL with a single pcrs command + * and returns the result if it differs from the + * original and isn't obviously invalid. + * + * Parameters : + * 1 : old_url = URL to rewrite. + * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@) + * + * + * Returns : NULL if the pcrs_command didn't change the url, or + * the result of the modification. + * + *********************************************************************/ +char *rewrite_url(char *old_url, const char *pcrs_command) +{ + char *new_url = NULL; + int hits; + + assert(old_url); + assert(pcrs_command); + + new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits); + + if (hits == 0) + { + log_error(LOG_LEVEL_REDIRECTS, + "pcrs command \"%s\" didn't change \"%s\".", + pcrs_command, old_url); + freez(new_url); + } + else if (hits < 0) + { + log_error(LOG_LEVEL_REDIRECTS, + "executing pcrs command \"%s\" to rewrite %s failed: %s", + pcrs_command, old_url, pcrs_strerror(hits)); + freez(new_url); + } + else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8)) + { + log_error(LOG_LEVEL_ERROR, + "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), " + "but the result doesn't look like a valid URL and will be ignored.", + pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts"); + freez(new_url); + } + else + { + log_error(LOG_LEVEL_REDIRECTS, + "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).", + pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts"); + } + + return new_url; + +} + + +#ifdef FEATURE_FAST_REDIRECTS +/********************************************************************* + * + * Function : get_last_url + * + * Description : Search for the last URL inside a string. + * If the string already is a URL, it will + * be the first URL found. + * + * Parameters : + * 1 : subject = the string to check + * 2 : redirect_mode = +fast-redirect{} mode + * + * Returns : NULL if no URL was found, or + * the last URL found. + * + *********************************************************************/ +char *get_last_url(char *subject, const char *redirect_mode) +{ + char *new_url = NULL; + char *tmp; + + assert(subject); + assert(redirect_mode); + + subject = strdup(subject); + if (subject == NULL) + { + log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects."); + return NULL; + } + + if (0 == strcmpic(redirect_mode, "check-decoded-url")) + { + log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject); + new_url = url_decode(subject); + if (new_url != NULL) + { + freez(subject); + subject = new_url; + } + else + { + log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject); + } + } + + log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject); + + /* + * Find the last URL encoded in the request */ - while ((p = strstr(p, "http://"))) + tmp = subject; + while ((tmp = strstr(tmp, "http://")) != NULL) { - q = p++; + new_url = tmp++; + } + tmp = (new_url != NULL) ? new_url : subject; + while ((tmp = strstr(tmp, "https://")) != NULL) + { + new_url = tmp++; } - /* - * if there was any, generate and return a HTTP redirect + if ((new_url != NULL) + && ( (new_url != subject) + || (0 == strncmpic(subject, "http://", 7)) + || (0 == strncmpic(subject, "https://", 8)) + )) + { + /* + * Return new URL if we found a redirect + * or if the subject already was a URL. + * + * The second case makes sure that we can + * chain get_last_url after another redirection check + * (like rewrite_url) without losing earlier redirects. + */ + new_url = strdup(new_url); + freez(subject); + return new_url; + } + + freez(subject); + return NULL; + +} +#endif /* def FEATURE_FAST_REDIRECTS */ + + +/********************************************************************* + * + * Function : redirect_url + * + * Description : Checks if Privoxy should answer the request with + * a HTTP redirect and generates the redirect if + * necessary. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * + * Returns : NULL if the request can pass, HTTP redirect otherwise. + * + *********************************************************************/ +struct http_response *redirect_url(struct client_state *csp) +{ + struct http_response *rsp; +#ifdef FEATURE_FAST_REDIRECTS + /* + * XXX: Do we still need FEATURE_FAST_REDIRECTS + * as compile-time option? The user can easily disable + * it in his action file. */ - if (q != csp->http->path) + char * redirect_mode; +#endif /* def FEATURE_FAST_REDIRECTS */ + char *old_url = NULL; + char *new_url = NULL; + char *redirection_string; + + if ((csp->action->flags & ACTION_REDIRECT)) { - log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q); + redirection_string = csp->action->string[ACTION_STRING_REDIRECT]; + + /* + * If the redirection string begins with 's', + * assume it's a pcrs command, otherwise treat it as + * properly formatted URL and use it for the redirection + * directly. + * + * According to RFC 2616 section 14.30 the URL + * has to be absolute and if the user tries: + * +redirect{shit/this/will/be/parsed/as/pcrs_command.html} + * she would get undefined results anyway. + * + */ - if (NULL == (rsp = alloc_http_response())) + if (*redirection_string == 's') + { + old_url = csp->http->url; + new_url = rewrite_url(old_url, redirection_string); + } + else { - return NULL; + log_error(LOG_LEVEL_REDIRECTS, + "No pcrs command recognized, assuming that \"%s\" is already properly formatted.", + redirection_string); + new_url = strdup(redirection_string); } + } - rsp->status = strdup("302 Local Redirect from Junkbuster"); - enlist_unique_header(rsp->headers, "Location", q); +#ifdef FEATURE_FAST_REDIRECTS + if ((csp->action->flags & ACTION_FAST_REDIRECTS)) + { + redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS]; - return(finish_http_response(rsp)); + /* + * If it exists, use the previously rewritten URL as input + * otherwise just use the old path. + */ + old_url = (new_url != NULL) ? new_url : strdup(csp->http->path); + new_url = get_last_url(old_url, redirect_mode); + freez(old_url); } - else + + /* + * Disable redirect checkers, so that they + * will be only run more than once if the user + * also enables them through tags. + * + * From a performance point of view + * it doesn't matter, but the duplicated + * log messages are annoying. + */ + csp->action->flags &= ~ACTION_FAST_REDIRECTS; +#endif /* def FEATURE_FAST_REDIRECTS */ + csp->action->flags &= ~ACTION_REDIRECT; + + /* Did any redirect action trigger? */ + if (new_url) { - return(NULL); + if (0 == strcmpic(new_url, csp->http->url)) + { + log_error(LOG_LEVEL_ERROR, + "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.", + csp->http->url, new_url); + freez(new_url); + } + else + { + log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url); + + if (NULL == (rsp = alloc_http_response())) + { + freez(new_url); + return cgi_error_memory(); + } + + if ( enlist_unique_header(rsp->headers, "Location", new_url) + || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) ) + { + freez(new_url); + free_http_response(rsp); + return cgi_error_memory(); + } + rsp->reason = RSP_REASON_REDIRECTED; + freez(new_url); + + return finish_http_response(csp, rsp); + } } + /* Only reached if no redirect is required */ + return NULL; + } -#endif /* def FEATURE_FAST_REDIRECTS */ #ifdef FEATURE_IMAGE_BLOCKING @@ -754,9 +1315,10 @@ struct http_response *redirect_url(struct client_state *csp) * * 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. - * + * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser + * is MSIE and not on a Mac, tell from the browser's accept + * header. + * * Parameters : * 1 : csp = Current client state (buffers, headers, etc...) * @@ -764,20 +1326,26 @@ struct http_response *redirect_url(struct client_state *csp) * otherwise * *********************************************************************/ -int is_imageurl(struct client_state *csp) +int is_imageurl(const struct client_state *csp) { #ifdef FEATURE_IMAGE_DETECT_MSIE - if ((csp->accept_types - & (ACCEPT_TYPE_IS_MSIE|ACCEPT_TYPE_MSIE_IMAGE|ACCEPT_TYPE_MSIE_HTML)) - == (ACCEPT_TYPE_IS_MSIE|ACCEPT_TYPE_MSIE_IMAGE)) - { - return 1; - } - 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)) + char *tmp; + + tmp = get_header_value(csp->headers, "User-Agent:"); + if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_")) { - return 0; + 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 */ @@ -787,7 +1355,7 @@ int is_imageurl(struct client_state *csp) #endif /* def FEATURE_IMAGE_BLOCKING */ -#ifdef FEATURE_COOKIE_JAR +#ifdef FEATURE_TRUST /********************************************************************* * * Function : is_untrusted_url @@ -805,234 +1373,301 @@ int is_imageurl(struct client_state *csp) * Returns : 0 => trusted, 1 => untrusted * *********************************************************************/ -int is_untrusted_url(struct client_state *csp) +int is_untrusted_url(const struct client_state *csp) { struct file_list *fl; struct block_spec *b; - struct url_spec url[1], **tl, *t; + struct url_spec **trusted_url; struct http_request rhttp[1]; - char *p, *h; + 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); + return 0; } + memset(rhttp, '\0', sizeof(*rhttp)); /* * 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 (url_match(b->url, csp->http)) { - 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); - } - } + return b->reject; } } - freez(url->dbuf); - freez(url->dvec); - - if ((csp->referrer == NULL)|| (strlen(csp->referrer) <= 9)) + if (NULL == (referer = 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, csp->referrer + 9); /* skip over "Referer: " */ - p = strsav(p, " HTTP/1.0"); - - parse_http_request(p, rhttp, csp); - freez(p); - - if (rhttp->cmd == NULL) - { - return(1); + 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); + err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL); + if (err) + { + return 1; + } - for (tl = csp->config->trust_list; (t = *tl) ; tl++) + for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++) { - if ((t->port == 0) || (t->port == rhttp->port)) + if (url_match(*trusted_url, rhttp)) { - 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). - */ + /* if the URL's referrer is from a trusted referrer, then + * add the target spec to the trustfile as an unblocked + * domain and return 0 (which means it's OK). + */ - FILE *fp; + FILE *fp; - freez(url->dbuf); - freez(url->dvec); + if (NULL != (fp = fopen(csp->config->trustfile, "a"))) + { + char * path; + char * path_end; + char * new_entry = strdup("~"); - if ((fp = fopen(csp->config->trustfile, "a"))) - { - h = NULL; + string_append(&new_entry, csp->http->hostport); - h = strsav(h, "~"); - h = strsav(h, csp->http->hostport); + 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. + */ + long 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); + } - 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, "/"); - } - } + /* + * Give a reason for generating this entry. + */ + string_append(&new_entry, " # Trusted referrer was: "); + string_append(&new_entry, referer); - fprintf(fp, "%s\n", h); - freez(h); - fclose(fp); + if (new_entry != NULL) + { + if (-1 == fprintf(fp, "%s\n", new_entry)) + { + log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E", + new_entry, csp->config->trustfile); } - return(0); + freez(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); + } + else + { + log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E", + csp->http->hostport, csp->config->trustfile); } + return 0; } } - return(1); + + return 1; } -#endif /* def FEATURE_COOKIE_JAR */ +#endif /* def FEATURE_TRUST */ /********************************************************************* * * 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. + * Description : Execute all text substitutions from all applying + * +filter actions on the text buffer that's been + * accumulated in csp->iob->buf. * * 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 - * + * or NULL if there were no hits or something went wrong + * *********************************************************************/ -char *pcrs_filter_response(struct client_state *csp) +static char *pcrs_filter_response(struct client_state *csp) { int hits=0; - size_t size; + size_t size, prev_size; - char *old = csp->iob->cur, *new = NULL; + char *old = NULL; + char *new = NULL; pcrs_job *job; struct file_list *fl; struct re_filterfile_spec *b; + struct list_entry *filtername; + + int i, found_filters = 0; - /* Sanity first */ + /* + * 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 + * Need to check the set of re_filterfiles... */ - if (csp->flags & CSP_FLAG_CHUNKED) + for (i = 0; i < MAX_AF_FILES; i++) { - log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first"); - if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size))) + fl = csp->rlist[i]; + if (NULL != fl) { - return(NULL); + if (NULL != fl->f) + { + found_filters = 1; + break; + } } - csp->iob->eod = csp->iob->cur + size; - csp->flags |= CSP_FLAG_MODIFIED; } - if ( ( NULL == (fl = csp->rlist) ) || ( NULL == (b = fl->f) ) ) + if (0 == found_filters) { - log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering."); + log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: " + "content filtering enabled, but no content filters available."); return(NULL); } - if ( NULL == b->joblist ) + size = (size_t)(csp->iob->eod - csp->iob->cur); + old = csp->iob->cur; + + for (i = 0; i < MAX_AF_FILES; i++) { - log_error(LOG_LEVEL_RE_FILTER, "Empty joblist. Nothing to do."); - return(NULL); - } + fl = csp->rlist[i]; + if ((NULL == fl) || (NULL == fl->f)) + { + /* + * Either there are no filter files + * left, or this filter file just + * contains no valid filters. + * + * Continue to be sure we don't miss + * valid filter files that are chained + * after empty or invalid ones. + */ + continue; + } + /* + * For all applying +filter actions, look if a filter by that + * name exists and if yes, execute it's pcrs_joblist on the + * buffer. + */ + for (b = fl->f; b; b = b->next) + { + if (b->type != FT_CONTENT_FILTER) + { + /* Skip header filters */ + continue; + } + + for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first; + filtername ; filtername = filtername->next) + { + if (strcmp(b->name, filtername->str) == 0) + { + int current_hits = 0; /* Number of hits caused by this filter */ + int job_number = 0; /* Which job we're currently executing */ + int job_hits = 0; /* How many hits the current job caused */ + pcrs_job *joblist = b->joblist; - log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) ...", - csp->http->hostport, csp->http->path, size); + if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b); - /* 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; - } + if (NULL == joblist) + { + log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name); + continue; + } - log_error(LOG_LEVEL_RE_FILTER, " produced %d hits (new size %d).", hits, size); + prev_size = size; + /* Apply all jobs from the joblist */ + for (job = joblist; NULL != job; job = job->next) + { + job_number++; + job_hits = pcrs_execute(job, old, size, &new, &size); - /* + if (job_hits >= 0) + { + /* + * That went well. Continue filtering + * and use the result of this job as + * input for the next one. + */ + current_hits += job_hits; + if (old != csp->iob->cur) + { + freez(old); + } + old = new; + } + else + { + /* + * This job caused an unexpected error. Inform the user + * and skip the rest of the jobs in this filter. We could + * continue with the next job, but usually the jobs + * depend on each other or are similar enough to + * fail for the same reason. + * + * At the moment our pcrs expects the error codes of pcre 3.4, + * but newer pcre versions can return additional error codes. + * As a result pcrs_strerror()'s error message might be + * "Unknown error ...", therefore we print the numerical value + * as well. + * + * XXX: Is this important enough for LOG_LEVEL_ERROR or + * should we use LOG_LEVEL_RE_FILTER instead? + */ + log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)", + b->name, job_number, pcrs_strerror(job_hits), job_hits); + break; + } + } + + if (b->dynamic) pcrs_free_joblist(joblist); + + log_error(LOG_LEVEL_RE_FILTER, + "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).", + csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size); + + hits += current_hits; + } + } + } + } + + /* * If there were no hits, destroy our copy and let * chat() use the original in csp->iob */ if (!hits) { - free(new); + freez(new); return(NULL); } @@ -1049,7 +1684,7 @@ char *pcrs_filter_response(struct client_state *csp) * * Function : gif_deanimate_response * - * Description : Deanimate the GIF image that has been accumulated in + * 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. * @@ -1058,57 +1693,127 @@ char *pcrs_filter_response(struct client_state *csp) * * Returns : a pointer to the (newly allocated) modified buffer. * or NULL in case something went wrong. - * + * *********************************************************************/ -char *gif_deanimate_response(struct client_state *csp) +static char *gif_deanimate_response(struct client_state *csp) { struct binbuffer *in, *out; char *p; - int size = csp->iob->eod - csp->iob->cur; + size_t size; + + size = (size_t)(csp->iob->eod - csp->iob->cur); + + 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)"); + freez(in); + buf_free(out); + return(NULL); + } + else + { + if ((int)size == out->offset) + { + log_error(LOG_LEVEL_DEANIMATE, "GIF not changed."); + } + 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; + freez(in); + freez(out); + return(p); + } + +} + + +/********************************************************************* + * + * Function : get_filter_function + * + * Description : Decides which content filter function has + * to be applied (if any). + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) + * + * Returns : The content filter function to run, or + * NULL if no content filter is active + * + *********************************************************************/ +filter_function_ptr get_filter_function(struct client_state *csp) +{ + filter_function_ptr filter_function = NULL; + + if ((csp->content_type & CT_TABOO) + && !(csp->action->flags & ACTION_FORCE_TEXT_MODE)) + { + return NULL; + } /* - * If the body has a "chunked" transfer-encoding, - * get rid of it first, adjusting size and iob->eod + * Are we enabling text mode by force? */ - if (csp->flags & CSP_FLAG_CHUNKED) + if (csp->action->flags & ACTION_FORCE_TEXT_MODE) { - log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first"); - if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size))) + /* + * Do we really have to? + */ + if (csp->content_type & CT_TEXT) { - return(NULL); + log_error(LOG_LEVEL_HEADER, "Text mode is already enabled."); + } + else + { + csp->content_type |= CT_TEXT; + log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!"); } - 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))) ) + if (!(csp->content_type & CT_DECLARED)) { - log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)"); - return NULL; + /* + * The server didn't bother to declare a MIME-Type. + * Assume it's text that can be filtered. + * + * This also regulary happens with 304 responses, + * therefore logging anything here would cause + * too much noise. + */ + csp->content_type |= CT_TEXT; } - in->buffer = csp->iob->cur; - in->size = size; - - if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4))) + /* + * Choose the applying filter function based on + * the content type and action settings. + */ + if ((csp->content_type & CT_TEXT) && + (csp->rlist != NULL) && + (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER]))) { - log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)"); - free(in); - buf_free(out); - return(NULL); + filter_function = pcrs_filter_response; } - else + else if ((csp->content_type & CT_GIF) && + (csp->action->flags & ACTION_DEANIMATE)) { - 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); - } + filter_function = gif_deanimate_response; + } + return filter_function; } @@ -1121,14 +1826,15 @@ char *gif_deanimate_response(struct client_state *csp) * * Parameters : * 1 : buffer = Pointer to the text buffer - * 2 : size = Number of bytes to be processed + * 2 : size = In: Number of bytes to be processed, + * Out: Number of bytes after de-chunking. + * (undefined in case of errors) + * + * Returns : JB_ERR_OK for success, + * JB_ERR_PARSE otherwise * - * 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) +static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size) { size_t newsize = 0; unsigned int chunksize = 0; @@ -1140,369 +1846,497 @@ int remove_chunked_transfer_coding(char *buffer, const size_t size) if (sscanf(buffer, "%x", &chunksize) != 1) { log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding"); - return(0); + return JB_ERR_PARSE; } - while (chunksize > 0) + while (chunksize > 0U) { if (NULL == (from_p = strstr(from_p, "\r\n"))) { log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding"); - return(0); + return JB_ERR_PARSE; + } + + if ((newsize += chunksize) >= *size) + { + log_error(LOG_LEVEL_ERROR, + "Chunk size %d exceeds buffer size %d in \"chunked\" transfer coding", + chunksize, *size); + return JB_ERR_PARSE; } - 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); + log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored."); + break; } } + + /* XXX: Should get its own loglevel. */ + log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize); + + *size = newsize; - /* 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); + return JB_ERR_OK; } /********************************************************************* * - * Function : url_actions + * Function : prepare_for_filtering * - * Description : Gets the actions for this URL. + * Description : If necessary, de-chunks and decompresses + * the content so it can get filterd. * * 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 : JB_ERR_OK for success, + * JB_ERR_PARSE otherwise * *********************************************************************/ -void url_actions(struct http_request *http, - struct client_state *csp) +static jb_err prepare_for_filtering(struct client_state *csp) { - struct file_list *fl; - struct url_actions *b; + jb_err err = JB_ERR_OK; - init_current_action(csp->action); - - if (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL)) + /* + * If the body has a "chunked" transfer-encoding, + * get rid of it, adjusting size and iob->eod + */ + if (csp->flags & CSP_FLAG_CHUNKED) { - return; + size_t size = (size_t)(csp->iob->eod - csp->iob->cur); + + log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first"); + err = remove_chunked_transfer_coding(csp->iob->cur, &size); + if (JB_ERR_OK == err) + { + csp->iob->eod = csp->iob->cur + size; + csp->flags |= CSP_FLAG_MODIFIED; + } + else + { + return JB_ERR_PARSE; + } } - apply_url_actions(csp->action, http, b); +#ifdef FEATURE_ZLIB + /* + * If the body has a supported transfer-encoding, + * decompress it, adjusting size and iob->eod. + */ + if (csp->content_type & (CT_GZIP|CT_DEFLATE)) + { + if (0 == csp->iob->eod - csp->iob->cur) + { + /* Nothing left after de-chunking. */ + return JB_ERR_OK; + } + + err = decompress_iob(csp); + + if (JB_ERR_OK == err) + { + csp->flags |= CSP_FLAG_MODIFIED; + csp->content_type &= ~CT_TABOO; + } + else + { + /* + * Unset CT_GZIP and CT_DEFLATE to remember not + * to modify the Content-Encoding header later. + */ + csp->content_type &= ~CT_GZIP; + csp->content_type &= ~CT_DEFLATE; + } + } +#endif + return err; } /********************************************************************* * - * Function : apply_url_actions + * Function : execute_content_filter * - * Description : Applies a list of URL actions. + * Description : Executes a given content filter. * * Parameters : - * 1 : action = Destination. - * 2 : http = Current URL - * 3 : b = list of URL actions to apply + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : content_filter = The filter function to execute * - * Returns : N/A + * Returns : Pointer to the modified buffer, or + * NULL if filtering failed or wasn't necessary. * *********************************************************************/ -void apply_url_actions(struct current_action_spec *action, - struct http_request *http, - struct url_actions *b) +char *execute_content_filter(struct client_state *csp, filter_function_ptr content_filter) { - struct url_spec url[1]; - - if (b == NULL) + if (0 == csp->iob->eod - csp->iob->cur) { - /* Should never happen */ - return; + /* + * No content (probably status code 301, 302 ...), + * no filtering necessary. + */ + return NULL; } - *url = dsplit(http->host); - - /* if splitting the domain fails, punt */ - if (url->dbuf == NULL) + if (JB_ERR_OK != prepare_for_filtering(csp)) { - return; + /* + * failed to de-chunk or decompress. + */ + return NULL; } - for (b = b->next; NULL != b; b = b->next) + if (0 == csp->iob->eod - csp->iob->cur) { - 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); - } - } - } + /* + * Clown alarm: chunked and/or compressed nothing delivered. + */ + return NULL; } - freez(url->dbuf); - freez(url->dvec); + return ((*content_filter)(csp)); } /********************************************************************* * - * Function : forward_url + * Function : get_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 - * 2 : csp = Current client state (buffers, headers, etc...) + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : http = http_request request for blocked URLs * - * Returns : Pointer to forwarding information. + * Returns : N/A * *********************************************************************/ -const struct forward_spec * forward_url(struct http_request *http, - struct client_state *csp) +void get_url_actions(struct client_state *csp, struct http_request *http) { - static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER }; - struct forward_spec *fwd = csp->config->forward; - struct url_spec url[1]; + struct file_list *fl; + struct url_actions *b; + int i; - if (fwd == NULL) + init_current_action(csp->action); + + for (i = 0; i < MAX_AF_FILES; i++) { - return(fwd_default); + if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL)) + { + return; + } + + apply_url_actions(csp->action, http, b); } - *url = dsplit(http->host); + return; +} + - /* if splitting the domain fails, punt */ - if (url->dbuf == NULL) +/********************************************************************* + * + * 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) +{ + if (b == NULL) { - return(fwd_default); + /* Should never happen */ + return; } - while (fwd != NULL) + for (b = b->next; NULL != b; b = b->next) { - if ((fwd->url->port == 0) || (fwd->url->port == http->port)) + if (url_match(b->url, http)) { - 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); - } - } + merge_current_action(action, b->action); } - fwd = fwd->next; } - - freez(url->dbuf); - freez(url->dvec); - return(fwd_default); - } /********************************************************************* * - * Function : dsplit + * Function : get_forward_override_settings + * + * Description : Returns forward settings as specified with the + * forward-override{} action. forward-override accepts + * forward lines similar to the one used in the + * configuration file, but without the URL pattern. + * + * For example: + * + * forward / . * - * 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. + * in the configuration file can be replaced with + * the action section: + * + * {+forward-override{forward .}} + * / * * Parameters : - * 1 : domain = a URL address + * 1 : csp = Current client state (buffers, headers, etc...) * - * Returns : url_spec structure populated with dbuf, dcnt and dvec. + * Returns : Pointer to forwarding structure in case of success. + * Invalid syntax is fatal. * *********************************************************************/ -struct url_spec dsplit(char *domain) +const static struct forward_spec *get_forward_override_settings(struct client_state *csp) { - struct url_spec ret[1]; - char *v[BUFFER_SIZE]; - int size; - char *p; - - memset(ret, '\0', sizeof(*ret)); + const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE]; + char forward_settings[BUFFER_SIZE]; + char *http_parent = NULL; + /* variable names were chosen for consistency reasons. */ + struct forward_spec *fwd = NULL; + int vec_count; + char *vec[3]; + + assert(csp->action->flags & ACTION_FORWARD_OVERRIDE); + /* Should be enforced by load_one_actions_file() */ + assert(strlen(forward_override_line) < sizeof(forward_settings) - 1); + + /* Create a copy ssplit can modify */ + strlcpy(forward_settings, forward_override_line, sizeof(forward_settings)); + + if (NULL != csp->fwd) + { + /* + * XXX: Currently necessary to prevent memory + * leaks when the show-url-info cgi page is visited. + */ + unload_forward_spec(csp->fwd); + } - if (domain[strlen(domain) - 1] == '.') + /* + * allocate a new forward node, valid only for + * the lifetime of this request. Save its location + * in csp as well, so sweep() can free it later on. + */ + fwd = csp->fwd = zalloc(sizeof(*fwd)); + if (NULL == fwd) { - ret->unanchored |= ANCHOR_RIGHT; + log_error(LOG_LEVEL_FATAL, + "can't allocate memory for forward-override{%s}", forward_override_line); + /* Never get here - LOG_LEVEL_FATAL causes program exit */ + return NULL; } - if (domain[0] == '.') + vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1); + if ((vec_count == 2) && !strcasecmp(vec[0], "forward")) { - ret->unanchored |= ANCHOR_LEFT; + fwd->type = SOCKS_NONE; + + /* Parse the parent HTTP proxy host:port */ + http_parent = vec[1]; + } + else if (vec_count == 3) + { + char *socks_proxy = NULL; - ret->dbuf = strdup(domain); + if (!strcasecmp(vec[0], "forward-socks4")) + { + fwd->type = SOCKS_4; + socks_proxy = vec[1]; + } + else if (!strcasecmp(vec[0], "forward-socks4a")) + { + fwd->type = SOCKS_4A; + socks_proxy = vec[1]; + } + else if (!strcasecmp(vec[0], "forward-socks5")) + { + fwd->type = SOCKS_5; + socks_proxy = vec[1]; + } - /* map to lower case */ - for (p = ret->dbuf; *p ; p++) *p = tolower(*p); + if (NULL != socks_proxy) + { + /* Parse the SOCKS proxy host[:port] */ + fwd->gateway_port = 1080; + parse_forwarder_address(socks_proxy, + &fwd->gateway_host, &fwd->gateway_port); - /* split the domain name into components */ - ret->dcnt = ssplit(ret->dbuf, ".", v, SZ(v), 1, 1); + http_parent = vec[2]; + } + } - if (ret->dcnt <= 0) + if (NULL == http_parent) { - memset(ret, '\0', sizeof(ret)); - return(*ret); + log_error(LOG_LEVEL_FATAL, + "Invalid forward-override syntax in: %s", forward_override_line); + /* Never get here - LOG_LEVEL_FATAL causes program exit */ } - /* save a copy of the pointers in dvec */ - size = ret->dcnt * sizeof(*ret->dvec); - - if ((ret->dvec = (char **)malloc(size))) + /* Parse http forwarding settings */ + if (strcmp(http_parent, ".") != 0) { - memcpy(ret->dvec, v, size); + fwd->forward_port = 8000; + parse_forwarder_address(http_parent, + &fwd->forward_host, &fwd->forward_port); } - return(*ret); + assert (NULL != fwd); + log_error(LOG_LEVEL_CONNECT, + "Overriding forwarding settings based on \'%s\'", forward_override_line); + + return fwd; } /********************************************************************* * - * Function : simple_domaincmp + * Function : forward_url * - * 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. + * Description : Should we forward this to another proxy? * * 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). + * 1 : csp = Current client state (buffers, headers, etc...) + * 2 : http = http_request request for current URL * - * Returns : 0 => domains are equivalent, else no match. + * Returns : Pointer to forwarding information. * *********************************************************************/ -static int simple_domaincmp(char **pv, char **fv, int len) +const struct forward_spec *forward_url(struct client_state *csp, + const struct http_request *http) { - int n; + static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER }; + struct forward_spec *fwd = csp->config->forward; + + if (csp->action->flags & ACTION_FORWARD_OVERRIDE) + { + return get_forward_override_settings(csp); + } + + if (fwd == NULL) + { + return fwd_default; + } - for (n = 0; n < len; n++) + while (fwd != NULL) { - if (simplematch(pv[n], fv[n])) + if (url_match(fwd->url, http)) { - return 1; + return fwd; } + fwd = fwd->next; } - return 0; - + return fwd_default; } /********************************************************************* * - * Function : domaincmp + * Function : direct_response * - * 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(). + * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE + * request and if so, return a HTTP 501 to the client. * - * Parameters : - * 1 : pattern = a domain that may contain a '*' as a wildcard. - * 2 : fqdn = domain name against which the patterns are compared. + * FIXME: I have a stupid name and I should handle the + * requests properly. Still, what we do here is rfc- + * compliant, whereas ignoring or forwarding are not. + * + * Parameters : + * 1 : csp = Current client state (buffers, headers, etc...) * - * Returns : 0 => domains are equivalent, else no match. + * Returns : http_response if , NULL if nonmatch or handler fail * *********************************************************************/ -int domaincmp(struct url_spec *pattern, struct url_spec *fqdn) +struct http_response *direct_response(struct client_state *csp) { - 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; + struct http_response *rsp; + struct list_entry *p; - 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) + if ((0 == strcmpic(csp->http->gpc, "trace")) + || (0 == strcmpic(csp->http->gpc, "options"))) { - /* Fully anchored, check length */ - if (flen != plen) + for (p = csp->headers->first; (p != NULL) ; p = p->next) { - 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)) + if (!strncmpic("Max-Forwards:", p->str, 13)) { - return 0; + unsigned int max_forwards; + + /* + * If it's a Max-Forwards value of zero, + * we have to intercept the request. + */ + if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0) + { + /* + * FIXME: We could handle at least TRACE here, + * but that would require a verbatim copy of + * the request which we don't have anymore + */ + log_error(LOG_LEVEL_HEADER, + "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.", + p->str); + + /* Get mem for response or fail*/ + if (NULL == (rsp = alloc_http_response())) + { + return cgi_error_memory(); + } + + if (NULL == (rsp->status = strdup("501 Not Implemented"))) + { + free_http_response(rsp); + return cgi_error_memory(); + } + + rsp->is_static = 1; + rsp->reason = RSP_REASON_UNSUPPORTED; + + return(finish_http_response(csp, rsp)); + } } - /* - * Doesn't match from start of fqdn - * Try skipping first part of fqdn - */ - fv++; } - return 1; } - + return NULL; } +/********************************************************************* + * + * Function : content_filters_enabled + * + * Description : Checks whether there are any content filters + * enabled for the current request. + * + * Parameters : + * 1 : action = Action spec to check. + * + * Returns : TRUE for yes, FALSE otherwise + * + *********************************************************************/ +int content_filters_enabled(const struct current_action_spec *action) +{ + return ((action->flags & ACTION_DEANIMATE) || + !list_is_empty(action->multi[ACTION_MULTI_FILTER])); +} + /* Local Variables: tab-width: 3