1 const char filters_rcs[] = "$Id: filters.c,v 1.163 2011/12/26 17:03:08 fabiankeil Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/filters.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
8 * Copyright : Written by and Copyright (C) 2001-2011 the
9 * Privoxy team. http://www.privoxy.org/
11 * Based on the Internet Junkbuster originally written
12 * by and Copyright (C) 1997 Anonymous Coders and
13 * Junkbusters Corporation. http://www.junkbusters.com
15 * This program is free software; you can redistribute it
16 * and/or modify it under the terms of the GNU General
17 * Public License as published by the Free Software
18 * Foundation; either version 2 of the License, or (at
19 * your option) any later version.
21 * This program is distributed in the hope that it will
22 * be useful, but WITHOUT ANY WARRANTY; without even the
23 * implied warranty of MERCHANTABILITY or FITNESS FOR A
24 * PARTICULAR PURPOSE. See the GNU General Public
25 * License for more details.
27 * The GNU General Public License should be included with
28 * this file. If not, you can view it at
29 * http://www.gnu.org/copyleft/gpl.html
30 * or write to the Free Software Foundation, Inc., 59
31 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
33 *********************************************************************/
39 #include <sys/types.h>
48 #endif /* ndef __OS2__ */
49 #include <netinet/in.h>
52 #endif /* ndef _WIN32 */
56 #endif /* def __OS2__ */
64 #include "jbsockets.h"
69 #include "deanimate.h"
77 const char filters_h_rcs[] = FILTERS_H_VERSION;
79 /* Fix a problem with Solaris. There should be no effect on other
81 * Solaris's isspace() is a macro which uses it's argument directly
82 * as an array index. Therefore we need to make sure that high-bit
83 * characters generate +ve values, and ideally we also want to make
84 * the argument match the declared parameter type of "int".
86 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
88 typedef char *(*filter_function_ptr)();
89 static filter_function_ptr get_filter_function(const struct client_state *csp);
90 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size);
91 static jb_err prepare_for_filtering(struct client_state *csp);
95 /*********************************************************************
97 * Function : sockaddr_storage_to_ip
99 * Description : Access internal structure of sockaddr_storage
102 * 1 : addr = socket address
103 * 2 : ip = IP address as array of octets in network order
104 * (it points into addr)
105 * 3 : len = length of IP address in octets
106 * 4 : port = port number in network order;
108 * Returns : 0 = no errror; -1 otherwise.
110 *********************************************************************/
111 static int sockaddr_storage_to_ip(const struct sockaddr_storage *addr,
112 uint8_t **ip, unsigned int *len,
120 switch (addr->ss_family)
130 &(((struct sockaddr_in *)addr)->sin_addr.s_addr);
134 *port = &((struct sockaddr_in *)addr)->sin_port;
145 *ip = ((struct sockaddr_in6 *)addr)->sin6_addr.s6_addr;
149 *port = &((struct sockaddr_in6 *)addr)->sin6_port;
154 /* Unsupported address family */
162 /*********************************************************************
164 * Function : match_sockaddr
166 * Description : Check whether address matches network (IP address and port)
169 * 1 : network = socket address of subnework
170 * 2 : netmask = network mask as socket address
171 * 3 : address = checked socket address against given network
173 * Returns : 0 = doesn't match; 1 = does match
175 *********************************************************************/
176 static int match_sockaddr(const struct sockaddr_storage *network,
177 const struct sockaddr_storage *netmask,
178 const struct sockaddr_storage *address)
180 uint8_t *network_addr, *netmask_addr, *address_addr;
181 unsigned int addr_len;
182 in_port_t *network_port, *netmask_port, *address_port;
185 if (network->ss_family != netmask->ss_family)
187 /* This should never happen */
188 log_error(LOG_LEVEL_ERROR,
189 "Internal error at %s:%llu: network and netmask differ in family",
194 sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
195 sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
196 sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
198 /* Check for family */
199 if ((network->ss_family == AF_INET) && (address->ss_family == AF_INET6)
200 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr))
202 /* Map AF_INET6 V4MAPPED address into AF_INET */
206 else if ((network->ss_family == AF_INET6) && (address->ss_family == AF_INET)
207 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr))
209 /* Map AF_INET6 V4MAPPED network into AF_INET */
214 else if (network->ss_family != address->ss_family)
219 /* XXX: Port check is signaled in netmask */
220 if (*netmask_port && *network_port != *address_port)
225 /* TODO: Optimize by checking by words insted of octets */
226 for (i = 0; (i < addr_len) && netmask_addr[i]; i++)
228 if ((network_addr[i] & netmask_addr[i]) !=
229 (address_addr[i] & netmask_addr[i]))
237 #endif /* def HAVE_RFC2553 */
240 /*********************************************************************
242 * Function : block_acl
244 * Description : Block this request?
245 * Decide yes or no based on ACL file.
248 * 1 : dst = The proxy or gateway address this is going to.
249 * Or NULL to check all possible targets.
250 * 2 : csp = Current client state (buffers, headers, etc...)
251 * Also includes the client IP address.
253 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
255 *********************************************************************/
256 int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
258 struct access_control_list *acl = csp->config->acl;
260 /* if not using an access control list, then permit the connection */
266 /* search the list */
271 match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
273 (csp->ip_addr_long & acl->src->mask) == acl->src->addr
279 /* Just want to check if they have any access */
280 if (acl->action == ACL_PERMIT)
292 * XXX: An undefined acl->dst is full of zeros and should be
293 * considered a wildcard address. sockaddr_storage_to_ip()
294 * fails on such destinations because of unknown sa_familly
295 * (glibc only?). However this test is not portable.
297 * So, we signal the acl->dst is wildcard in wildcard_dst.
300 match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
302 ((dst->addr & acl->dst->mask) == acl->dst->addr)
303 && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
307 if (acl->action == ACL_PERMIT)
325 /*********************************************************************
327 * Function : acl_addr
329 * Description : Called from `load_config' to parse an ACL address.
332 * 1 : aspec = String specifying ACL address.
333 * 2 : aca = struct access_control_addr to fill in.
335 * Returns : 0 => Ok, everything else is an error.
337 *********************************************************************/
338 int acl_addr(const char *aspec, struct access_control_addr *aca)
342 struct addrinfo hints, *result;
344 in_port_t *mask_port;
345 unsigned int addr_len;
348 #endif /* def HAVE_RFC2553 */
350 char *acl_spec = NULL;
353 /* XXX: Depend on ai_family */
361 * Use a temporary acl spec copy so we can log
362 * the unmodified original in case of parse errors.
364 acl_spec = strdup(aspec);
365 if (acl_spec == NULL)
367 /* XXX: This will be logged as parse error. */
371 if ((p = strchr(acl_spec, '/')) != NULL)
374 if (ijb_isdigit(*p) == 0)
379 masklength = atoi(p);
382 if ((masklength < 0) ||
394 if ((*acl_spec == '[') && (NULL != (p = strchr(acl_spec, ']'))))
397 memmove(acl_spec, acl_spec + 1, (size_t)(p - acl_spec));
406 p = strchr(acl_spec, ':');
410 memset(&hints, 0, sizeof(struct addrinfo));
411 hints.ai_family = AF_UNSPEC;
412 hints.ai_socktype = SOCK_STREAM;
414 i = getaddrinfo(acl_spec, ((p) ? ++p : NULL), &hints, &result);
418 log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s",
419 acl_spec, p, gai_strerror(i));
425 /* TODO: Allow multihomed hostnames */
426 memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
427 freeaddrinfo(result);
434 port = strtol(p, &endptr, 10);
436 if (port <= 0 || port > 65535 || *endptr != '\0')
443 aca->port = (unsigned long)port;
445 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
448 if (aca->addr == INADDR_NONE)
450 /* XXX: This will be logged as parse error. */
453 #endif /* def HAVE_RFC2553 */
455 /* build the netmask */
457 /* Clip masklength according to current family. */
458 if ((aca->addr.ss_family == AF_INET) && (masklength > 32))
463 aca->mask.ss_family = aca->addr.ss_family;
464 if (sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port))
471 /* ACL contains a port number, check ports in the future. */
476 * XXX: This could be optimized to operate on whole words instead
477 * of octets (128-bit CPU could do it in one iteration).
480 * Octets after prefix can be omitted because of
481 * previous initialization to zeros.
483 for (i = 0; (i < addr_len) && masklength; i++)
493 * XXX: This assumes MSB of octet is on the left side.
494 * This should be true for all architectures or solved
497 mask_data[i] = (uint8_t)~((1 << (8 - masklength)) - 1);
504 for (i=1; i <= masklength ; i++)
506 aca->mask |= (1U << (32 - i));
509 /* now mask off the host portion of the ip address
510 * (i.e. save on the network portion of the address).
512 aca->addr = aca->addr & aca->mask;
513 #endif /* def HAVE_RFC2553 */
518 #endif /* def FEATURE_ACL */
521 /*********************************************************************
523 * Function : connect_port_is_forbidden
525 * Description : Check to see if CONNECT requests to the destination
526 * port of this request are forbidden. The check is
527 * independend of the actual request method.
530 * 1 : csp = Current client state (buffers, headers, etc...)
532 * Returns : True if yes, false otherwise.
534 *********************************************************************/
535 int connect_port_is_forbidden(const struct client_state *csp)
537 return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
538 !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
543 /*********************************************************************
545 * Function : block_url
547 * Description : Called from `chat'. Check to see if we need to block this.
550 * 1 : csp = Current client state (buffers, headers, etc...)
552 * Returns : NULL => unblocked, else HTTP block response
554 *********************************************************************/
555 struct http_response *block_url(struct client_state *csp)
557 struct http_response *rsp;
558 const char *new_content_type = NULL;
561 * If it's not blocked, don't block it ;-)
563 if ((csp->action->flags & ACTION_BLOCK) == 0)
567 if (csp->action->flags & ACTION_REDIRECT)
569 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
572 * Else, prepare a response
574 if (NULL == (rsp = alloc_http_response()))
576 return cgi_error_memory();
580 * If it's an image-url, send back an image or redirect
581 * as specified by the relevant +image action
583 #ifdef FEATURE_IMAGE_BLOCKING
584 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
588 /* determine HOW images should be blocked */
589 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
591 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
593 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
596 /* and handle accordingly: */
597 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
599 rsp->status = strdup("403 Request blocked by Privoxy");
600 if (rsp->status == NULL)
602 free_http_response(rsp);
603 return cgi_error_memory();
605 rsp->body = bindup(image_pattern_data, image_pattern_length);
606 if (rsp->body == NULL)
608 free_http_response(rsp);
609 return cgi_error_memory();
611 rsp->content_length = image_pattern_length;
613 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
615 free_http_response(rsp);
616 return cgi_error_memory();
619 else if (0 == strcmpic(p, "blank"))
621 rsp->status = strdup("403 Request blocked by Privoxy");
622 if (rsp->status == NULL)
624 free_http_response(rsp);
625 return cgi_error_memory();
627 rsp->body = bindup(image_blank_data, image_blank_length);
628 if (rsp->body == NULL)
630 free_http_response(rsp);
631 return cgi_error_memory();
633 rsp->content_length = image_blank_length;
635 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
637 free_http_response(rsp);
638 return cgi_error_memory();
643 rsp->status = strdup("302 Local Redirect from Privoxy");
644 if (rsp->status == NULL)
646 free_http_response(rsp);
647 return cgi_error_memory();
650 if (enlist_unique_header(rsp->headers, "Location", p))
652 free_http_response(rsp);
653 return cgi_error_memory();
659 #endif /* def FEATURE_IMAGE_BLOCKING */
660 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
663 * Send empty document.
665 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
668 rsp->body = strdup(" ");
669 rsp->content_length = 1;
671 if (csp->config->feature_flags & RUNTIME_FEATURE_EMPTY_DOC_RETURNS_OK)
674 * Workaround for firefox bug 492459
675 * https://bugzilla.mozilla.org/show_bug.cgi?id=492459
676 * Return a 200 OK status for pages blocked with +handle-as-empty-document
677 * if the "handle-as-empty-doc-returns-ok" runtime config option is set.
679 rsp->status = strdup("200 Request blocked by Privoxy");
683 rsp->status = strdup("403 Request blocked by Privoxy");
686 if (rsp->status == NULL)
688 free_http_response(rsp);
689 return cgi_error_memory();
691 if (new_content_type != 0)
693 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
694 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
696 free_http_response(rsp);
697 return cgi_error_memory();
704 * Else, generate an HTML "blocked" message:
708 struct map * exports;
710 rsp->status = strdup("403 Request blocked by Privoxy");
711 if (rsp->status == NULL)
713 free_http_response(rsp);
714 return cgi_error_memory();
717 exports = default_exports(csp, NULL);
720 free_http_response(rsp);
721 return cgi_error_memory();
724 #ifdef FEATURE_FORCE_LOAD
725 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
727 * Export the force conditional block killer if
729 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
730 * - Privoxy is configured to enforce blocks, or
731 * - it's a CONNECT request and enforcing wouldn't work anyway.
733 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
734 || (0 == strcmpic(csp->http->gpc, "connect")))
735 #endif /* ndef FEATURE_FORCE_LOAD */
737 err = map_block_killer(exports, "force-support");
740 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
741 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
742 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
743 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
746 const char *block_reason;
747 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
749 block_reason = csp->action->string[ACTION_STRING_BLOCK];
753 assert(connect_port_is_forbidden(csp));
754 block_reason = "Forbidden CONNECT port.";
756 err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
761 free_http_response(rsp);
762 return cgi_error_memory();
765 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
768 free_http_response(rsp);
769 return cgi_error_memory();
772 rsp->crunch_reason = BLOCKED;
774 return finish_http_response(csp, rsp);
780 /*********************************************************************
782 * Function : trust_url FIXME: I should be called distrust_url
784 * Description : Calls is_untrusted_url to determine if the URL is trusted
785 * and if not, returns a HTTP 403 response with a reject message.
788 * 1 : csp = Current client state (buffers, headers, etc...)
790 * Returns : NULL => trusted, else http_response.
792 *********************************************************************/
793 struct http_response *trust_url(struct client_state *csp)
795 struct http_response *rsp;
796 struct map * exports;
797 char buf[BUFFER_SIZE];
799 struct url_spec **tl;
804 * Don't bother to work on trusted URLs
806 if (!is_untrusted_url(csp))
812 * Else, prepare a response:
814 if (NULL == (rsp = alloc_http_response()))
816 return cgi_error_memory();
819 rsp->status = strdup("403 Request blocked by Privoxy");
820 exports = default_exports(csp, NULL);
821 if (exports == NULL || rsp->status == NULL)
823 free_http_response(rsp);
824 return cgi_error_memory();
828 * Export the protocol, host, port, and referrer information
830 err = map(exports, "hostport", 1, csp->http->hostport, 1);
831 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
832 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
834 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
836 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
840 if (!err) err = map(exports, "referrer", 1, "none set", 1);
846 free_http_response(rsp);
847 return cgi_error_memory();
851 * Export the trust list
854 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
856 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
857 string_append(&p, buf);
859 err = map(exports, "trusted-referrers", 1, p, 0);
864 free_http_response(rsp);
865 return cgi_error_memory();
869 * Export the trust info, if available
871 if (csp->config->trust_info->first)
873 struct list_entry *l;
876 for (l = csp->config->trust_info->first; l ; l = l->next)
878 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
879 string_append(&p, buf);
881 err = map(exports, "trust-info", 1, p, 0);
885 err = map_block_killer(exports, "have-trust-info");
891 free_http_response(rsp);
892 return cgi_error_memory();
896 * Export the force conditional block killer if
898 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
899 * - Privoxy is configured to enforce blocks, or
900 * - it's a CONNECT request and enforcing wouldn't work anyway.
902 #ifdef FEATURE_FORCE_LOAD
903 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
904 || (0 == strcmpic(csp->http->gpc, "connect")))
906 err = map_block_killer(exports, "force-support");
910 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
912 #else /* ifndef FEATURE_FORCE_LOAD */
913 err = map_block_killer(exports, "force-support");
914 #endif /* ndef FEATURE_FORCE_LOAD */
919 free_http_response(rsp);
920 return cgi_error_memory();
926 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
929 free_http_response(rsp);
930 return cgi_error_memory();
932 rsp->crunch_reason = UNTRUSTED;
934 return finish_http_response(csp, rsp);
936 #endif /* def FEATURE_TRUST */
939 /*********************************************************************
941 * Function : compile_dynamic_pcrs_job_list
943 * Description : Compiles a dynamic pcrs job list (one with variables
944 * resolved at request time)
947 * 1 : csp = Current client state (buffers, headers, etc...)
948 * 2 : b = The filter list to compile
950 * Returns : NULL in case of errors, otherwise the
953 *********************************************************************/
954 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
956 struct list_entry *pattern;
957 pcrs_job *job_list = NULL;
958 pcrs_job *dummy = NULL;
959 pcrs_job *lastjob = NULL;
962 const struct pcrs_variable variables[] =
964 {"url", csp->http->url, 1},
965 {"path", csp->http->path, 1},
966 {"host", csp->http->host, 1},
967 {"origin", csp->ip_addr_str, 1},
971 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
973 assert(pattern->str != NULL);
975 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
978 log_error(LOG_LEVEL_ERROR,
979 "Compiling dynamic pcrs job '%s' for '%s' failed with error code %d: %s",
980 pattern->str, b->name, error, pcrs_strerror(error));
985 if (error == PCRS_WARN_TRUNCATION)
987 log_error(LOG_LEVEL_ERROR,
988 "At least one of the variables in \'%s\' had to "
989 "be truncated before compilation", pattern->str);
991 if (job_list == NULL)
997 lastjob->next = dummy;
1007 /*********************************************************************
1009 * Function : rewrite_url
1011 * Description : Rewrites a URL with a single pcrs command
1012 * and returns the result if it differs from the
1013 * original and isn't obviously invalid.
1016 * 1 : old_url = URL to rewrite.
1017 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1020 * Returns : NULL if the pcrs_command didn't change the url, or
1021 * the result of the modification.
1023 *********************************************************************/
1024 char *rewrite_url(char *old_url, const char *pcrs_command)
1026 char *new_url = NULL;
1030 assert(pcrs_command);
1032 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1036 log_error(LOG_LEVEL_REDIRECTS,
1037 "pcrs command \"%s\" didn't change \"%s\".",
1038 pcrs_command, old_url);
1043 log_error(LOG_LEVEL_REDIRECTS,
1044 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1045 pcrs_command, old_url, pcrs_strerror(hits));
1048 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1050 log_error(LOG_LEVEL_ERROR,
1051 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1052 "but the result doesn't look like a valid URL and will be ignored.",
1053 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1058 log_error(LOG_LEVEL_REDIRECTS,
1059 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1060 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1068 #ifdef FEATURE_FAST_REDIRECTS
1069 /*********************************************************************
1071 * Function : get_last_url
1073 * Description : Search for the last URL inside a string.
1074 * If the string already is a URL, it will
1075 * be the first URL found.
1078 * 1 : subject = the string to check
1079 * 2 : redirect_mode = +fast-redirect{} mode
1081 * Returns : NULL if no URL was found, or
1082 * the last URL found.
1084 *********************************************************************/
1085 char *get_last_url(char *subject, const char *redirect_mode)
1087 char *new_url = NULL;
1091 assert(redirect_mode);
1093 subject = strdup(subject);
1094 if (subject == NULL)
1096 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1100 if (0 == strcmpic(redirect_mode, "check-decoded-url") && strchr(subject, '%'))
1102 log_error(LOG_LEVEL_REDIRECTS,
1103 "Checking \"%s\" for encoded redirects.", subject);
1106 * Check each parameter in the URL separately.
1107 * Sectionize the URL at "?" and "&",
1108 * go backwards through the segments, URL-decode them
1109 * and look for a URL in the decoded result.
1110 * Stop the search after the first match.
1112 char *url_segment = NULL;
1114 * XXX: This estimate is guaranteed to be high enough as we
1115 * let ssplit() ignore empty fields, but also a bit wasteful.
1117 size_t max_segments = strlen(subject) / 2;
1118 char **url_segments = malloc(max_segments * sizeof(char *));
1121 if (NULL == url_segments)
1123 log_error(LOG_LEVEL_ERROR, "Out of memory while decoding URL: %s", new_url);
1128 segments = ssplit(subject, "?&", url_segments, max_segments, 1, 1);
1130 while (segments-- > 0)
1132 char *dtoken = url_decode(url_segments[segments]);
1135 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", url_segments[segments]);
1138 url_segment = strstr(dtoken, "http://");
1139 if (NULL == url_segment)
1141 url_segment = strstr(dtoken, "https://");
1143 if (NULL != url_segment)
1145 url_segment = strdup(url_segment);
1147 if (url_segment == NULL)
1149 log_error(LOG_LEVEL_ERROR,
1150 "Out of memory while searching for redirects.");
1158 freez(url_segments);
1160 if (url_segment == NULL)
1164 subject = url_segment;
1168 /* Look for a URL inside this one, without decoding anything. */
1169 log_error(LOG_LEVEL_REDIRECTS,
1170 "Checking \"%s\" for unencoded redirects.", subject);
1174 * Find the last URL encoded in the request
1177 while ((tmp = strstr(tmp, "http://")) != NULL)
1181 tmp = (new_url != NULL) ? new_url : subject;
1182 while ((tmp = strstr(tmp, "https://")) != NULL)
1187 if ((new_url != NULL)
1188 && ( (new_url != subject)
1189 || (0 == strncmpic(subject, "http://", 7))
1190 || (0 == strncmpic(subject, "https://", 8))
1194 * Return new URL if we found a redirect
1195 * or if the subject already was a URL.
1197 * The second case makes sure that we can
1198 * chain get_last_url after another redirection check
1199 * (like rewrite_url) without losing earlier redirects.
1201 new_url = strdup(new_url);
1210 #endif /* def FEATURE_FAST_REDIRECTS */
1213 /*********************************************************************
1215 * Function : redirect_url
1217 * Description : Checks if Privoxy should answer the request with
1218 * a HTTP redirect and generates the redirect if
1222 * 1 : csp = Current client state (buffers, headers, etc...)
1224 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1226 *********************************************************************/
1227 struct http_response *redirect_url(struct client_state *csp)
1229 struct http_response *rsp;
1230 #ifdef FEATURE_FAST_REDIRECTS
1232 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1233 * as compile-time option? The user can easily disable
1234 * it in his action file.
1236 char * redirect_mode;
1237 #endif /* def FEATURE_FAST_REDIRECTS */
1238 char *old_url = NULL;
1239 char *new_url = NULL;
1240 char *redirection_string;
1242 if ((csp->action->flags & ACTION_REDIRECT))
1244 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1247 * If the redirection string begins with 's',
1248 * assume it's a pcrs command, otherwise treat it as
1249 * properly formatted URL and use it for the redirection
1252 * According to RFC 2616 section 14.30 the URL
1253 * has to be absolute and if the user tries:
1254 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1255 * she would get undefined results anyway.
1259 if (*redirection_string == 's')
1261 old_url = csp->http->url;
1262 new_url = rewrite_url(old_url, redirection_string);
1266 log_error(LOG_LEVEL_REDIRECTS,
1267 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1268 redirection_string);
1269 new_url = strdup(redirection_string);
1273 #ifdef FEATURE_FAST_REDIRECTS
1274 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1276 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1279 * If it exists, use the previously rewritten URL as input
1280 * otherwise just use the old path.
1282 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1283 new_url = get_last_url(old_url, redirect_mode);
1288 * Disable redirect checkers, so that they
1289 * will be only run more than once if the user
1290 * also enables them through tags.
1292 * From a performance point of view
1293 * it doesn't matter, but the duplicated
1294 * log messages are annoying.
1296 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1297 #endif /* def FEATURE_FAST_REDIRECTS */
1298 csp->action->flags &= ~ACTION_REDIRECT;
1300 /* Did any redirect action trigger? */
1303 if (url_requires_percent_encoding(new_url))
1306 log_error(LOG_LEVEL_REDIRECTS, "Percent-encoding redirect URL: %N",
1307 strlen(new_url), new_url);
1308 encoded_url = percent_encode_url(new_url);
1310 if (encoded_url == NULL)
1312 return cgi_error_memory();
1314 new_url = encoded_url;
1315 assert(FALSE == url_requires_percent_encoding(new_url));
1318 if (0 == strcmpic(new_url, csp->http->url))
1320 log_error(LOG_LEVEL_ERROR,
1321 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1322 csp->http->url, new_url);
1327 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1329 if (NULL == (rsp = alloc_http_response()))
1332 return cgi_error_memory();
1335 if (enlist_unique_header(rsp->headers, "Location", new_url)
1336 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))))
1339 free_http_response(rsp);
1340 return cgi_error_memory();
1342 rsp->crunch_reason = REDIRECTED;
1345 return finish_http_response(csp, rsp);
1349 /* Only reached if no redirect is required */
1355 #ifdef FEATURE_IMAGE_BLOCKING
1356 /*********************************************************************
1358 * Function : is_imageurl
1360 * Description : Given a URL, decide whether it is an image or not,
1361 * using either the info from a previous +image action
1362 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1363 * is MSIE and not on a Mac, tell from the browser's accept
1367 * 1 : csp = Current client state (buffers, headers, etc...)
1369 * Returns : True (nonzero) if URL is an image, false (0)
1372 *********************************************************************/
1373 int is_imageurl(const struct client_state *csp)
1375 #ifdef FEATURE_IMAGE_DETECT_MSIE
1378 tmp = get_header_value(csp->headers, "User-Agent:");
1379 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1381 tmp = get_header_value(csp->headers, "Accept:");
1382 if (tmp && strstr(tmp, "image/gif"))
1384 /* Client will accept HTML. If this seems counterintuitive,
1394 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1396 return ((csp->action->flags & ACTION_IMAGE) != 0);
1399 #endif /* def FEATURE_IMAGE_BLOCKING */
1402 #ifdef FEATURE_TRUST
1403 /*********************************************************************
1405 * Function : is_untrusted_url
1407 * Description : Should we "distrust" this URL (and block it)?
1409 * Yes if it matches a line in the trustfile, or if the
1410 * referrer matches a line starting with "+" in the
1415 * 1 : csp = Current client state (buffers, headers, etc...)
1417 * Returns : 0 => trusted, 1 => untrusted
1419 *********************************************************************/
1420 int is_untrusted_url(const struct client_state *csp)
1422 struct file_list *fl;
1423 struct block_spec *b;
1424 struct url_spec **trusted_url;
1425 struct http_request rhttp[1];
1426 const char * referer;
1430 * If we don't have a trustlist, we trust everybody
1432 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1437 memset(rhttp, '\0', sizeof(*rhttp));
1440 * Do we trust the request URL itself?
1442 for (b = b->next; b ; b = b->next)
1444 if (url_match(b->url, csp->http))
1450 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1452 /* no referrer was supplied */
1458 * If not, do we maybe trust its referrer?
1460 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
1466 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1468 if (url_match(*trusted_url, rhttp))
1470 /* if the URL's referrer is from a trusted referrer, then
1471 * add the target spec to the trustfile as an unblocked
1472 * domain and return 0 (which means it's OK).
1477 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1481 char * new_entry = strdup("~");
1483 string_append(&new_entry, csp->http->hostport);
1485 path = csp->http->path;
1486 if ( (path[0] == '/')
1488 && ((path_end = strchr(path + 2, '/')) != NULL))
1490 /* since this path points into a user's home space
1491 * be sure to include this spec in the trustfile.
1493 long path_len = path_end - path; /* save offset */
1494 path = strdup(path); /* Copy string */
1497 path_end = path + path_len; /* regenerate ptr to new buffer */
1498 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1500 string_join(&new_entry, path);
1504 * Give a reason for generating this entry.
1506 string_append(&new_entry, " # Trusted referrer was: ");
1507 string_append(&new_entry, referer);
1509 if (new_entry != NULL)
1511 if (-1 == fprintf(fp, "%s\n", new_entry))
1513 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1514 new_entry, csp->config->trustfile);
1520 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1521 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1528 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1529 csp->http->hostport, csp->config->trustfile);
1537 #endif /* def FEATURE_TRUST */
1540 /*********************************************************************
1542 * Function : pcrs_filter_response
1544 * Description : Execute all text substitutions from all applying
1545 * +filter actions on the text buffer that's been
1546 * accumulated in csp->iob->buf.
1549 * 1 : csp = Current client state (buffers, headers, etc...)
1551 * Returns : a pointer to the (newly allocated) modified buffer.
1552 * or NULL if there were no hits or something went wrong
1554 *********************************************************************/
1555 static char *pcrs_filter_response(struct client_state *csp)
1559 size_t size, prev_size;
1565 struct file_list *fl;
1566 struct re_filterfile_spec *b;
1567 struct list_entry *filtername;
1572 if (csp->iob->cur >= csp->iob->eod)
1577 if (filters_available(csp) == FALSE)
1579 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1580 "content filtering enabled, but no content filters available.");
1584 size = (size_t)(csp->iob->eod - csp->iob->cur);
1585 old = csp->iob->cur;
1587 for (i = 0; i < MAX_AF_FILES; i++)
1590 if ((NULL == fl) || (NULL == fl->f))
1593 * Either there are no filter files
1594 * left, or this filter file just
1595 * contains no valid filters.
1597 * Continue to be sure we don't miss
1598 * valid filter files that are chained
1599 * after empty or invalid ones.
1604 * For all applying +filter actions, look if a filter by that
1605 * name exists and if yes, execute it's pcrs_joblist on the
1608 for (b = fl->f; b; b = b->next)
1610 if (b->type != FT_CONTENT_FILTER)
1612 /* Skip header filters */
1616 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1617 filtername ; filtername = filtername->next)
1619 if (strcmp(b->name, filtername->str) == 0)
1621 int current_hits = 0; /* Number of hits caused by this filter */
1622 int job_number = 0; /* Which job we're currently executing */
1623 int job_hits = 0; /* How many hits the current job caused */
1624 pcrs_job *joblist = b->joblist;
1626 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1628 if (NULL == joblist)
1630 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1635 /* Apply all jobs from the joblist */
1636 for (job = joblist; NULL != job; job = job->next)
1639 job_hits = pcrs_execute(job, old, size, &new, &size);
1644 * That went well. Continue filtering
1645 * and use the result of this job as
1646 * input for the next one.
1648 current_hits += job_hits;
1649 if (old != csp->iob->cur)
1658 * This job caused an unexpected error. Inform the user
1659 * and skip the rest of the jobs in this filter. We could
1660 * continue with the next job, but usually the jobs
1661 * depend on each other or are similar enough to
1662 * fail for the same reason.
1664 * At the moment our pcrs expects the error codes of pcre 3.4,
1665 * but newer pcre versions can return additional error codes.
1666 * As a result pcrs_strerror()'s error message might be
1667 * "Unknown error ...", therefore we print the numerical value
1670 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1671 * should we use LOG_LEVEL_RE_FILTER instead?
1673 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1674 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1679 if (b->dynamic) pcrs_free_joblist(joblist);
1681 log_error(LOG_LEVEL_RE_FILTER,
1682 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
1683 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1685 hits += current_hits;
1692 * If there were no hits, destroy our copy and let
1693 * chat() use the original in csp->iob
1701 csp->flags |= CSP_FLAG_MODIFIED;
1702 csp->content_length = size;
1710 /*********************************************************************
1712 * Function : gif_deanimate_response
1714 * Description : Deanimate the GIF image that has been accumulated in
1715 * csp->iob->buf, set csp->content_length to the modified
1716 * size and raise the CSP_FLAG_MODIFIED flag.
1719 * 1 : csp = Current client state (buffers, headers, etc...)
1721 * Returns : a pointer to the (newly allocated) modified buffer.
1722 * or NULL in case something went wrong.
1724 *********************************************************************/
1725 static char *gif_deanimate_response(struct client_state *csp)
1727 struct binbuffer *in, *out;
1731 size = (size_t)(csp->iob->eod - csp->iob->cur);
1733 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1734 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1736 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1740 in->buffer = csp->iob->cur;
1743 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1745 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1752 if ((int)size == out->offset)
1754 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1758 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1760 csp->content_length = out->offset;
1761 csp->flags |= CSP_FLAG_MODIFIED;
1771 /*********************************************************************
1773 * Function : get_filter_function
1775 * Description : Decides which content filter function has
1776 * to be applied (if any).
1779 * 1 : csp = Current client state (buffers, headers, etc...)
1781 * Returns : The content filter function to run, or
1782 * NULL if no content filter is active
1784 *********************************************************************/
1785 static filter_function_ptr get_filter_function(const struct client_state *csp)
1787 filter_function_ptr filter_function = NULL;
1790 * Choose the applying filter function based on
1791 * the content type and action settings.
1793 if ((csp->content_type & CT_TEXT) &&
1794 (csp->rlist != NULL) &&
1795 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
1797 filter_function = pcrs_filter_response;
1799 else if ((csp->content_type & CT_GIF) &&
1800 (csp->action->flags & ACTION_DEANIMATE))
1802 filter_function = gif_deanimate_response;
1805 return filter_function;
1809 /*********************************************************************
1811 * Function : remove_chunked_transfer_coding
1813 * Description : In-situ remove the "chunked" transfer coding as defined
1814 * in rfc2616 from a buffer.
1817 * 1 : buffer = Pointer to the text buffer
1818 * 2 : size = In: Number of bytes to be processed,
1819 * Out: Number of bytes after de-chunking.
1820 * (undefined in case of errors)
1822 * Returns : JB_ERR_OK for success,
1823 * JB_ERR_PARSE otherwise
1825 *********************************************************************/
1826 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
1829 unsigned int chunksize = 0;
1830 char *from_p, *to_p;
1833 from_p = to_p = buffer;
1835 if (sscanf(buffer, "%x", &chunksize) != 1)
1837 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1838 return JB_ERR_PARSE;
1841 while (chunksize > 0U)
1843 if (NULL == (from_p = strstr(from_p, "\r\n")))
1845 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1846 return JB_ERR_PARSE;
1849 if (chunksize >= *size - newsize)
1851 log_error(LOG_LEVEL_ERROR,
1852 "Chunk size %u exceeds buffered data left. "
1853 "Already digested %u of %u buffered bytes.",
1854 chunksize, (unsigned int)newsize, (unsigned int)*size);
1855 return JB_ERR_PARSE;
1857 newsize += chunksize;
1860 memmove(to_p, from_p, (size_t) chunksize);
1861 to_p = buffer + newsize;
1862 from_p += chunksize + 2;
1864 if (sscanf(from_p, "%x", &chunksize) != 1)
1866 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
1871 /* XXX: Should get its own loglevel. */
1872 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
1881 /*********************************************************************
1883 * Function : prepare_for_filtering
1885 * Description : If necessary, de-chunks and decompresses
1886 * the content so it can get filterd.
1889 * 1 : csp = Current client state (buffers, headers, etc...)
1891 * Returns : JB_ERR_OK for success,
1892 * JB_ERR_PARSE otherwise
1894 *********************************************************************/
1895 static jb_err prepare_for_filtering(struct client_state *csp)
1897 jb_err err = JB_ERR_OK;
1900 * If the body has a "chunked" transfer-encoding,
1901 * get rid of it, adjusting size and iob->eod
1903 if (csp->flags & CSP_FLAG_CHUNKED)
1905 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
1907 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1908 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
1909 if (JB_ERR_OK == err)
1911 csp->iob->eod = csp->iob->cur + size;
1912 csp->flags |= CSP_FLAG_MODIFIED;
1916 return JB_ERR_PARSE;
1922 * If the body has a supported transfer-encoding,
1923 * decompress it, adjusting size and iob->eod.
1925 if (csp->content_type & (CT_GZIP|CT_DEFLATE))
1927 if (0 == csp->iob->eod - csp->iob->cur)
1929 /* Nothing left after de-chunking. */
1933 err = decompress_iob(csp);
1935 if (JB_ERR_OK == err)
1937 csp->flags |= CSP_FLAG_MODIFIED;
1938 csp->content_type &= ~CT_TABOO;
1943 * Unset CT_GZIP and CT_DEFLATE to remember not
1944 * to modify the Content-Encoding header later.
1946 csp->content_type &= ~CT_GZIP;
1947 csp->content_type &= ~CT_DEFLATE;
1956 /*********************************************************************
1958 * Function : execute_content_filters
1960 * Description : Executes a given content filter.
1963 * 1 : csp = Current client state (buffers, headers, etc...)
1965 * Returns : Pointer to the modified buffer, or
1966 * NULL if filtering failed or wasn't necessary.
1968 *********************************************************************/
1969 char *execute_content_filters(struct client_state *csp)
1971 filter_function_ptr content_filter;
1973 assert(content_filters_enabled(csp->action));
1975 if (0 == csp->iob->eod - csp->iob->cur)
1978 * No content (probably status code 301, 302 ...),
1979 * no filtering necessary.
1984 if (JB_ERR_OK != prepare_for_filtering(csp))
1987 * failed to de-chunk or decompress.
1992 if (0 == csp->iob->eod - csp->iob->cur)
1995 * Clown alarm: chunked and/or compressed nothing delivered.
2000 content_filter = get_filter_function(csp);
2002 return ((*content_filter)(csp));
2006 /*********************************************************************
2008 * Function : get_url_actions
2010 * Description : Gets the actions for this URL.
2013 * 1 : csp = Current client state (buffers, headers, etc...)
2014 * 2 : http = http_request request for blocked URLs
2018 *********************************************************************/
2019 void get_url_actions(struct client_state *csp, struct http_request *http)
2021 struct file_list *fl;
2022 struct url_actions *b;
2025 init_current_action(csp->action);
2027 for (i = 0; i < MAX_AF_FILES; i++)
2029 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2034 apply_url_actions(csp->action, http, b);
2041 /*********************************************************************
2043 * Function : apply_url_actions
2045 * Description : Applies a list of URL actions.
2048 * 1 : action = Destination.
2049 * 2 : http = Current URL
2050 * 3 : b = list of URL actions to apply
2054 *********************************************************************/
2055 void apply_url_actions(struct current_action_spec *action,
2056 struct http_request *http,
2057 struct url_actions *b)
2061 /* Should never happen */
2065 for (b = b->next; NULL != b; b = b->next)
2067 if (url_match(b->url, http))
2069 merge_current_action(action, b->action);
2075 /*********************************************************************
2077 * Function : get_forward_override_settings
2079 * Description : Returns forward settings as specified with the
2080 * forward-override{} action. forward-override accepts
2081 * forward lines similar to the one used in the
2082 * configuration file, but without the URL pattern.
2088 * in the configuration file can be replaced with
2089 * the action section:
2091 * {+forward-override{forward .}}
2095 * 1 : csp = Current client state (buffers, headers, etc...)
2097 * Returns : Pointer to forwarding structure in case of success.
2098 * Invalid syntax is fatal.
2100 *********************************************************************/
2101 const static struct forward_spec *get_forward_override_settings(struct client_state *csp)
2103 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2104 char forward_settings[BUFFER_SIZE];
2105 char *http_parent = NULL;
2106 /* variable names were chosen for consistency reasons. */
2107 struct forward_spec *fwd = NULL;
2111 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2112 /* Should be enforced by load_one_actions_file() */
2113 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2115 /* Create a copy ssplit can modify */
2116 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2118 if (NULL != csp->fwd)
2121 * XXX: Currently necessary to prevent memory
2122 * leaks when the show-url-info cgi page is visited.
2124 unload_forward_spec(csp->fwd);
2128 * allocate a new forward node, valid only for
2129 * the lifetime of this request. Save its location
2130 * in csp as well, so sweep() can free it later on.
2132 fwd = csp->fwd = zalloc(sizeof(*fwd));
2135 log_error(LOG_LEVEL_FATAL,
2136 "can't allocate memory for forward-override{%s}", forward_override_line);
2137 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2141 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
2142 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2144 fwd->type = SOCKS_NONE;
2146 /* Parse the parent HTTP proxy host:port */
2147 http_parent = vec[1];
2150 else if (vec_count == 3)
2152 char *socks_proxy = NULL;
2154 if (!strcasecmp(vec[0], "forward-socks4"))
2156 fwd->type = SOCKS_4;
2157 socks_proxy = vec[1];
2159 else if (!strcasecmp(vec[0], "forward-socks4a"))
2161 fwd->type = SOCKS_4A;
2162 socks_proxy = vec[1];
2164 else if (!strcasecmp(vec[0], "forward-socks5"))
2166 fwd->type = SOCKS_5;
2167 socks_proxy = vec[1];
2170 if (NULL != socks_proxy)
2172 /* Parse the SOCKS proxy host[:port] */
2173 fwd->gateway_port = 1080;
2174 parse_forwarder_address(socks_proxy,
2175 &fwd->gateway_host, &fwd->gateway_port);
2177 http_parent = vec[2];
2181 if (NULL == http_parent)
2183 log_error(LOG_LEVEL_FATAL,
2184 "Invalid forward-override syntax in: %s", forward_override_line);
2185 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2188 /* Parse http forwarding settings */
2189 if (strcmp(http_parent, ".") != 0)
2191 fwd->forward_port = 8000;
2192 parse_forwarder_address(http_parent,
2193 &fwd->forward_host, &fwd->forward_port);
2196 assert (NULL != fwd);
2198 log_error(LOG_LEVEL_CONNECT,
2199 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2205 /*********************************************************************
2207 * Function : forward_url
2209 * Description : Should we forward this to another proxy?
2212 * 1 : csp = Current client state (buffers, headers, etc...)
2213 * 2 : http = http_request request for current URL
2215 * Returns : Pointer to forwarding information.
2217 *********************************************************************/
2218 const struct forward_spec *forward_url(struct client_state *csp,
2219 const struct http_request *http)
2221 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2222 struct forward_spec *fwd = csp->config->forward;
2224 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2226 return get_forward_override_settings(csp);
2236 if (url_match(fwd->url, http))
2247 /*********************************************************************
2249 * Function : direct_response
2251 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2252 * request and if so, return a HTTP 501 to the client.
2254 * FIXME: I have a stupid name and I should handle the
2255 * requests properly. Still, what we do here is rfc-
2256 * compliant, whereas ignoring or forwarding are not.
2259 * 1 : csp = Current client state (buffers, headers, etc...)
2261 * Returns : http_response if , NULL if nonmatch or handler fail
2263 *********************************************************************/
2264 struct http_response *direct_response(struct client_state *csp)
2266 struct http_response *rsp;
2267 struct list_entry *p;
2269 if ((0 == strcmpic(csp->http->gpc, "trace"))
2270 || (0 == strcmpic(csp->http->gpc, "options")))
2272 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2274 if (!strncmpic(p->str, "Max-Forwards:", 13))
2276 unsigned int max_forwards;
2279 * If it's a Max-Forwards value of zero,
2280 * we have to intercept the request.
2282 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2285 * FIXME: We could handle at least TRACE here,
2286 * but that would require a verbatim copy of
2287 * the request which we don't have anymore
2289 log_error(LOG_LEVEL_HEADER,
2290 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2293 /* Get mem for response or fail*/
2294 if (NULL == (rsp = alloc_http_response()))
2296 return cgi_error_memory();
2299 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2301 free_http_response(rsp);
2302 return cgi_error_memory();
2306 rsp->crunch_reason = UNSUPPORTED;
2308 return(finish_http_response(csp, rsp));
2317 /*********************************************************************
2319 * Function : content_requires_filtering
2321 * Description : Checks whether there are any content filters
2322 * enabled for the current request and if they
2323 * can actually be applied..
2326 * 1 : csp = Current client state (buffers, headers, etc...)
2328 * Returns : TRUE for yes, FALSE otherwise
2330 *********************************************************************/
2331 int content_requires_filtering(struct client_state *csp)
2333 if ((csp->content_type & CT_TABOO)
2334 && !(csp->action->flags & ACTION_FORCE_TEXT_MODE))
2340 * Are we enabling text mode by force?
2342 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2345 * Do we really have to?
2347 if (csp->content_type & CT_TEXT)
2349 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2353 csp->content_type |= CT_TEXT;
2354 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2358 if (!(csp->content_type & CT_DECLARED))
2361 * The server didn't bother to declare a MIME-Type.
2362 * Assume it's text that can be filtered.
2364 * This also regulary happens with 304 responses,
2365 * therefore logging anything here would cause
2368 csp->content_type |= CT_TEXT;
2372 * Choose the applying filter function based on
2373 * the content type and action settings.
2375 if ((csp->content_type & CT_TEXT) &&
2376 (csp->rlist != NULL) &&
2377 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2381 else if ((csp->content_type & CT_GIF) &&
2382 (csp->action->flags & ACTION_DEANIMATE))
2392 /*********************************************************************
2394 * Function : content_filters_enabled
2396 * Description : Checks whether there are any content filters
2397 * enabled for the current request.
2400 * 1 : action = Action spec to check.
2402 * Returns : TRUE for yes, FALSE otherwise
2404 *********************************************************************/
2405 int content_filters_enabled(const struct current_action_spec *action)
2407 return ((action->flags & ACTION_DEANIMATE) ||
2408 !list_is_empty(action->multi[ACTION_MULTI_FILTER]));
2412 /*********************************************************************
2414 * Function : filters_available
2416 * Description : Checks whether there are any filters available.
2419 * 1 : csp = Current client state (buffers, headers, etc...)
2421 * Returns : TRUE for yes, FALSE otherwise.
2423 *********************************************************************/
2424 int filters_available(const struct client_state *csp)
2427 for (i = 0; i < MAX_AF_FILES; i++)
2429 const struct file_list *fl = csp->rlist[i];
2430 if ((NULL != fl) && (NULL != fl->f))