1 const char filters_rcs[] = "$Id: filters.c,v 1.164 2011/12/31 14:47:44 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 assert(network->ss_family == netmask->ss_family);
189 log_error(LOG_LEVEL_FATAL, "Network and netmask differ in family.");
192 sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
193 sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
194 sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
196 /* Check for family */
197 if ((network->ss_family == AF_INET) && (address->ss_family == AF_INET6)
198 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr))
200 /* Map AF_INET6 V4MAPPED address into AF_INET */
204 else if ((network->ss_family == AF_INET6) && (address->ss_family == AF_INET)
205 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr))
207 /* Map AF_INET6 V4MAPPED network into AF_INET */
212 else if (network->ss_family != address->ss_family)
217 /* XXX: Port check is signaled in netmask */
218 if (*netmask_port && *network_port != *address_port)
223 /* TODO: Optimize by checking by words insted of octets */
224 for (i = 0; (i < addr_len) && netmask_addr[i]; i++)
226 if ((network_addr[i] & netmask_addr[i]) !=
227 (address_addr[i] & netmask_addr[i]))
235 #endif /* def HAVE_RFC2553 */
238 /*********************************************************************
240 * Function : block_acl
242 * Description : Block this request?
243 * Decide yes or no based on ACL file.
246 * 1 : dst = The proxy or gateway address this is going to.
247 * Or NULL to check all possible targets.
248 * 2 : csp = Current client state (buffers, headers, etc...)
249 * Also includes the client IP address.
251 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
253 *********************************************************************/
254 int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
256 struct access_control_list *acl = csp->config->acl;
258 /* if not using an access control list, then permit the connection */
264 /* search the list */
269 match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
271 (csp->ip_addr_long & acl->src->mask) == acl->src->addr
277 /* Just want to check if they have any access */
278 if (acl->action == ACL_PERMIT)
290 * XXX: An undefined acl->dst is full of zeros and should be
291 * considered a wildcard address. sockaddr_storage_to_ip()
292 * fails on such destinations because of unknown sa_familly
293 * (glibc only?). However this test is not portable.
295 * So, we signal the acl->dst is wildcard in wildcard_dst.
298 match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
300 ((dst->addr & acl->dst->mask) == acl->dst->addr)
301 && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
305 if (acl->action == ACL_PERMIT)
323 /*********************************************************************
325 * Function : acl_addr
327 * Description : Called from `load_config' to parse an ACL address.
330 * 1 : aspec = String specifying ACL address.
331 * 2 : aca = struct access_control_addr to fill in.
333 * Returns : 0 => Ok, everything else is an error.
335 *********************************************************************/
336 int acl_addr(const char *aspec, struct access_control_addr *aca)
340 struct addrinfo hints, *result;
342 in_port_t *mask_port;
343 unsigned int addr_len;
346 #endif /* def HAVE_RFC2553 */
348 char *acl_spec = NULL;
351 /* XXX: Depend on ai_family */
359 * Use a temporary acl spec copy so we can log
360 * the unmodified original in case of parse errors.
362 acl_spec = strdup(aspec);
363 if (acl_spec == NULL)
365 /* XXX: This will be logged as parse error. */
369 if ((p = strchr(acl_spec, '/')) != NULL)
372 if (ijb_isdigit(*p) == 0)
377 masklength = atoi(p);
380 if ((masklength < 0) ||
392 if ((*acl_spec == '[') && (NULL != (p = strchr(acl_spec, ']'))))
395 memmove(acl_spec, acl_spec + 1, (size_t)(p - acl_spec));
404 p = strchr(acl_spec, ':');
408 memset(&hints, 0, sizeof(struct addrinfo));
409 hints.ai_family = AF_UNSPEC;
410 hints.ai_socktype = SOCK_STREAM;
412 i = getaddrinfo(acl_spec, ((p) ? ++p : NULL), &hints, &result);
416 log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s",
417 acl_spec, p, gai_strerror(i));
423 /* TODO: Allow multihomed hostnames */
424 memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
425 freeaddrinfo(result);
432 port = strtol(p, &endptr, 10);
434 if (port <= 0 || port > 65535 || *endptr != '\0')
441 aca->port = (unsigned long)port;
443 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
446 if (aca->addr == INADDR_NONE)
448 /* XXX: This will be logged as parse error. */
451 #endif /* def HAVE_RFC2553 */
453 /* build the netmask */
455 /* Clip masklength according to current family. */
456 if ((aca->addr.ss_family == AF_INET) && (masklength > 32))
461 aca->mask.ss_family = aca->addr.ss_family;
462 if (sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port))
469 /* ACL contains a port number, check ports in the future. */
474 * XXX: This could be optimized to operate on whole words instead
475 * of octets (128-bit CPU could do it in one iteration).
478 * Octets after prefix can be omitted because of
479 * previous initialization to zeros.
481 for (i = 0; (i < addr_len) && masklength; i++)
491 * XXX: This assumes MSB of octet is on the left side.
492 * This should be true for all architectures or solved
495 mask_data[i] = (uint8_t)~((1 << (8 - masklength)) - 1);
502 for (i=1; i <= masklength ; i++)
504 aca->mask |= (1U << (32 - i));
507 /* now mask off the host portion of the ip address
508 * (i.e. save on the network portion of the address).
510 aca->addr = aca->addr & aca->mask;
511 #endif /* def HAVE_RFC2553 */
516 #endif /* def FEATURE_ACL */
519 /*********************************************************************
521 * Function : connect_port_is_forbidden
523 * Description : Check to see if CONNECT requests to the destination
524 * port of this request are forbidden. The check is
525 * independend of the actual request method.
528 * 1 : csp = Current client state (buffers, headers, etc...)
530 * Returns : True if yes, false otherwise.
532 *********************************************************************/
533 int connect_port_is_forbidden(const struct client_state *csp)
535 return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
536 !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
541 /*********************************************************************
543 * Function : block_url
545 * Description : Called from `chat'. Check to see if we need to block this.
548 * 1 : csp = Current client state (buffers, headers, etc...)
550 * Returns : NULL => unblocked, else HTTP block response
552 *********************************************************************/
553 struct http_response *block_url(struct client_state *csp)
555 struct http_response *rsp;
556 const char *new_content_type = NULL;
559 * If it's not blocked, don't block it ;-)
561 if ((csp->action->flags & ACTION_BLOCK) == 0)
565 if (csp->action->flags & ACTION_REDIRECT)
567 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
570 * Else, prepare a response
572 if (NULL == (rsp = alloc_http_response()))
574 return cgi_error_memory();
578 * If it's an image-url, send back an image or redirect
579 * as specified by the relevant +image action
581 #ifdef FEATURE_IMAGE_BLOCKING
582 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
586 /* determine HOW images should be blocked */
587 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
589 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
591 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
594 /* and handle accordingly: */
595 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
597 rsp->status = strdup("403 Request blocked by Privoxy");
598 if (rsp->status == NULL)
600 free_http_response(rsp);
601 return cgi_error_memory();
603 rsp->body = bindup(image_pattern_data, image_pattern_length);
604 if (rsp->body == NULL)
606 free_http_response(rsp);
607 return cgi_error_memory();
609 rsp->content_length = image_pattern_length;
611 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
613 free_http_response(rsp);
614 return cgi_error_memory();
617 else if (0 == strcmpic(p, "blank"))
619 rsp->status = strdup("403 Request blocked by Privoxy");
620 if (rsp->status == NULL)
622 free_http_response(rsp);
623 return cgi_error_memory();
625 rsp->body = bindup(image_blank_data, image_blank_length);
626 if (rsp->body == NULL)
628 free_http_response(rsp);
629 return cgi_error_memory();
631 rsp->content_length = image_blank_length;
633 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
635 free_http_response(rsp);
636 return cgi_error_memory();
641 rsp->status = strdup("302 Local Redirect from Privoxy");
642 if (rsp->status == NULL)
644 free_http_response(rsp);
645 return cgi_error_memory();
648 if (enlist_unique_header(rsp->headers, "Location", p))
650 free_http_response(rsp);
651 return cgi_error_memory();
657 #endif /* def FEATURE_IMAGE_BLOCKING */
658 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
661 * Send empty document.
663 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
666 rsp->body = strdup(" ");
667 rsp->content_length = 1;
669 if (csp->config->feature_flags & RUNTIME_FEATURE_EMPTY_DOC_RETURNS_OK)
672 * Workaround for firefox bug 492459
673 * https://bugzilla.mozilla.org/show_bug.cgi?id=492459
674 * Return a 200 OK status for pages blocked with +handle-as-empty-document
675 * if the "handle-as-empty-doc-returns-ok" runtime config option is set.
677 rsp->status = strdup("200 Request blocked by Privoxy");
681 rsp->status = strdup("403 Request blocked by Privoxy");
684 if (rsp->status == NULL)
686 free_http_response(rsp);
687 return cgi_error_memory();
689 if (new_content_type != 0)
691 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
692 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
694 free_http_response(rsp);
695 return cgi_error_memory();
702 * Else, generate an HTML "blocked" message:
706 struct map * exports;
708 rsp->status = strdup("403 Request blocked by Privoxy");
709 if (rsp->status == NULL)
711 free_http_response(rsp);
712 return cgi_error_memory();
715 exports = default_exports(csp, NULL);
718 free_http_response(rsp);
719 return cgi_error_memory();
722 #ifdef FEATURE_FORCE_LOAD
723 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
725 * Export the force conditional block killer if
727 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
728 * - Privoxy is configured to enforce blocks, or
729 * - it's a CONNECT request and enforcing wouldn't work anyway.
731 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
732 || (0 == strcmpic(csp->http->gpc, "connect")))
733 #endif /* ndef FEATURE_FORCE_LOAD */
735 err = map_block_killer(exports, "force-support");
738 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
739 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
740 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
741 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
744 const char *block_reason;
745 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
747 block_reason = csp->action->string[ACTION_STRING_BLOCK];
751 assert(connect_port_is_forbidden(csp));
752 block_reason = "Forbidden CONNECT port.";
754 err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
759 free_http_response(rsp);
760 return cgi_error_memory();
763 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
766 free_http_response(rsp);
767 return cgi_error_memory();
770 rsp->crunch_reason = BLOCKED;
772 return finish_http_response(csp, rsp);
778 /*********************************************************************
780 * Function : trust_url FIXME: I should be called distrust_url
782 * Description : Calls is_untrusted_url to determine if the URL is trusted
783 * and if not, returns a HTTP 403 response with a reject message.
786 * 1 : csp = Current client state (buffers, headers, etc...)
788 * Returns : NULL => trusted, else http_response.
790 *********************************************************************/
791 struct http_response *trust_url(struct client_state *csp)
793 struct http_response *rsp;
794 struct map * exports;
795 char buf[BUFFER_SIZE];
797 struct url_spec **tl;
802 * Don't bother to work on trusted URLs
804 if (!is_untrusted_url(csp))
810 * Else, prepare a response:
812 if (NULL == (rsp = alloc_http_response()))
814 return cgi_error_memory();
817 rsp->status = strdup("403 Request blocked by Privoxy");
818 exports = default_exports(csp, NULL);
819 if (exports == NULL || rsp->status == NULL)
821 free_http_response(rsp);
822 return cgi_error_memory();
826 * Export the protocol, host, port, and referrer information
828 err = map(exports, "hostport", 1, csp->http->hostport, 1);
829 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
830 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
832 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
834 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
838 if (!err) err = map(exports, "referrer", 1, "none set", 1);
844 free_http_response(rsp);
845 return cgi_error_memory();
849 * Export the trust list
852 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
854 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
855 string_append(&p, buf);
857 err = map(exports, "trusted-referrers", 1, p, 0);
862 free_http_response(rsp);
863 return cgi_error_memory();
867 * Export the trust info, if available
869 if (csp->config->trust_info->first)
871 struct list_entry *l;
874 for (l = csp->config->trust_info->first; l ; l = l->next)
876 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
877 string_append(&p, buf);
879 err = map(exports, "trust-info", 1, p, 0);
883 err = map_block_killer(exports, "have-trust-info");
889 free_http_response(rsp);
890 return cgi_error_memory();
894 * Export the force conditional block killer if
896 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
897 * - Privoxy is configured to enforce blocks, or
898 * - it's a CONNECT request and enforcing wouldn't work anyway.
900 #ifdef FEATURE_FORCE_LOAD
901 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
902 || (0 == strcmpic(csp->http->gpc, "connect")))
904 err = map_block_killer(exports, "force-support");
908 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
910 #else /* ifndef FEATURE_FORCE_LOAD */
911 err = map_block_killer(exports, "force-support");
912 #endif /* ndef FEATURE_FORCE_LOAD */
917 free_http_response(rsp);
918 return cgi_error_memory();
924 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
927 free_http_response(rsp);
928 return cgi_error_memory();
930 rsp->crunch_reason = UNTRUSTED;
932 return finish_http_response(csp, rsp);
934 #endif /* def FEATURE_TRUST */
937 /*********************************************************************
939 * Function : compile_dynamic_pcrs_job_list
941 * Description : Compiles a dynamic pcrs job list (one with variables
942 * resolved at request time)
945 * 1 : csp = Current client state (buffers, headers, etc...)
946 * 2 : b = The filter list to compile
948 * Returns : NULL in case of errors, otherwise the
951 *********************************************************************/
952 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
954 struct list_entry *pattern;
955 pcrs_job *job_list = NULL;
956 pcrs_job *dummy = NULL;
957 pcrs_job *lastjob = NULL;
960 const struct pcrs_variable variables[] =
962 {"url", csp->http->url, 1},
963 {"path", csp->http->path, 1},
964 {"host", csp->http->host, 1},
965 {"origin", csp->ip_addr_str, 1},
969 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
971 assert(pattern->str != NULL);
973 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
976 log_error(LOG_LEVEL_ERROR,
977 "Compiling dynamic pcrs job '%s' for '%s' failed with error code %d: %s",
978 pattern->str, b->name, error, pcrs_strerror(error));
983 if (error == PCRS_WARN_TRUNCATION)
985 log_error(LOG_LEVEL_ERROR,
986 "At least one of the variables in \'%s\' had to "
987 "be truncated before compilation", pattern->str);
989 if (job_list == NULL)
995 lastjob->next = dummy;
1005 /*********************************************************************
1007 * Function : rewrite_url
1009 * Description : Rewrites a URL with a single pcrs command
1010 * and returns the result if it differs from the
1011 * original and isn't obviously invalid.
1014 * 1 : old_url = URL to rewrite.
1015 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1018 * Returns : NULL if the pcrs_command didn't change the url, or
1019 * the result of the modification.
1021 *********************************************************************/
1022 char *rewrite_url(char *old_url, const char *pcrs_command)
1024 char *new_url = NULL;
1028 assert(pcrs_command);
1030 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1034 log_error(LOG_LEVEL_REDIRECTS,
1035 "pcrs command \"%s\" didn't change \"%s\".",
1036 pcrs_command, old_url);
1041 log_error(LOG_LEVEL_REDIRECTS,
1042 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1043 pcrs_command, old_url, pcrs_strerror(hits));
1046 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1048 log_error(LOG_LEVEL_ERROR,
1049 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1050 "but the result doesn't look like a valid URL and will be ignored.",
1051 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1056 log_error(LOG_LEVEL_REDIRECTS,
1057 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1058 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1066 #ifdef FEATURE_FAST_REDIRECTS
1067 /*********************************************************************
1069 * Function : get_last_url
1071 * Description : Search for the last URL inside a string.
1072 * If the string already is a URL, it will
1073 * be the first URL found.
1076 * 1 : subject = the string to check
1077 * 2 : redirect_mode = +fast-redirect{} mode
1079 * Returns : NULL if no URL was found, or
1080 * the last URL found.
1082 *********************************************************************/
1083 char *get_last_url(char *subject, const char *redirect_mode)
1085 char *new_url = NULL;
1089 assert(redirect_mode);
1091 subject = strdup(subject);
1092 if (subject == NULL)
1094 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1098 if (0 == strcmpic(redirect_mode, "check-decoded-url") && strchr(subject, '%'))
1100 log_error(LOG_LEVEL_REDIRECTS,
1101 "Checking \"%s\" for encoded redirects.", subject);
1104 * Check each parameter in the URL separately.
1105 * Sectionize the URL at "?" and "&",
1106 * go backwards through the segments, URL-decode them
1107 * and look for a URL in the decoded result.
1108 * Stop the search after the first match.
1110 char *url_segment = NULL;
1112 * XXX: This estimate is guaranteed to be high enough as we
1113 * let ssplit() ignore empty fields, but also a bit wasteful.
1115 size_t max_segments = strlen(subject) / 2;
1116 char **url_segments = malloc(max_segments * sizeof(char *));
1119 if (NULL == url_segments)
1121 log_error(LOG_LEVEL_ERROR, "Out of memory while decoding URL: %s", new_url);
1126 segments = ssplit(subject, "?&", url_segments, max_segments, 1, 1);
1128 while (segments-- > 0)
1130 char *dtoken = url_decode(url_segments[segments]);
1133 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", url_segments[segments]);
1136 url_segment = strstr(dtoken, "http://");
1137 if (NULL == url_segment)
1139 url_segment = strstr(dtoken, "https://");
1141 if (NULL != url_segment)
1143 url_segment = strdup(url_segment);
1145 if (url_segment == NULL)
1147 log_error(LOG_LEVEL_ERROR,
1148 "Out of memory while searching for redirects.");
1156 freez(url_segments);
1158 if (url_segment == NULL)
1162 subject = url_segment;
1166 /* Look for a URL inside this one, without decoding anything. */
1167 log_error(LOG_LEVEL_REDIRECTS,
1168 "Checking \"%s\" for unencoded redirects.", subject);
1172 * Find the last URL encoded in the request
1175 while ((tmp = strstr(tmp, "http://")) != NULL)
1179 tmp = (new_url != NULL) ? new_url : subject;
1180 while ((tmp = strstr(tmp, "https://")) != NULL)
1185 if ((new_url != NULL)
1186 && ( (new_url != subject)
1187 || (0 == strncmpic(subject, "http://", 7))
1188 || (0 == strncmpic(subject, "https://", 8))
1192 * Return new URL if we found a redirect
1193 * or if the subject already was a URL.
1195 * The second case makes sure that we can
1196 * chain get_last_url after another redirection check
1197 * (like rewrite_url) without losing earlier redirects.
1199 new_url = strdup(new_url);
1208 #endif /* def FEATURE_FAST_REDIRECTS */
1211 /*********************************************************************
1213 * Function : redirect_url
1215 * Description : Checks if Privoxy should answer the request with
1216 * a HTTP redirect and generates the redirect if
1220 * 1 : csp = Current client state (buffers, headers, etc...)
1222 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1224 *********************************************************************/
1225 struct http_response *redirect_url(struct client_state *csp)
1227 struct http_response *rsp;
1228 #ifdef FEATURE_FAST_REDIRECTS
1230 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1231 * as compile-time option? The user can easily disable
1232 * it in his action file.
1234 char * redirect_mode;
1235 #endif /* def FEATURE_FAST_REDIRECTS */
1236 char *old_url = NULL;
1237 char *new_url = NULL;
1238 char *redirection_string;
1240 if ((csp->action->flags & ACTION_REDIRECT))
1242 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1245 * If the redirection string begins with 's',
1246 * assume it's a pcrs command, otherwise treat it as
1247 * properly formatted URL and use it for the redirection
1250 * According to RFC 2616 section 14.30 the URL
1251 * has to be absolute and if the user tries:
1252 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1253 * she would get undefined results anyway.
1257 if (*redirection_string == 's')
1259 old_url = csp->http->url;
1260 new_url = rewrite_url(old_url, redirection_string);
1264 log_error(LOG_LEVEL_REDIRECTS,
1265 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1266 redirection_string);
1267 new_url = strdup(redirection_string);
1271 #ifdef FEATURE_FAST_REDIRECTS
1272 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1274 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1277 * If it exists, use the previously rewritten URL as input
1278 * otherwise just use the old path.
1280 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1281 new_url = get_last_url(old_url, redirect_mode);
1286 * Disable redirect checkers, so that they
1287 * will be only run more than once if the user
1288 * also enables them through tags.
1290 * From a performance point of view
1291 * it doesn't matter, but the duplicated
1292 * log messages are annoying.
1294 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1295 #endif /* def FEATURE_FAST_REDIRECTS */
1296 csp->action->flags &= ~ACTION_REDIRECT;
1298 /* Did any redirect action trigger? */
1301 if (url_requires_percent_encoding(new_url))
1304 log_error(LOG_LEVEL_REDIRECTS, "Percent-encoding redirect URL: %N",
1305 strlen(new_url), new_url);
1306 encoded_url = percent_encode_url(new_url);
1308 if (encoded_url == NULL)
1310 return cgi_error_memory();
1312 new_url = encoded_url;
1313 assert(FALSE == url_requires_percent_encoding(new_url));
1316 if (0 == strcmpic(new_url, csp->http->url))
1318 log_error(LOG_LEVEL_ERROR,
1319 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1320 csp->http->url, new_url);
1325 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1327 if (NULL == (rsp = alloc_http_response()))
1330 return cgi_error_memory();
1333 if (enlist_unique_header(rsp->headers, "Location", new_url)
1334 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))))
1337 free_http_response(rsp);
1338 return cgi_error_memory();
1340 rsp->crunch_reason = REDIRECTED;
1343 return finish_http_response(csp, rsp);
1347 /* Only reached if no redirect is required */
1353 #ifdef FEATURE_IMAGE_BLOCKING
1354 /*********************************************************************
1356 * Function : is_imageurl
1358 * Description : Given a URL, decide whether it is an image or not,
1359 * using either the info from a previous +image action
1360 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1361 * is MSIE and not on a Mac, tell from the browser's accept
1365 * 1 : csp = Current client state (buffers, headers, etc...)
1367 * Returns : True (nonzero) if URL is an image, false (0)
1370 *********************************************************************/
1371 int is_imageurl(const struct client_state *csp)
1373 #ifdef FEATURE_IMAGE_DETECT_MSIE
1376 tmp = get_header_value(csp->headers, "User-Agent:");
1377 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1379 tmp = get_header_value(csp->headers, "Accept:");
1380 if (tmp && strstr(tmp, "image/gif"))
1382 /* Client will accept HTML. If this seems counterintuitive,
1392 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1394 return ((csp->action->flags & ACTION_IMAGE) != 0);
1397 #endif /* def FEATURE_IMAGE_BLOCKING */
1400 #ifdef FEATURE_TRUST
1401 /*********************************************************************
1403 * Function : is_untrusted_url
1405 * Description : Should we "distrust" this URL (and block it)?
1407 * Yes if it matches a line in the trustfile, or if the
1408 * referrer matches a line starting with "+" in the
1413 * 1 : csp = Current client state (buffers, headers, etc...)
1415 * Returns : 0 => trusted, 1 => untrusted
1417 *********************************************************************/
1418 int is_untrusted_url(const struct client_state *csp)
1420 struct file_list *fl;
1421 struct block_spec *b;
1422 struct url_spec **trusted_url;
1423 struct http_request rhttp[1];
1424 const char * referer;
1428 * If we don't have a trustlist, we trust everybody
1430 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1435 memset(rhttp, '\0', sizeof(*rhttp));
1438 * Do we trust the request URL itself?
1440 for (b = b->next; b ; b = b->next)
1442 if (url_match(b->url, csp->http))
1448 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1450 /* no referrer was supplied */
1456 * If not, do we maybe trust its referrer?
1458 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
1464 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1466 if (url_match(*trusted_url, rhttp))
1468 /* if the URL's referrer is from a trusted referrer, then
1469 * add the target spec to the trustfile as an unblocked
1470 * domain and return 0 (which means it's OK).
1475 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1479 char * new_entry = strdup("~");
1481 string_append(&new_entry, csp->http->hostport);
1483 path = csp->http->path;
1484 if ( (path[0] == '/')
1486 && ((path_end = strchr(path + 2, '/')) != NULL))
1488 /* since this path points into a user's home space
1489 * be sure to include this spec in the trustfile.
1491 long path_len = path_end - path; /* save offset */
1492 path = strdup(path); /* Copy string */
1495 path_end = path + path_len; /* regenerate ptr to new buffer */
1496 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1498 string_join(&new_entry, path);
1502 * Give a reason for generating this entry.
1504 string_append(&new_entry, " # Trusted referrer was: ");
1505 string_append(&new_entry, referer);
1507 if (new_entry != NULL)
1509 if (-1 == fprintf(fp, "%s\n", new_entry))
1511 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1512 new_entry, csp->config->trustfile);
1518 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1519 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1526 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1527 csp->http->hostport, csp->config->trustfile);
1535 #endif /* def FEATURE_TRUST */
1538 /*********************************************************************
1540 * Function : pcrs_filter_response
1542 * Description : Execute all text substitutions from all applying
1543 * +filter actions on the text buffer that's been
1544 * accumulated in csp->iob->buf.
1547 * 1 : csp = Current client state (buffers, headers, etc...)
1549 * Returns : a pointer to the (newly allocated) modified buffer.
1550 * or NULL if there were no hits or something went wrong
1552 *********************************************************************/
1553 static char *pcrs_filter_response(struct client_state *csp)
1557 size_t size, prev_size;
1563 struct file_list *fl;
1564 struct re_filterfile_spec *b;
1565 struct list_entry *filtername;
1570 if (csp->iob->cur >= csp->iob->eod)
1575 if (filters_available(csp) == FALSE)
1577 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1578 "content filtering enabled, but no content filters available.");
1582 size = (size_t)(csp->iob->eod - csp->iob->cur);
1583 old = csp->iob->cur;
1585 for (i = 0; i < MAX_AF_FILES; i++)
1588 if ((NULL == fl) || (NULL == fl->f))
1591 * Either there are no filter files
1592 * left, or this filter file just
1593 * contains no valid filters.
1595 * Continue to be sure we don't miss
1596 * valid filter files that are chained
1597 * after empty or invalid ones.
1602 * For all applying +filter actions, look if a filter by that
1603 * name exists and if yes, execute it's pcrs_joblist on the
1606 for (b = fl->f; b; b = b->next)
1608 if (b->type != FT_CONTENT_FILTER)
1610 /* Skip header filters */
1614 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1615 filtername ; filtername = filtername->next)
1617 if (strcmp(b->name, filtername->str) == 0)
1619 int current_hits = 0; /* Number of hits caused by this filter */
1620 int job_number = 0; /* Which job we're currently executing */
1621 int job_hits = 0; /* How many hits the current job caused */
1622 pcrs_job *joblist = b->joblist;
1624 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1626 if (NULL == joblist)
1628 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1633 /* Apply all jobs from the joblist */
1634 for (job = joblist; NULL != job; job = job->next)
1637 job_hits = pcrs_execute(job, old, size, &new, &size);
1642 * That went well. Continue filtering
1643 * and use the result of this job as
1644 * input for the next one.
1646 current_hits += job_hits;
1647 if (old != csp->iob->cur)
1656 * This job caused an unexpected error. Inform the user
1657 * and skip the rest of the jobs in this filter. We could
1658 * continue with the next job, but usually the jobs
1659 * depend on each other or are similar enough to
1660 * fail for the same reason.
1662 * At the moment our pcrs expects the error codes of pcre 3.4,
1663 * but newer pcre versions can return additional error codes.
1664 * As a result pcrs_strerror()'s error message might be
1665 * "Unknown error ...", therefore we print the numerical value
1668 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1669 * should we use LOG_LEVEL_RE_FILTER instead?
1671 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1672 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1677 if (b->dynamic) pcrs_free_joblist(joblist);
1679 log_error(LOG_LEVEL_RE_FILTER,
1680 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
1681 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1683 hits += current_hits;
1690 * If there were no hits, destroy our copy and let
1691 * chat() use the original in csp->iob
1699 csp->flags |= CSP_FLAG_MODIFIED;
1700 csp->content_length = size;
1708 /*********************************************************************
1710 * Function : gif_deanimate_response
1712 * Description : Deanimate the GIF image that has been accumulated in
1713 * csp->iob->buf, set csp->content_length to the modified
1714 * size and raise the CSP_FLAG_MODIFIED flag.
1717 * 1 : csp = Current client state (buffers, headers, etc...)
1719 * Returns : a pointer to the (newly allocated) modified buffer.
1720 * or NULL in case something went wrong.
1722 *********************************************************************/
1723 static char *gif_deanimate_response(struct client_state *csp)
1725 struct binbuffer *in, *out;
1729 size = (size_t)(csp->iob->eod - csp->iob->cur);
1731 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1732 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1734 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1738 in->buffer = csp->iob->cur;
1741 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1743 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1750 if ((int)size == out->offset)
1752 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1756 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1758 csp->content_length = out->offset;
1759 csp->flags |= CSP_FLAG_MODIFIED;
1769 /*********************************************************************
1771 * Function : get_filter_function
1773 * Description : Decides which content filter function has
1774 * to be applied (if any).
1777 * 1 : csp = Current client state (buffers, headers, etc...)
1779 * Returns : The content filter function to run, or
1780 * NULL if no content filter is active
1782 *********************************************************************/
1783 static filter_function_ptr get_filter_function(const struct client_state *csp)
1785 filter_function_ptr filter_function = NULL;
1788 * Choose the applying filter function based on
1789 * the content type and action settings.
1791 if ((csp->content_type & CT_TEXT) &&
1792 (csp->rlist != NULL) &&
1793 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
1795 filter_function = pcrs_filter_response;
1797 else if ((csp->content_type & CT_GIF) &&
1798 (csp->action->flags & ACTION_DEANIMATE))
1800 filter_function = gif_deanimate_response;
1803 return filter_function;
1807 /*********************************************************************
1809 * Function : remove_chunked_transfer_coding
1811 * Description : In-situ remove the "chunked" transfer coding as defined
1812 * in rfc2616 from a buffer.
1815 * 1 : buffer = Pointer to the text buffer
1816 * 2 : size = In: Number of bytes to be processed,
1817 * Out: Number of bytes after de-chunking.
1818 * (undefined in case of errors)
1820 * Returns : JB_ERR_OK for success,
1821 * JB_ERR_PARSE otherwise
1823 *********************************************************************/
1824 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
1827 unsigned int chunksize = 0;
1828 char *from_p, *to_p;
1831 from_p = to_p = buffer;
1833 if (sscanf(buffer, "%x", &chunksize) != 1)
1835 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1836 return JB_ERR_PARSE;
1839 while (chunksize > 0U)
1841 if (NULL == (from_p = strstr(from_p, "\r\n")))
1843 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1844 return JB_ERR_PARSE;
1847 if (chunksize >= *size - newsize)
1849 log_error(LOG_LEVEL_ERROR,
1850 "Chunk size %u exceeds buffered data left. "
1851 "Already digested %u of %u buffered bytes.",
1852 chunksize, (unsigned int)newsize, (unsigned int)*size);
1853 return JB_ERR_PARSE;
1855 newsize += chunksize;
1858 memmove(to_p, from_p, (size_t) chunksize);
1859 to_p = buffer + newsize;
1860 from_p += chunksize + 2;
1862 if (sscanf(from_p, "%x", &chunksize) != 1)
1864 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
1869 /* XXX: Should get its own loglevel. */
1870 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
1879 /*********************************************************************
1881 * Function : prepare_for_filtering
1883 * Description : If necessary, de-chunks and decompresses
1884 * the content so it can get filterd.
1887 * 1 : csp = Current client state (buffers, headers, etc...)
1889 * Returns : JB_ERR_OK for success,
1890 * JB_ERR_PARSE otherwise
1892 *********************************************************************/
1893 static jb_err prepare_for_filtering(struct client_state *csp)
1895 jb_err err = JB_ERR_OK;
1898 * If the body has a "chunked" transfer-encoding,
1899 * get rid of it, adjusting size and iob->eod
1901 if (csp->flags & CSP_FLAG_CHUNKED)
1903 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
1905 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1906 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
1907 if (JB_ERR_OK == err)
1909 csp->iob->eod = csp->iob->cur + size;
1910 csp->flags |= CSP_FLAG_MODIFIED;
1914 return JB_ERR_PARSE;
1920 * If the body has a supported transfer-encoding,
1921 * decompress it, adjusting size and iob->eod.
1923 if (csp->content_type & (CT_GZIP|CT_DEFLATE))
1925 if (0 == csp->iob->eod - csp->iob->cur)
1927 /* Nothing left after de-chunking. */
1931 err = decompress_iob(csp);
1933 if (JB_ERR_OK == err)
1935 csp->flags |= CSP_FLAG_MODIFIED;
1936 csp->content_type &= ~CT_TABOO;
1941 * Unset CT_GZIP and CT_DEFLATE to remember not
1942 * to modify the Content-Encoding header later.
1944 csp->content_type &= ~CT_GZIP;
1945 csp->content_type &= ~CT_DEFLATE;
1954 /*********************************************************************
1956 * Function : execute_content_filters
1958 * Description : Executes a given content filter.
1961 * 1 : csp = Current client state (buffers, headers, etc...)
1963 * Returns : Pointer to the modified buffer, or
1964 * NULL if filtering failed or wasn't necessary.
1966 *********************************************************************/
1967 char *execute_content_filters(struct client_state *csp)
1969 filter_function_ptr content_filter;
1971 assert(content_filters_enabled(csp->action));
1973 if (0 == csp->iob->eod - csp->iob->cur)
1976 * No content (probably status code 301, 302 ...),
1977 * no filtering necessary.
1982 if (JB_ERR_OK != prepare_for_filtering(csp))
1985 * failed to de-chunk or decompress.
1990 if (0 == csp->iob->eod - csp->iob->cur)
1993 * Clown alarm: chunked and/or compressed nothing delivered.
1998 content_filter = get_filter_function(csp);
2000 return ((*content_filter)(csp));
2004 /*********************************************************************
2006 * Function : get_url_actions
2008 * Description : Gets the actions for this URL.
2011 * 1 : csp = Current client state (buffers, headers, etc...)
2012 * 2 : http = http_request request for blocked URLs
2016 *********************************************************************/
2017 void get_url_actions(struct client_state *csp, struct http_request *http)
2019 struct file_list *fl;
2020 struct url_actions *b;
2023 init_current_action(csp->action);
2025 for (i = 0; i < MAX_AF_FILES; i++)
2027 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2032 apply_url_actions(csp->action, http, b);
2039 /*********************************************************************
2041 * Function : apply_url_actions
2043 * Description : Applies a list of URL actions.
2046 * 1 : action = Destination.
2047 * 2 : http = Current URL
2048 * 3 : b = list of URL actions to apply
2052 *********************************************************************/
2053 void apply_url_actions(struct current_action_spec *action,
2054 struct http_request *http,
2055 struct url_actions *b)
2059 /* Should never happen */
2063 for (b = b->next; NULL != b; b = b->next)
2065 if (url_match(b->url, http))
2067 merge_current_action(action, b->action);
2073 /*********************************************************************
2075 * Function : get_forward_override_settings
2077 * Description : Returns forward settings as specified with the
2078 * forward-override{} action. forward-override accepts
2079 * forward lines similar to the one used in the
2080 * configuration file, but without the URL pattern.
2086 * in the configuration file can be replaced with
2087 * the action section:
2089 * {+forward-override{forward .}}
2093 * 1 : csp = Current client state (buffers, headers, etc...)
2095 * Returns : Pointer to forwarding structure in case of success.
2096 * Invalid syntax is fatal.
2098 *********************************************************************/
2099 const static struct forward_spec *get_forward_override_settings(struct client_state *csp)
2101 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2102 char forward_settings[BUFFER_SIZE];
2103 char *http_parent = NULL;
2104 /* variable names were chosen for consistency reasons. */
2105 struct forward_spec *fwd = NULL;
2109 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2110 /* Should be enforced by load_one_actions_file() */
2111 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2113 /* Create a copy ssplit can modify */
2114 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2116 if (NULL != csp->fwd)
2119 * XXX: Currently necessary to prevent memory
2120 * leaks when the show-url-info cgi page is visited.
2122 unload_forward_spec(csp->fwd);
2126 * allocate a new forward node, valid only for
2127 * the lifetime of this request. Save its location
2128 * in csp as well, so sweep() can free it later on.
2130 fwd = csp->fwd = zalloc(sizeof(*fwd));
2133 log_error(LOG_LEVEL_FATAL,
2134 "can't allocate memory for forward-override{%s}", forward_override_line);
2135 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2139 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
2140 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2142 fwd->type = SOCKS_NONE;
2144 /* Parse the parent HTTP proxy host:port */
2145 http_parent = vec[1];
2148 else if (vec_count == 3)
2150 char *socks_proxy = NULL;
2152 if (!strcasecmp(vec[0], "forward-socks4"))
2154 fwd->type = SOCKS_4;
2155 socks_proxy = vec[1];
2157 else if (!strcasecmp(vec[0], "forward-socks4a"))
2159 fwd->type = SOCKS_4A;
2160 socks_proxy = vec[1];
2162 else if (!strcasecmp(vec[0], "forward-socks5"))
2164 fwd->type = SOCKS_5;
2165 socks_proxy = vec[1];
2168 if (NULL != socks_proxy)
2170 /* Parse the SOCKS proxy host[:port] */
2171 fwd->gateway_port = 1080;
2172 parse_forwarder_address(socks_proxy,
2173 &fwd->gateway_host, &fwd->gateway_port);
2175 http_parent = vec[2];
2179 if (NULL == http_parent)
2181 log_error(LOG_LEVEL_FATAL,
2182 "Invalid forward-override syntax in: %s", forward_override_line);
2183 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2186 /* Parse http forwarding settings */
2187 if (strcmp(http_parent, ".") != 0)
2189 fwd->forward_port = 8000;
2190 parse_forwarder_address(http_parent,
2191 &fwd->forward_host, &fwd->forward_port);
2194 assert (NULL != fwd);
2196 log_error(LOG_LEVEL_CONNECT,
2197 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2203 /*********************************************************************
2205 * Function : forward_url
2207 * Description : Should we forward this to another proxy?
2210 * 1 : csp = Current client state (buffers, headers, etc...)
2211 * 2 : http = http_request request for current URL
2213 * Returns : Pointer to forwarding information.
2215 *********************************************************************/
2216 const struct forward_spec *forward_url(struct client_state *csp,
2217 const struct http_request *http)
2219 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2220 struct forward_spec *fwd = csp->config->forward;
2222 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2224 return get_forward_override_settings(csp);
2234 if (url_match(fwd->url, http))
2245 /*********************************************************************
2247 * Function : direct_response
2249 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2250 * request and if so, return a HTTP 501 to the client.
2252 * FIXME: I have a stupid name and I should handle the
2253 * requests properly. Still, what we do here is rfc-
2254 * compliant, whereas ignoring or forwarding are not.
2257 * 1 : csp = Current client state (buffers, headers, etc...)
2259 * Returns : http_response if , NULL if nonmatch or handler fail
2261 *********************************************************************/
2262 struct http_response *direct_response(struct client_state *csp)
2264 struct http_response *rsp;
2265 struct list_entry *p;
2267 if ((0 == strcmpic(csp->http->gpc, "trace"))
2268 || (0 == strcmpic(csp->http->gpc, "options")))
2270 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2272 if (!strncmpic(p->str, "Max-Forwards:", 13))
2274 unsigned int max_forwards;
2277 * If it's a Max-Forwards value of zero,
2278 * we have to intercept the request.
2280 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2283 * FIXME: We could handle at least TRACE here,
2284 * but that would require a verbatim copy of
2285 * the request which we don't have anymore
2287 log_error(LOG_LEVEL_HEADER,
2288 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2291 /* Get mem for response or fail*/
2292 if (NULL == (rsp = alloc_http_response()))
2294 return cgi_error_memory();
2297 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2299 free_http_response(rsp);
2300 return cgi_error_memory();
2304 rsp->crunch_reason = UNSUPPORTED;
2306 return(finish_http_response(csp, rsp));
2315 /*********************************************************************
2317 * Function : content_requires_filtering
2319 * Description : Checks whether there are any content filters
2320 * enabled for the current request and if they
2321 * can actually be applied..
2324 * 1 : csp = Current client state (buffers, headers, etc...)
2326 * Returns : TRUE for yes, FALSE otherwise
2328 *********************************************************************/
2329 int content_requires_filtering(struct client_state *csp)
2331 if ((csp->content_type & CT_TABOO)
2332 && !(csp->action->flags & ACTION_FORCE_TEXT_MODE))
2338 * Are we enabling text mode by force?
2340 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2343 * Do we really have to?
2345 if (csp->content_type & CT_TEXT)
2347 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2351 csp->content_type |= CT_TEXT;
2352 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2356 if (!(csp->content_type & CT_DECLARED))
2359 * The server didn't bother to declare a MIME-Type.
2360 * Assume it's text that can be filtered.
2362 * This also regulary happens with 304 responses,
2363 * therefore logging anything here would cause
2366 csp->content_type |= CT_TEXT;
2370 * Choose the applying filter function based on
2371 * the content type and action settings.
2373 if ((csp->content_type & CT_TEXT) &&
2374 (csp->rlist != NULL) &&
2375 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2379 else if ((csp->content_type & CT_GIF) &&
2380 (csp->action->flags & ACTION_DEANIMATE))
2390 /*********************************************************************
2392 * Function : content_filters_enabled
2394 * Description : Checks whether there are any content filters
2395 * enabled for the current request.
2398 * 1 : action = Action spec to check.
2400 * Returns : TRUE for yes, FALSE otherwise
2402 *********************************************************************/
2403 int content_filters_enabled(const struct current_action_spec *action)
2405 return ((action->flags & ACTION_DEANIMATE) ||
2406 !list_is_empty(action->multi[ACTION_MULTI_FILTER]));
2410 /*********************************************************************
2412 * Function : filters_available
2414 * Description : Checks whether there are any filters available.
2417 * 1 : csp = Current client state (buffers, headers, etc...)
2419 * Returns : TRUE for yes, FALSE otherwise.
2421 *********************************************************************/
2422 int filters_available(const struct client_state *csp)
2425 for (i = 0; i < MAX_AF_FILES; i++)
2427 const struct file_list *fl = csp->rlist[i];
2428 if ((NULL != fl) && (NULL != fl->f))