1 const char filters_rcs[] = "$Id: filters.c,v 1.199 2016/01/16 12:33:35 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) 2001-2016 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"
70 #include "deanimate.h"
78 const char filters_h_rcs[] = FILTERS_H_VERSION;
80 typedef char *(*filter_function_ptr)();
81 static filter_function_ptr get_filter_function(const struct client_state *csp);
82 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size);
83 static jb_err prepare_for_filtering(struct client_state *csp);
87 /*********************************************************************
89 * Function : sockaddr_storage_to_ip
91 * Description : Access internal structure of sockaddr_storage
94 * 1 : addr = socket address
95 * 2 : ip = IP address as array of octets in network order
96 * (it points into addr)
97 * 3 : len = length of IP address in octets
98 * 4 : port = port number in network order;
100 * Returns : 0 = no errror; -1 otherwise.
102 *********************************************************************/
103 static int sockaddr_storage_to_ip(const struct sockaddr_storage *addr,
104 uint8_t **ip, unsigned int *len,
112 switch (addr->ss_family)
122 &(((struct sockaddr_in *)addr)->sin_addr.s_addr);
126 *port = &((struct sockaddr_in *)addr)->sin_port;
137 *ip = ((struct sockaddr_in6 *)addr)->sin6_addr.s6_addr;
141 *port = &((struct sockaddr_in6 *)addr)->sin6_port;
146 /* Unsupported address family */
154 /*********************************************************************
156 * Function : match_sockaddr
158 * Description : Check whether address matches network (IP address and port)
161 * 1 : network = socket address of subnework
162 * 2 : netmask = network mask as socket address
163 * 3 : address = checked socket address against given network
165 * Returns : 0 = doesn't match; 1 = does match
167 *********************************************************************/
168 static int match_sockaddr(const struct sockaddr_storage *network,
169 const struct sockaddr_storage *netmask,
170 const struct sockaddr_storage *address)
172 uint8_t *network_addr, *netmask_addr, *address_addr;
173 unsigned int addr_len;
174 in_port_t *network_port, *netmask_port, *address_port;
177 if (network->ss_family != netmask->ss_family)
179 /* This should never happen */
180 assert(network->ss_family == netmask->ss_family);
181 log_error(LOG_LEVEL_FATAL, "Network and netmask differ in family.");
184 sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
185 sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
186 sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
188 /* Check for family */
189 if ((network->ss_family == AF_INET) && (address->ss_family == AF_INET6)
190 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr))
192 /* Map AF_INET6 V4MAPPED address into AF_INET */
196 else if ((network->ss_family == AF_INET6) && (address->ss_family == AF_INET)
197 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr))
199 /* Map AF_INET6 V4MAPPED network into AF_INET */
205 /* XXX: Port check is signaled in netmask */
206 if (*netmask_port && *network_port != *address_port)
211 /* TODO: Optimize by checking by words insted of octets */
212 for (i = 0; (i < addr_len) && netmask_addr[i]; i++)
214 if ((network_addr[i] & netmask_addr[i]) !=
215 (address_addr[i] & netmask_addr[i]))
223 #endif /* def HAVE_RFC2553 */
226 /*********************************************************************
228 * Function : block_acl
230 * Description : Block this request?
231 * Decide yes or no based on ACL file.
234 * 1 : dst = The proxy or gateway address this is going to.
235 * Or NULL to check all possible targets.
236 * 2 : csp = Current client state (buffers, headers, etc...)
237 * Also includes the client IP address.
239 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
241 *********************************************************************/
242 int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
244 struct access_control_list *acl = csp->config->acl;
246 /* if not using an access control list, then permit the connection */
252 /* search the list */
257 match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
259 (csp->ip_addr_long & acl->src->mask) == acl->src->addr
265 /* Just want to check if they have any access */
266 if (acl->action == ACL_PERMIT)
278 * XXX: An undefined acl->dst is full of zeros and should be
279 * considered a wildcard address. sockaddr_storage_to_ip()
280 * fails on such destinations because of unknown sa_familly
281 * (glibc only?). However this test is not portable.
283 * So, we signal the acl->dst is wildcard in wildcard_dst.
286 match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
288 ((dst->addr & acl->dst->mask) == acl->dst->addr)
289 && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
293 if (acl->action == ACL_PERMIT)
311 /*********************************************************************
313 * Function : acl_addr
315 * Description : Called from `load_config' to parse an ACL address.
318 * 1 : aspec = String specifying ACL address.
319 * 2 : aca = struct access_control_addr to fill in.
321 * Returns : 0 => Ok, everything else is an error.
323 *********************************************************************/
324 int acl_addr(const char *aspec, struct access_control_addr *aca)
328 struct addrinfo hints, *result;
330 in_port_t *mask_port;
331 unsigned int addr_len;
334 #endif /* def HAVE_RFC2553 */
336 char *acl_spec = NULL;
339 /* XXX: Depend on ai_family */
347 * Use a temporary acl spec copy so we can log
348 * the unmodified original in case of parse errors.
350 acl_spec = strdup_or_die(aspec);
352 if ((p = strchr(acl_spec, '/')) != NULL)
355 if (privoxy_isdigit(*p) == 0)
360 masklength = atoi(p);
363 if ((masklength < 0) ||
375 if ((*acl_spec == '[') && (NULL != (p = strchr(acl_spec, ']'))))
378 memmove(acl_spec, acl_spec + 1, (size_t)(p - acl_spec));
387 p = strchr(acl_spec, ':');
397 memset(&hints, 0, sizeof(struct addrinfo));
398 hints.ai_family = AF_UNSPEC;
399 hints.ai_socktype = SOCK_STREAM;
401 i = getaddrinfo(acl_spec, p, &hints, &result);
405 log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s",
406 acl_spec, p, gai_strerror(i));
412 /* TODO: Allow multihomed hostnames */
413 memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
414 freeaddrinfo(result);
420 port = strtol(p, &endptr, 10);
422 if (port <= 0 || port > 65535 || *endptr != '\0')
429 aca->port = (unsigned long)port;
431 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
434 if (aca->addr == INADDR_NONE)
436 /* XXX: This will be logged as parse error. */
439 #endif /* def HAVE_RFC2553 */
441 /* build the netmask */
443 /* Clip masklength according to current family. */
444 if ((aca->addr.ss_family == AF_INET) && (masklength > 32))
449 aca->mask.ss_family = aca->addr.ss_family;
450 if (sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port))
457 /* ACL contains a port number, check ports in the future. */
462 * XXX: This could be optimized to operate on whole words instead
463 * of octets (128-bit CPU could do it in one iteration).
466 * Octets after prefix can be omitted because of
467 * previous initialization to zeros.
469 for (i = 0; (i < addr_len) && masklength; i++)
479 * XXX: This assumes MSB of octet is on the left side.
480 * This should be true for all architectures or solved
483 mask_data[i] = (uint8_t)~((1 << (8 - masklength)) - 1);
490 for (i=1; i <= masklength ; i++)
492 aca->mask |= (1U << (32 - i));
495 /* now mask off the host portion of the ip address
496 * (i.e. save on the network portion of the address).
498 aca->addr = aca->addr & aca->mask;
499 #endif /* def HAVE_RFC2553 */
504 #endif /* def FEATURE_ACL */
507 /*********************************************************************
509 * Function : connect_port_is_forbidden
511 * Description : Check to see if CONNECT requests to the destination
512 * port of this request are forbidden. The check is
513 * independend of the actual request method.
516 * 1 : csp = Current client state (buffers, headers, etc...)
518 * Returns : True if yes, false otherwise.
520 *********************************************************************/
521 int connect_port_is_forbidden(const struct client_state *csp)
523 return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
524 !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
529 /*********************************************************************
531 * Function : block_url
533 * Description : Called from `chat'. Check to see if we need to block this.
536 * 1 : csp = Current client state (buffers, headers, etc...)
538 * Returns : NULL => unblocked, else HTTP block response
540 *********************************************************************/
541 struct http_response *block_url(struct client_state *csp)
543 struct http_response *rsp;
544 const char *new_content_type = NULL;
547 * If it's not blocked, don't block it ;-)
549 if ((csp->action->flags & ACTION_BLOCK) == 0)
553 if (csp->action->flags & ACTION_REDIRECT)
555 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
558 * Else, prepare a response
560 if (NULL == (rsp = alloc_http_response()))
562 return cgi_error_memory();
566 * If it's an image-url, send back an image or redirect
567 * as specified by the relevant +image action
569 #ifdef FEATURE_IMAGE_BLOCKING
570 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
574 /* determine HOW images should be blocked */
575 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
577 if (csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
579 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
582 /* and handle accordingly: */
583 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
585 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
586 rsp->body = bindup(image_pattern_data, image_pattern_length);
587 if (rsp->body == NULL)
589 free_http_response(rsp);
590 return cgi_error_memory();
592 rsp->content_length = image_pattern_length;
594 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
596 free_http_response(rsp);
597 return cgi_error_memory();
600 else if (0 == strcmpic(p, "blank"))
602 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
603 rsp->body = bindup(image_blank_data, image_blank_length);
604 if (rsp->body == NULL)
606 free_http_response(rsp);
607 return cgi_error_memory();
609 rsp->content_length = image_blank_length;
611 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
613 free_http_response(rsp);
614 return cgi_error_memory();
619 rsp->status = strdup_or_die("302 Local Redirect from Privoxy");
621 if (enlist_unique_header(rsp->headers, "Location", p))
623 free_http_response(rsp);
624 return cgi_error_memory();
630 #endif /* def FEATURE_IMAGE_BLOCKING */
631 if (csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
634 * Send empty document.
636 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
639 rsp->body = strdup_or_die(" ");
640 rsp->content_length = 1;
642 if (csp->config->feature_flags & RUNTIME_FEATURE_EMPTY_DOC_RETURNS_OK)
645 * Workaround for firefox bug 492459
646 * https://bugzilla.mozilla.org/show_bug.cgi?id=492459
647 * Return a 200 OK status for pages blocked with +handle-as-empty-document
648 * if the "handle-as-empty-doc-returns-ok" runtime config option is set.
650 rsp->status = strdup_or_die("200 Request blocked by Privoxy");
654 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
657 if (new_content_type != 0)
659 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
660 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
662 free_http_response(rsp);
663 return cgi_error_memory();
670 * Else, generate an HTML "blocked" message:
674 struct map * exports;
676 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
678 exports = default_exports(csp, NULL);
681 free_http_response(rsp);
682 return cgi_error_memory();
685 #ifdef FEATURE_FORCE_LOAD
686 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
688 * Export the force conditional block killer if
690 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
691 * - Privoxy is configured to enforce blocks, or
692 * - it's a CONNECT request and enforcing wouldn't work anyway.
694 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
695 || (0 == strcmpic(csp->http->gpc, "connect")))
696 #endif /* ndef FEATURE_FORCE_LOAD */
698 err = map_block_killer(exports, "force-support");
701 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
702 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
703 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
704 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
707 const char *block_reason;
708 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
710 block_reason = csp->action->string[ACTION_STRING_BLOCK];
714 assert(connect_port_is_forbidden(csp));
715 block_reason = "Forbidden CONNECT port.";
717 err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
722 free_http_response(rsp);
723 return cgi_error_memory();
726 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
729 free_http_response(rsp);
730 return cgi_error_memory();
733 rsp->crunch_reason = BLOCKED;
735 return finish_http_response(csp, rsp);
741 /*********************************************************************
743 * Function : trust_url FIXME: I should be called distrust_url
745 * Description : Calls is_untrusted_url to determine if the URL is trusted
746 * and if not, returns a HTTP 403 response with a reject message.
749 * 1 : csp = Current client state (buffers, headers, etc...)
751 * Returns : NULL => trusted, else http_response.
753 *********************************************************************/
754 struct http_response *trust_url(struct client_state *csp)
756 struct http_response *rsp;
757 struct map * exports;
758 char buf[BUFFER_SIZE];
760 struct pattern_spec **tl;
761 struct pattern_spec *t;
765 * Don't bother to work on trusted URLs
767 if (!is_untrusted_url(csp))
773 * Else, prepare a response:
775 if (NULL == (rsp = alloc_http_response()))
777 return cgi_error_memory();
780 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
781 exports = default_exports(csp, NULL);
784 free_http_response(rsp);
785 return cgi_error_memory();
789 * Export the protocol, host, port, and referrer information
791 err = map(exports, "hostport", 1, csp->http->hostport, 1);
792 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
793 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
795 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
797 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
801 if (!err) err = map(exports, "referrer", 1, "none set", 1);
807 free_http_response(rsp);
808 return cgi_error_memory();
812 * Export the trust list
814 p = strdup_or_die("");
815 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
817 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
818 string_append(&p, buf);
820 err = map(exports, "trusted-referrers", 1, p, 0);
825 free_http_response(rsp);
826 return cgi_error_memory();
830 * Export the trust info, if available
832 if (csp->config->trust_info->first)
834 struct list_entry *l;
836 p = strdup_or_die("");
837 for (l = csp->config->trust_info->first; l ; l = l->next)
839 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
840 string_append(&p, buf);
842 err = map(exports, "trust-info", 1, p, 0);
846 err = map_block_killer(exports, "have-trust-info");
852 free_http_response(rsp);
853 return cgi_error_memory();
857 * Export the force conditional block killer if
859 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
860 * - Privoxy is configured to enforce blocks, or
861 * - it's a CONNECT request and enforcing wouldn't work anyway.
863 #ifdef FEATURE_FORCE_LOAD
864 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
865 || (0 == strcmpic(csp->http->gpc, "connect")))
867 err = map_block_killer(exports, "force-support");
871 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
873 #else /* ifndef FEATURE_FORCE_LOAD */
874 err = map_block_killer(exports, "force-support");
875 #endif /* ndef FEATURE_FORCE_LOAD */
880 free_http_response(rsp);
881 return cgi_error_memory();
887 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
890 free_http_response(rsp);
891 return cgi_error_memory();
893 rsp->crunch_reason = UNTRUSTED;
895 return finish_http_response(csp, rsp);
897 #endif /* def FEATURE_TRUST */
900 /*********************************************************************
902 * Function : compile_dynamic_pcrs_job_list
904 * Description : Compiles a dynamic pcrs job list (one with variables
905 * resolved at request time)
908 * 1 : csp = Current client state (buffers, headers, etc...)
909 * 2 : b = The filter list to compile
911 * Returns : NULL in case of errors, otherwise the
914 *********************************************************************/
915 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
917 struct list_entry *pattern;
918 pcrs_job *job_list = NULL;
919 pcrs_job *dummy = NULL;
920 pcrs_job *lastjob = NULL;
923 const struct pcrs_variable variables[] =
925 {"url", csp->http->url, 1},
926 {"path", csp->http->path, 1},
927 {"host", csp->http->host, 1},
928 {"origin", csp->ip_addr_str, 1},
932 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
934 assert(pattern->str != NULL);
936 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
939 log_error(LOG_LEVEL_ERROR,
940 "Compiling dynamic pcrs job '%s' for '%s' failed with error code %d: %s",
941 pattern->str, b->name, error, pcrs_strerror(error));
946 if (error == PCRS_WARN_TRUNCATION)
948 log_error(LOG_LEVEL_ERROR,
949 "At least one of the variables in \'%s\' had to "
950 "be truncated before compilation", pattern->str);
952 if (job_list == NULL)
958 lastjob->next = dummy;
968 /*********************************************************************
970 * Function : rewrite_url
972 * Description : Rewrites a URL with a single pcrs command
973 * and returns the result if it differs from the
974 * original and isn't obviously invalid.
977 * 1 : old_url = URL to rewrite.
978 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
981 * Returns : NULL if the pcrs_command didn't change the url, or
982 * the result of the modification.
984 *********************************************************************/
985 char *rewrite_url(char *old_url, const char *pcrs_command)
987 char *new_url = NULL;
991 assert(pcrs_command);
993 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
997 log_error(LOG_LEVEL_REDIRECTS,
998 "pcrs command \"%s\" didn't change \"%s\".",
999 pcrs_command, old_url);
1004 log_error(LOG_LEVEL_REDIRECTS,
1005 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1006 pcrs_command, old_url, pcrs_strerror(hits));
1009 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1011 log_error(LOG_LEVEL_ERROR,
1012 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1013 "but the result doesn't look like a valid URL and will be ignored.",
1014 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1019 log_error(LOG_LEVEL_REDIRECTS,
1020 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1021 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1029 #ifdef FEATURE_FAST_REDIRECTS
1030 /*********************************************************************
1032 * Function : get_last_url
1034 * Description : Search for the last URL inside a string.
1035 * If the string already is a URL, it will
1036 * be the first URL found.
1039 * 1 : subject = the string to check
1040 * 2 : redirect_mode = +fast-redirect{} mode
1042 * Returns : NULL if no URL was found, or
1043 * the last URL found.
1045 *********************************************************************/
1046 char *get_last_url(char *subject, const char *redirect_mode)
1048 char *new_url = NULL;
1052 assert(redirect_mode);
1054 subject = strdup(subject);
1055 if (subject == NULL)
1057 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1061 if (0 == strcmpic(redirect_mode, "check-decoded-url") && strchr(subject, '%'))
1063 char *url_segment = NULL;
1064 char **url_segments;
1065 size_t max_segments;
1068 log_error(LOG_LEVEL_REDIRECTS,
1069 "Checking \"%s\" for encoded redirects.", subject);
1072 * Check each parameter in the URL separately.
1073 * Sectionize the URL at "?" and "&",
1074 * go backwards through the segments, URL-decode them
1075 * and look for a URL in the decoded result.
1076 * Stop the search after the first match.
1078 * XXX: This estimate is guaranteed to be high enough as we
1079 * let ssplit() ignore empty fields, but also a bit wasteful.
1081 max_segments = strlen(subject) / 2;
1082 url_segments = malloc(max_segments * sizeof(char *));
1084 if (NULL == url_segments)
1086 log_error(LOG_LEVEL_ERROR,
1087 "Out of memory while decoding URL: %s", subject);
1092 segments = ssplit(subject, "?&", url_segments, max_segments);
1094 while (segments-- > 0)
1096 char *dtoken = url_decode(url_segments[segments]);
1099 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", url_segments[segments]);
1102 url_segment = strstr(dtoken, "http://");
1103 if (NULL == url_segment)
1105 url_segment = strstr(dtoken, "https://");
1107 if (NULL != url_segment)
1109 url_segment = strdup_or_die(url_segment);
1116 freez(url_segments);
1118 if (url_segment == NULL)
1122 subject = url_segment;
1126 /* Look for a URL inside this one, without decoding anything. */
1127 log_error(LOG_LEVEL_REDIRECTS,
1128 "Checking \"%s\" for unencoded redirects.", subject);
1132 * Find the last URL encoded in the request
1135 while ((tmp = strstr(tmp, "http://")) != NULL)
1139 tmp = (new_url != NULL) ? new_url : subject;
1140 while ((tmp = strstr(tmp, "https://")) != NULL)
1145 if ((new_url != NULL)
1146 && ( (new_url != subject)
1147 || (0 == strncmpic(subject, "http://", 7))
1148 || (0 == strncmpic(subject, "https://", 8))
1152 * Return new URL if we found a redirect
1153 * or if the subject already was a URL.
1155 * The second case makes sure that we can
1156 * chain get_last_url after another redirection check
1157 * (like rewrite_url) without losing earlier redirects.
1159 new_url = strdup(new_url);
1168 #endif /* def FEATURE_FAST_REDIRECTS */
1171 /*********************************************************************
1173 * Function : redirect_url
1175 * Description : Checks if Privoxy should answer the request with
1176 * a HTTP redirect and generates the redirect if
1180 * 1 : csp = Current client state (buffers, headers, etc...)
1182 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1184 *********************************************************************/
1185 struct http_response *redirect_url(struct client_state *csp)
1187 struct http_response *rsp;
1188 #ifdef FEATURE_FAST_REDIRECTS
1190 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1191 * as compile-time option? The user can easily disable
1192 * it in his action file.
1194 char * redirect_mode;
1195 #endif /* def FEATURE_FAST_REDIRECTS */
1196 char *old_url = NULL;
1197 char *new_url = NULL;
1198 char *redirection_string;
1200 if ((csp->action->flags & ACTION_REDIRECT))
1202 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1205 * If the redirection string begins with 's',
1206 * assume it's a pcrs command, otherwise treat it as
1207 * properly formatted URL and use it for the redirection
1210 * According to (the now obsolete) RFC 2616 section 14.30
1211 * the URL has to be absolute and if the user tries:
1212 * +redirect{sadly/this/will/be/parsed/as/pcrs_command.html}
1213 * she would get undefined results anyway.
1215 * RFC 7231 7.1.2 actually allows relative references,
1216 * but those start with a leading slash (RFC 3986 4.2) and
1217 * thus can't be mistaken for pcrs commands either.
1220 if (*redirection_string == 's')
1222 old_url = csp->http->url;
1223 new_url = rewrite_url(old_url, redirection_string);
1227 log_error(LOG_LEVEL_REDIRECTS,
1228 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1229 redirection_string);
1230 new_url = strdup(redirection_string);
1234 #ifdef FEATURE_FAST_REDIRECTS
1235 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1237 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1240 * If it exists, use the previously rewritten URL as input
1241 * otherwise just use the old path.
1243 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1244 new_url = get_last_url(old_url, redirect_mode);
1249 * Disable redirect checkers, so that they
1250 * will be only run more than once if the user
1251 * also enables them through tags.
1253 * From a performance point of view
1254 * it doesn't matter, but the duplicated
1255 * log messages are annoying.
1257 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1258 #endif /* def FEATURE_FAST_REDIRECTS */
1259 csp->action->flags &= ~ACTION_REDIRECT;
1261 /* Did any redirect action trigger? */
1264 if (url_requires_percent_encoding(new_url))
1267 log_error(LOG_LEVEL_REDIRECTS, "Percent-encoding redirect URL: %N",
1268 strlen(new_url), new_url);
1269 encoded_url = percent_encode_url(new_url);
1271 if (encoded_url == NULL)
1273 return cgi_error_memory();
1275 new_url = encoded_url;
1276 assert(FALSE == url_requires_percent_encoding(new_url));
1279 if (0 == strcmpic(new_url, csp->http->url))
1281 log_error(LOG_LEVEL_ERROR,
1282 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1283 csp->http->url, new_url);
1288 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1290 if (NULL == (rsp = alloc_http_response()))
1293 return cgi_error_memory();
1296 rsp->status = strdup_or_die("302 Local Redirect from Privoxy");
1297 if (enlist_unique_header(rsp->headers, "Location", new_url))
1300 free_http_response(rsp);
1301 return cgi_error_memory();
1303 rsp->crunch_reason = REDIRECTED;
1306 return finish_http_response(csp, rsp);
1310 /* Only reached if no redirect is required */
1316 #ifdef FEATURE_IMAGE_BLOCKING
1317 /*********************************************************************
1319 * Function : is_imageurl
1321 * Description : Given a URL, decide whether it is an image or not,
1322 * using either the info from a previous +image action
1323 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1324 * is MSIE and not on a Mac, tell from the browser's accept
1328 * 1 : csp = Current client state (buffers, headers, etc...)
1330 * Returns : True (nonzero) if URL is an image, false (0)
1333 *********************************************************************/
1334 int is_imageurl(const struct client_state *csp)
1336 #ifdef FEATURE_IMAGE_DETECT_MSIE
1339 tmp = get_header_value(csp->headers, "User-Agent:");
1340 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1342 tmp = get_header_value(csp->headers, "Accept:");
1343 if (tmp && strstr(tmp, "image/gif"))
1345 /* Client will accept HTML. If this seems counterintuitive,
1355 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1357 return ((csp->action->flags & ACTION_IMAGE) != 0);
1360 #endif /* def FEATURE_IMAGE_BLOCKING */
1363 #ifdef FEATURE_TRUST
1364 /*********************************************************************
1366 * Function : is_untrusted_url
1368 * Description : Should we "distrust" this URL (and block it)?
1370 * Yes if it matches a line in the trustfile, or if the
1371 * referrer matches a line starting with "+" in the
1376 * 1 : csp = Current client state (buffers, headers, etc...)
1378 * Returns : 0 => trusted, 1 => untrusted
1380 *********************************************************************/
1381 int is_untrusted_url(const struct client_state *csp)
1383 struct file_list *fl;
1384 struct block_spec *b;
1385 struct pattern_spec **trusted_url;
1386 struct http_request rhttp[1];
1387 const char * referer;
1391 * If we don't have a trustlist, we trust everybody
1393 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1398 memset(rhttp, '\0', sizeof(*rhttp));
1401 * Do we trust the request URL itself?
1403 for (b = b->next; b ; b = b->next)
1405 if (url_match(b->url, csp->http))
1411 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1413 /* no referrer was supplied */
1419 * If not, do we maybe trust its referrer?
1421 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
1427 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1429 if (url_match(*trusted_url, rhttp))
1431 /* if the URL's referrer is from a trusted referrer, then
1432 * add the target spec to the trustfile as an unblocked
1433 * domain and return 0 (which means it's OK).
1438 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1442 char * new_entry = strdup_or_die("~");
1444 string_append(&new_entry, csp->http->hostport);
1446 path = csp->http->path;
1447 if ( (path[0] == '/')
1449 && ((path_end = strchr(path + 2, '/')) != NULL))
1451 /* since this path points into a user's home space
1452 * be sure to include this spec in the trustfile.
1454 long path_len = path_end - path; /* save offset */
1455 path = strdup(path); /* Copy string */
1458 path_end = path + path_len; /* regenerate ptr to new buffer */
1459 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1461 string_join(&new_entry, path);
1465 * Give a reason for generating this entry.
1467 string_append(&new_entry, " # Trusted referrer was: ");
1468 string_append(&new_entry, referer);
1470 if (new_entry != NULL)
1472 if (-1 == fprintf(fp, "%s\n", new_entry))
1474 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1475 new_entry, csp->config->trustfile);
1481 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1482 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1489 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1490 csp->http->hostport, csp->config->trustfile);
1498 #endif /* def FEATURE_TRUST */
1501 /*********************************************************************
1503 * Function : get_filter
1505 * Description : Get a filter with a given name and type.
1506 * Note that taggers are filters, too.
1509 * 1 : csp = Current client state (buffers, headers, etc...)
1510 * 2 : requested_name = Name of the content filter to get
1511 * 3 : requested_type = Type of the filter to tagger to lookup
1513 * Returns : A pointer to the requested filter
1514 * or NULL if the filter wasn't found
1516 *********************************************************************/
1517 struct re_filterfile_spec *get_filter(const struct client_state *csp,
1518 const char *requested_name,
1519 enum filter_type requested_type)
1522 struct re_filterfile_spec *b;
1523 struct file_list *fl;
1525 for (i = 0; i < MAX_AF_FILES; i++)
1528 if ((NULL == fl) || (NULL == fl->f))
1531 * Either there are no filter files left or this
1532 * filter file just contains no valid filters.
1534 * Continue to be sure we don't miss valid filter
1535 * files that are chained after empty or invalid ones.
1540 for (b = fl->f; b != NULL; b = b->next)
1542 if (b->type != requested_type)
1544 /* The callers isn't interested in this filter type. */
1547 if (strcmp(b->name, requested_name) == 0)
1549 /* The requested filter has been found. Abort search. */
1555 /* No filter with the given name and type exists. */
1561 /*********************************************************************
1563 * Function : pcrs_filter_response
1565 * Description : Execute all text substitutions from all applying
1566 * +filter actions on the text buffer that's been
1567 * accumulated in csp->iob->buf.
1570 * 1 : csp = Current client state (buffers, headers, etc...)
1572 * Returns : a pointer to the (newly allocated) modified buffer.
1573 * or NULL if there were no hits or something went wrong
1575 *********************************************************************/
1576 static char *pcrs_filter_response(struct client_state *csp)
1579 size_t size, prev_size;
1585 struct re_filterfile_spec *b;
1586 struct list_entry *filtername;
1591 if (csp->iob->cur >= csp->iob->eod)
1596 if (filters_available(csp) == FALSE)
1598 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1599 "content filtering enabled, but no content filters available.");
1603 size = (size_t)(csp->iob->eod - csp->iob->cur);
1604 old = csp->iob->cur;
1607 * For all applying +filter actions, look if a filter by that
1608 * name exists and if yes, execute it's pcrs_joblist on the
1611 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1612 filtername != NULL; filtername = filtername->next)
1614 int current_hits = 0; /* Number of hits caused by this filter */
1615 int job_number = 0; /* Which job we're currently executing */
1616 int job_hits = 0; /* How many hits the current job caused */
1619 b = get_filter(csp, filtername->str, FT_CONTENT_FILTER);
1625 joblist = b->joblist;
1627 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1629 if (NULL == joblist)
1631 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1636 /* Apply all jobs from the joblist */
1637 for (job = joblist; NULL != job; job = job->next)
1640 job_hits = pcrs_execute(job, old, size, &new, &size);
1645 * That went well. Continue filtering
1646 * and use the result of this job as
1647 * input for the next one.
1649 current_hits += job_hits;
1650 if (old != csp->iob->cur)
1659 * This job caused an unexpected error. Inform the user
1660 * and skip the rest of the jobs in this filter. We could
1661 * continue with the next job, but usually the jobs
1662 * depend on each other or are similar enough to
1663 * fail for the same reason.
1665 * At the moment our pcrs expects the error codes of pcre 3.4,
1666 * but newer pcre versions can return additional error codes.
1667 * As a result pcrs_strerror()'s error message might be
1668 * "Unknown error ...", therefore we print the numerical value
1671 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1672 * should we use LOG_LEVEL_RE_FILTER instead?
1674 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1675 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1680 if (b->dynamic) pcrs_free_joblist(joblist);
1682 log_error(LOG_LEVEL_RE_FILTER,
1683 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
1684 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1686 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;
1701 clear_iob(csp->iob);
1708 #ifdef FEATURE_EXTERNAL_FILTERS
1709 /*********************************************************************
1711 * Function : get_external_filter
1713 * Description : Lookup the code to execute for an external filter.
1714 * Masks the misuse of the re_filterfile_spec.
1717 * 1 : csp = Current client state (buffers, headers, etc...)
1718 * 2 : name = Name of the content filter to get
1720 * Returns : A pointer to the requested code
1721 * or NULL if the filter wasn't found
1723 *********************************************************************/
1724 static const char *get_external_filter(const struct client_state *csp,
1727 struct re_filterfile_spec *external_filter;
1729 external_filter = get_filter(csp, name, FT_EXTERNAL_CONTENT_FILTER);
1730 if (external_filter == NULL)
1732 log_error(LOG_LEVEL_FATAL,
1733 "Didn't find stuff to execute for external filter: %s",
1737 return external_filter->patterns->first->str;
1742 /*********************************************************************
1744 * Function : set_privoxy_variables
1746 * Description : Sets a couple of privoxy-specific environment variables
1749 * 1 : csp = Current client state (buffers, headers, etc...)
1753 *********************************************************************/
1754 static void set_privoxy_variables(const struct client_state *csp)
1761 { "PRIVOXY_URL", csp->http->url },
1762 { "PRIVOXY_PATH", csp->http->path },
1763 { "PRIVOXY_HOST", csp->http->host },
1764 { "PRIVOXY_ORIGIN", csp->ip_addr_str },
1767 for (i = 0; i < SZ(env); i++)
1769 if (setenv(env[i].name, env[i].value, 1))
1771 log_error(LOG_LEVEL_ERROR, "Failed to set %s=%s: %E",
1772 env[i].name, env[i].value);
1778 /*********************************************************************
1780 * Function : execute_external_filter
1782 * Description : Pipe content into external filter and return the output
1785 * 1 : csp = Current client state (buffers, headers, etc...)
1786 * 2 : name = Name of the external filter to execute
1787 * 3 : content = The original content to filter
1788 * 4 : size = The size of the content buffer
1790 * Returns : a pointer to the (newly allocated) modified buffer.
1791 * or NULL if there were no hits or something went wrong
1793 *********************************************************************/
1794 static char *execute_external_filter(const struct client_state *csp,
1795 const char *name, char *content, size_t *size)
1798 char file_name[FILENAME_MAX];
1800 char *filter_output;
1804 const char *external_filter;
1806 if (csp->config->temporary_directory == NULL)
1808 log_error(LOG_LEVEL_ERROR,
1809 "No temporary-directory configured. Can't execute filter: %s",
1814 external_filter = get_external_filter(csp, name);
1816 if (sizeof(file_name) < snprintf(file_name, sizeof(file_name),
1817 "%s/privoxy-XXXXXXXX", csp->config->temporary_directory))
1819 log_error(LOG_LEVEL_ERROR, "temporary-directory path too long");
1823 fd = mkstemp(file_name);
1826 log_error(LOG_LEVEL_ERROR, "mkstemp() failed to create %s: %E", file_name);
1830 fp = fdopen(fd, "w");
1833 log_error(LOG_LEVEL_ERROR, "fdopen() failed: %E");
1839 * The size may be zero if a previous filter discarded everything.
1841 * This isn't necessary unintentional, so we just don't try
1842 * to fwrite() nothing and let the user deal with the rest.
1844 if ((*size != 0) && fwrite(content, *size, 1, fp) != 1)
1846 log_error(LOG_LEVEL_ERROR, "fwrite(..., %d, 1, ..) failed: %E", *size);
1853 if (sizeof(cmd) < snprintf(cmd, sizeof(cmd), "%s < %s", external_filter, file_name))
1855 log_error(LOG_LEVEL_ERROR,
1856 "temporary-directory or external filter path too long");
1861 log_error(LOG_LEVEL_RE_FILTER, "Executing '%s': %s", name, cmd);
1864 * The locking is necessary to prevent other threads
1865 * from overwriting the environment variables before
1866 * the popen fork. Afterwards this no longer matters.
1868 privoxy_mutex_lock(&external_filter_mutex);
1869 set_privoxy_variables(csp);
1870 fp = popen(cmd, "r");
1871 privoxy_mutex_unlock(&external_filter_mutex);
1874 log_error(LOG_LEVEL_ERROR, "popen(\"%s\", \"r\") failed: %E", cmd);
1879 /* Allocate at least one byte */
1880 filter_output = malloc_or_die(*size + 1);
1883 while (!feof(fp) && !ferror(fp))
1886 /* Could be bigger ... */
1887 enum { READ_LENGTH = 2048 };
1889 if (new_size + READ_LENGTH >= *size)
1893 /* Could be considered wasteful if the content is 'large'. */
1894 *size += (*size >= READ_LENGTH) ? *size : READ_LENGTH;
1896 p = realloc(filter_output, *size);
1899 log_error(LOG_LEVEL_ERROR, "Out of memory while reading "
1900 "external filter output. Using what we got so far.");
1905 assert(new_size + READ_LENGTH < *size);
1906 len = fread(&filter_output[new_size], 1, READ_LENGTH, fp);
1916 log_error(LOG_LEVEL_ERROR, "Executing %s failed: %E", cmd);
1920 log_error(LOG_LEVEL_RE_FILTER,
1921 "Executing '%s' resulted in return value %d. "
1922 "Read %d of up to %d bytes.", name, (ret >> 8), new_size, *size);
1928 return filter_output;
1931 #endif /* def FEATURE_EXTERNAL_FILTERS */
1934 /*********************************************************************
1936 * Function : gif_deanimate_response
1938 * Description : Deanimate the GIF image that has been accumulated in
1939 * csp->iob->buf, set csp->content_length to the modified
1940 * size and raise the CSP_FLAG_MODIFIED flag.
1943 * 1 : csp = Current client state (buffers, headers, etc...)
1945 * Returns : a pointer to the (newly allocated) modified buffer.
1946 * or NULL in case something went wrong.
1948 *********************************************************************/
1949 static char *gif_deanimate_response(struct client_state *csp)
1951 struct binbuffer *in, *out;
1955 size = (size_t)(csp->iob->eod - csp->iob->cur);
1957 in = zalloc_or_die(sizeof(*in));
1958 out = zalloc_or_die(sizeof(*out));
1960 in->buffer = csp->iob->cur;
1963 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1965 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1972 if ((int)size == out->offset)
1974 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1978 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1980 csp->content_length = out->offset;
1981 csp->flags |= CSP_FLAG_MODIFIED;
1991 /*********************************************************************
1993 * Function : get_filter_function
1995 * Description : Decides which content filter function has
1996 * to be applied (if any). Only considers functions
1997 * for internal filters which are mutually-exclusive.
2000 * 1 : csp = Current client state (buffers, headers, etc...)
2002 * Returns : The content filter function to run, or
2003 * NULL if no content filter is active
2005 *********************************************************************/
2006 static filter_function_ptr get_filter_function(const struct client_state *csp)
2008 filter_function_ptr filter_function = NULL;
2011 * Choose the applying filter function based on
2012 * the content type and action settings.
2014 if ((csp->content_type & CT_TEXT) &&
2015 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2017 filter_function = pcrs_filter_response;
2019 else if ((csp->content_type & CT_GIF) &&
2020 (csp->action->flags & ACTION_DEANIMATE))
2022 filter_function = gif_deanimate_response;
2025 return filter_function;
2029 /*********************************************************************
2031 * Function : remove_chunked_transfer_coding
2033 * Description : In-situ remove the "chunked" transfer coding as defined
2034 * in RFC 7230 4.1 from a buffer. XXX: The implementation
2035 * is neither complete nor compliant (TODO #129).
2038 * 1 : buffer = Pointer to the text buffer
2039 * 2 : size = In: Number of bytes to be processed,
2040 * Out: Number of bytes after de-chunking.
2041 * (undefined in case of errors)
2043 * Returns : JB_ERR_OK for success,
2044 * JB_ERR_PARSE otherwise
2046 *********************************************************************/
2047 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2050 unsigned int chunksize = 0;
2051 char *from_p, *to_p;
2052 const char *end_of_buffer = buffer + *size;
2055 from_p = to_p = buffer;
2057 if (sscanf(buffer, "%x", &chunksize) != 1)
2059 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2060 return JB_ERR_PARSE;
2063 while (chunksize > 0U)
2066 * If the chunk-size is valid, we should have at least
2067 * chunk-size bytes of chunk-data and five bytes of
2068 * meta data (chunk-size, CRLF, CRLF) left in the buffer.
2070 if (chunksize + 5 >= *size - newsize)
2072 log_error(LOG_LEVEL_ERROR,
2073 "Chunk size %u exceeds buffered data left. "
2074 "Already digested %u of %u buffered bytes.",
2075 chunksize, (unsigned int)newsize, (unsigned int)*size);
2076 return JB_ERR_PARSE;
2080 * Skip the chunk-size, the optional chunk-ext and the CRLF
2081 * that is supposed to be located directly before the start
2084 if (NULL == (from_p = strstr(from_p, "\r\n")))
2086 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2087 return JB_ERR_PARSE;
2092 * The previous strstr() does not enforce chunk-validity
2093 * and is sattisfied as long a CRLF is left in the buffer.
2095 * Make sure the bytes we consider chunk-data are within
2098 if (from_p + chunksize >= end_of_buffer)
2100 log_error(LOG_LEVEL_ERROR,
2101 "End of chunk is beyond the end of the buffer.");
2102 return JB_ERR_PARSE;
2105 memmove(to_p, from_p, (size_t) chunksize);
2106 newsize += chunksize;
2107 to_p = buffer + newsize;
2108 from_p += chunksize;
2111 * Not merging this check with the previous one allows us
2112 * to keep chunks without trailing CRLF. It's not clear
2113 * if we actually have to care about those, though.
2115 if (from_p + 2 >= end_of_buffer)
2117 log_error(LOG_LEVEL_ERROR, "Not enough room for trailing CRLF.");
2118 return JB_ERR_PARSE;
2121 if (sscanf(from_p, "%x", &chunksize) != 1)
2123 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
2128 /* XXX: Should get its own loglevel. */
2129 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
2138 /*********************************************************************
2140 * Function : prepare_for_filtering
2142 * Description : If necessary, de-chunks and decompresses
2143 * the content so it can get filterd.
2146 * 1 : csp = Current client state (buffers, headers, etc...)
2148 * Returns : JB_ERR_OK for success,
2149 * JB_ERR_PARSE otherwise
2151 *********************************************************************/
2152 static jb_err prepare_for_filtering(struct client_state *csp)
2154 jb_err err = JB_ERR_OK;
2157 * If the body has a "chunked" transfer-encoding,
2158 * get rid of it, adjusting size and iob->eod
2160 if (csp->flags & CSP_FLAG_CHUNKED)
2162 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2164 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2165 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2166 if (JB_ERR_OK == err)
2168 csp->iob->eod = csp->iob->cur + size;
2169 csp->flags |= CSP_FLAG_MODIFIED;
2173 return JB_ERR_PARSE;
2179 * If the body has a supported transfer-encoding,
2180 * decompress it, adjusting size and iob->eod.
2182 if (csp->content_type & (CT_GZIP|CT_DEFLATE))
2184 if (0 == csp->iob->eod - csp->iob->cur)
2186 /* Nothing left after de-chunking. */
2190 err = decompress_iob(csp);
2192 if (JB_ERR_OK == err)
2194 csp->flags |= CSP_FLAG_MODIFIED;
2195 csp->content_type &= ~CT_TABOO;
2200 * Unset CT_GZIP and CT_DEFLATE to remember not
2201 * to modify the Content-Encoding header later.
2203 csp->content_type &= ~CT_GZIP;
2204 csp->content_type &= ~CT_DEFLATE;
2213 /*********************************************************************
2215 * Function : execute_content_filters
2217 * Description : Executes a given content filter.
2220 * 1 : csp = Current client state (buffers, headers, etc...)
2222 * Returns : Pointer to the modified buffer, or
2223 * NULL if filtering failed or wasn't necessary.
2225 *********************************************************************/
2226 char *execute_content_filters(struct client_state *csp)
2229 filter_function_ptr content_filter;
2231 assert(content_filters_enabled(csp->action));
2233 if (0 == csp->iob->eod - csp->iob->cur)
2236 * No content (probably status code 301, 302 ...),
2237 * no filtering necessary.
2242 if (JB_ERR_OK != prepare_for_filtering(csp))
2245 * failed to de-chunk or decompress.
2250 if (0 == csp->iob->eod - csp->iob->cur)
2253 * Clown alarm: chunked and/or compressed nothing delivered.
2258 content_filter = get_filter_function(csp);
2259 content = (content_filter != NULL) ? (*content_filter)(csp) : NULL;
2261 #ifdef FEATURE_EXTERNAL_FILTERS
2262 if ((csp->content_type & CT_TEXT) &&
2263 !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]))
2265 struct list_entry *filtername;
2266 size_t size = (size_t)csp->content_length;
2268 if (content == NULL)
2270 content = csp->iob->cur;
2271 size = (size_t)(csp->iob->eod - csp->iob->cur);
2274 for (filtername = csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]->first;
2275 filtername ; filtername = filtername->next)
2277 char *result = execute_external_filter(csp, filtername->str, content, &size);
2280 if (content != csp->iob->cur)
2287 csp->flags |= CSP_FLAG_MODIFIED;
2288 csp->content_length = size;
2290 #endif /* def FEATURE_EXTERNAL_FILTERS */
2297 /*********************************************************************
2299 * Function : get_url_actions
2301 * Description : Gets the actions for this URL.
2304 * 1 : csp = Current client state (buffers, headers, etc...)
2305 * 2 : http = http_request request for blocked URLs
2309 *********************************************************************/
2310 void get_url_actions(struct client_state *csp, struct http_request *http)
2312 struct file_list *fl;
2313 struct url_actions *b;
2316 init_current_action(csp->action);
2318 for (i = 0; i < MAX_AF_FILES; i++)
2320 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2325 apply_url_actions(csp->action, http, b);
2332 /*********************************************************************
2334 * Function : apply_url_actions
2336 * Description : Applies a list of URL actions.
2339 * 1 : action = Destination.
2340 * 2 : http = Current URL
2341 * 3 : b = list of URL actions to apply
2345 *********************************************************************/
2346 void apply_url_actions(struct current_action_spec *action,
2347 struct http_request *http,
2348 struct url_actions *b)
2352 /* Should never happen */
2356 for (b = b->next; NULL != b; b = b->next)
2358 if (url_match(b->url, http))
2360 merge_current_action(action, b->action);
2366 /*********************************************************************
2368 * Function : get_forward_override_settings
2370 * Description : Returns forward settings as specified with the
2371 * forward-override{} action. forward-override accepts
2372 * forward lines similar to the one used in the
2373 * configuration file, but without the URL pattern.
2379 * in the configuration file can be replaced with
2380 * the action section:
2382 * {+forward-override{forward .}}
2386 * 1 : csp = Current client state (buffers, headers, etc...)
2388 * Returns : Pointer to forwarding structure in case of success.
2389 * Invalid syntax is fatal.
2391 *********************************************************************/
2392 static const struct forward_spec *get_forward_override_settings(struct client_state *csp)
2394 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2395 char forward_settings[BUFFER_SIZE];
2396 char *http_parent = NULL;
2397 /* variable names were chosen for consistency reasons. */
2398 struct forward_spec *fwd = NULL;
2402 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2403 /* Should be enforced by load_one_actions_file() */
2404 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2406 /* Create a copy ssplit can modify */
2407 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2409 if (NULL != csp->fwd)
2412 * XXX: Currently necessary to prevent memory
2413 * leaks when the show-url-info cgi page is visited.
2415 unload_forward_spec(csp->fwd);
2419 * allocate a new forward node, valid only for
2420 * the lifetime of this request. Save its location
2421 * in csp as well, so sweep() can free it later on.
2423 fwd = csp->fwd = zalloc_or_die(sizeof(*fwd));
2425 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec));
2426 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2428 fwd->type = SOCKS_NONE;
2430 /* Parse the parent HTTP proxy host:port */
2431 http_parent = vec[1];
2434 else if ((vec_count == 2) && !strcasecmp(vec[0], "forward-webserver"))
2436 fwd->type = FORWARD_WEBSERVER;
2438 /* Parse the parent HTTP server host:port */
2439 http_parent = vec[1];
2442 else if (vec_count == 3)
2444 char *socks_proxy = NULL;
2446 if (!strcasecmp(vec[0], "forward-socks4"))
2448 fwd->type = SOCKS_4;
2449 socks_proxy = vec[1];
2451 else if (!strcasecmp(vec[0], "forward-socks4a"))
2453 fwd->type = SOCKS_4A;
2454 socks_proxy = vec[1];
2456 else if (!strcasecmp(vec[0], "forward-socks5"))
2458 fwd->type = SOCKS_5;
2459 socks_proxy = vec[1];
2461 else if (!strcasecmp(vec[0], "forward-socks5t"))
2463 fwd->type = SOCKS_5T;
2464 socks_proxy = vec[1];
2467 if (NULL != socks_proxy)
2469 /* Parse the SOCKS proxy host[:port] */
2470 fwd->gateway_port = 1080;
2471 parse_forwarder_address(socks_proxy,
2472 &fwd->gateway_host, &fwd->gateway_port);
2474 http_parent = vec[2];
2478 if (NULL == http_parent)
2480 log_error(LOG_LEVEL_FATAL,
2481 "Invalid forward-override syntax in: %s", forward_override_line);
2482 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2485 /* Parse http forwarding settings */
2486 if (strcmp(http_parent, ".") != 0)
2488 fwd->forward_port = 8000;
2489 parse_forwarder_address(http_parent,
2490 &fwd->forward_host, &fwd->forward_port);
2493 assert (NULL != fwd);
2495 log_error(LOG_LEVEL_CONNECT,
2496 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2502 /*********************************************************************
2504 * Function : forward_url
2506 * Description : Should we forward this to another proxy?
2509 * 1 : csp = Current client state (buffers, headers, etc...)
2510 * 2 : http = http_request request for current URL
2512 * Returns : Pointer to forwarding information.
2514 *********************************************************************/
2515 const struct forward_spec *forward_url(struct client_state *csp,
2516 const struct http_request *http)
2518 static const struct forward_spec fwd_default[1]; /* Zero'ed due to being static. */
2519 struct forward_spec *fwd = csp->config->forward;
2521 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2523 return get_forward_override_settings(csp);
2533 if (url_match(fwd->url, http))
2544 /*********************************************************************
2546 * Function : direct_response
2548 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2549 * request and if so, return a HTTP 501 to the client.
2551 * FIXME: I have a stupid name and I should handle the
2552 * requests properly. Still, what we do here is rfc-
2553 * compliant, whereas ignoring or forwarding are not.
2556 * 1 : csp = Current client state (buffers, headers, etc...)
2558 * Returns : http_response if , NULL if nonmatch or handler fail
2560 *********************************************************************/
2561 struct http_response *direct_response(struct client_state *csp)
2563 struct http_response *rsp;
2564 struct list_entry *p;
2566 if ((0 == strcmpic(csp->http->gpc, "trace"))
2567 || (0 == strcmpic(csp->http->gpc, "options")))
2569 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2571 if (!strncmpic(p->str, "Max-Forwards:", 13))
2573 unsigned int max_forwards;
2576 * If it's a Max-Forwards value of zero,
2577 * we have to intercept the request.
2579 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2582 * FIXME: We could handle at least TRACE here,
2583 * but that would require a verbatim copy of
2584 * the request which we don't have anymore
2586 log_error(LOG_LEVEL_HEADER,
2587 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2590 /* Get mem for response or fail*/
2591 if (NULL == (rsp = alloc_http_response()))
2593 return cgi_error_memory();
2596 rsp->status = strdup_or_die("501 Not Implemented");
2598 rsp->crunch_reason = UNSUPPORTED;
2600 return(finish_http_response(csp, rsp));
2609 /*********************************************************************
2611 * Function : content_requires_filtering
2613 * Description : Checks whether there are any content filters
2614 * enabled for the current request and if they
2615 * can actually be applied..
2618 * 1 : csp = Current client state (buffers, headers, etc...)
2620 * Returns : TRUE for yes, FALSE otherwise
2622 *********************************************************************/
2623 int content_requires_filtering(struct client_state *csp)
2625 if ((csp->content_type & CT_TABOO)
2626 && !(csp->action->flags & ACTION_FORCE_TEXT_MODE))
2632 * Are we enabling text mode by force?
2634 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2637 * Do we really have to?
2639 if (csp->content_type & CT_TEXT)
2641 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2645 csp->content_type |= CT_TEXT;
2646 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2650 if (!(csp->content_type & CT_DECLARED))
2653 * The server didn't bother to declare a MIME-Type.
2654 * Assume it's text that can be filtered.
2656 * This also regulary happens with 304 responses,
2657 * therefore logging anything here would cause
2660 csp->content_type |= CT_TEXT;
2664 * Choose the applying filter function based on
2665 * the content type and action settings.
2667 if ((csp->content_type & CT_TEXT) &&
2668 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER]) ||
2669 !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER])))
2673 else if ((csp->content_type & CT_GIF) &&
2674 (csp->action->flags & ACTION_DEANIMATE))
2684 /*********************************************************************
2686 * Function : content_filters_enabled
2688 * Description : Checks whether there are any content filters
2689 * enabled for the current request.
2692 * 1 : action = Action spec to check.
2694 * Returns : TRUE for yes, FALSE otherwise
2696 *********************************************************************/
2697 int content_filters_enabled(const struct current_action_spec *action)
2699 return ((action->flags & ACTION_DEANIMATE) ||
2700 !list_is_empty(action->multi[ACTION_MULTI_FILTER]) ||
2701 !list_is_empty(action->multi[ACTION_MULTI_EXTERNAL_FILTER]));
2705 /*********************************************************************
2707 * Function : filters_available
2709 * Description : Checks whether there are any filters available.
2712 * 1 : csp = Current client state (buffers, headers, etc...)
2714 * Returns : TRUE for yes, FALSE otherwise.
2716 *********************************************************************/
2717 int filters_available(const struct client_state *csp)
2720 for (i = 0; i < MAX_AF_FILES; i++)
2722 const struct file_list *fl = csp->rlist[i];
2723 if ((NULL != fl) && (NULL != fl->f))