1 const char filters_rcs[] = "$Id: filters.c,v 1.206 2017/06/04 14:43:10 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"
73 #ifdef FEATURE_CLIENT_TAGS
74 #include "client-tags.h"
81 const char filters_h_rcs[] = FILTERS_H_VERSION;
83 typedef char *(*filter_function_ptr)();
84 static filter_function_ptr get_filter_function(const struct client_state *csp);
85 static jb_err prepare_for_filtering(struct client_state *csp);
86 static void apply_url_actions(struct current_action_spec *action,
87 struct http_request *http,
88 #ifdef FEATURE_CLIENT_TAGS
89 const struct list *client_tags,
91 struct url_actions *b);
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;
110 *********************************************************************/
111 static void sockaddr_storage_to_ip(const struct sockaddr_storage *addr,
112 uint8_t **ip, unsigned int *len,
115 assert(NULL != addr);
116 assert(addr->ss_family == AF_INET || addr->ss_family == AF_INET6);
118 switch (addr->ss_family)
128 &(((struct sockaddr_in *)addr)->sin_addr.s_addr);
132 *port = &((struct sockaddr_in *)addr)->sin_port;
143 *ip = ((struct sockaddr_in6 *)addr)->sin6_addr.s6_addr;
147 *port = &((struct sockaddr_in6 *)addr)->sin6_port;
155 /*********************************************************************
157 * Function : match_sockaddr
159 * Description : Check whether address matches network (IP address and port)
162 * 1 : network = socket address of subnework
163 * 2 : netmask = network mask as socket address
164 * 3 : address = checked socket address against given network
166 * Returns : 0 = doesn't match; 1 = does match
168 *********************************************************************/
169 static int match_sockaddr(const struct sockaddr_storage *network,
170 const struct sockaddr_storage *netmask,
171 const struct sockaddr_storage *address)
173 uint8_t *network_addr, *netmask_addr, *address_addr;
174 unsigned int addr_len;
175 in_port_t *network_port, *netmask_port, *address_port;
178 if (network->ss_family != netmask->ss_family)
180 /* This should never happen */
181 assert(network->ss_family == netmask->ss_family);
182 log_error(LOG_LEVEL_FATAL, "Network and netmask differ in family.");
185 sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
186 sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
187 sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
189 /* Check for family */
190 if ((network->ss_family == AF_INET) && (address->ss_family == AF_INET6)
191 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr))
193 /* Map AF_INET6 V4MAPPED address into AF_INET */
197 else if ((network->ss_family == AF_INET6) && (address->ss_family == AF_INET)
198 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr))
200 /* Map AF_INET6 V4MAPPED network into AF_INET */
206 /* XXX: Port check is signaled in netmask */
207 if (*netmask_port && *network_port != *address_port)
212 /* TODO: Optimize by checking by words insted of octets */
213 for (i = 0; (i < addr_len) && netmask_addr[i]; i++)
215 if ((network_addr[i] & netmask_addr[i]) !=
216 (address_addr[i] & netmask_addr[i]))
224 #endif /* def HAVE_RFC2553 */
227 /*********************************************************************
229 * Function : block_acl
231 * Description : Block this request?
232 * Decide yes or no based on ACL file.
235 * 1 : dst = The proxy or gateway address this is going to.
236 * Or NULL to check all possible targets.
237 * 2 : csp = Current client state (buffers, headers, etc...)
238 * Also includes the client IP address.
240 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
242 *********************************************************************/
243 int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
245 struct access_control_list *acl = csp->config->acl;
247 /* if not using an access control list, then permit the connection */
253 /* search the list */
258 match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
260 (csp->ip_addr_long & acl->src->mask) == acl->src->addr
266 /* Just want to check if they have any access */
267 if (acl->action == ACL_PERMIT)
279 * XXX: An undefined acl->dst is full of zeros and should be
280 * considered a wildcard address. sockaddr_storage_to_ip()
281 * fails on such destinations because of unknown sa_familly
282 * (glibc only?). However this test is not portable.
284 * So, we signal the acl->dst is wildcard in wildcard_dst.
287 match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
289 ((dst->addr & acl->dst->mask) == acl->dst->addr)
290 && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
294 if (acl->action == ACL_PERMIT)
312 /*********************************************************************
314 * Function : acl_addr
316 * Description : Called from `load_config' to parse an ACL address.
319 * 1 : aspec = String specifying ACL address.
320 * 2 : aca = struct access_control_addr to fill in.
322 * Returns : 0 => Ok, everything else is an error.
324 *********************************************************************/
325 int acl_addr(const char *aspec, struct access_control_addr *aca)
329 struct addrinfo hints, *result;
331 in_port_t *mask_port;
332 unsigned int addr_len;
335 #endif /* def HAVE_RFC2553 */
337 char *acl_spec = NULL;
340 /* XXX: Depend on ai_family */
348 * Use a temporary acl spec copy so we can log
349 * the unmodified original in case of parse errors.
351 acl_spec = strdup_or_die(aspec);
353 if ((p = strchr(acl_spec, '/')) != NULL)
356 if (privoxy_isdigit(*p) == 0)
361 masklength = atoi(p);
364 if ((masklength < 0) ||
376 if ((*acl_spec == '[') && (NULL != (p = strchr(acl_spec, ']'))))
379 memmove(acl_spec, acl_spec + 1, (size_t)(p - acl_spec));
388 p = strchr(acl_spec, ':');
398 memset(&hints, 0, sizeof(struct addrinfo));
399 hints.ai_family = AF_UNSPEC;
400 hints.ai_socktype = SOCK_STREAM;
402 i = getaddrinfo(acl_spec, p, &hints, &result);
406 log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s",
407 acl_spec, p, gai_strerror(i));
413 /* TODO: Allow multihomed hostnames */
414 memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
415 freeaddrinfo(result);
421 port = strtol(p, &endptr, 10);
423 if (port <= 0 || port > 65535 || *endptr != '\0')
430 aca->port = (unsigned long)port;
432 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
435 if (aca->addr == INADDR_NONE)
437 /* XXX: This will be logged as parse error. */
440 #endif /* def HAVE_RFC2553 */
442 /* build the netmask */
444 /* Clip masklength according to current family. */
445 if ((aca->addr.ss_family == AF_INET) && (masklength > 32))
450 aca->mask.ss_family = aca->addr.ss_family;
451 sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port);
455 /* ACL contains a port number, check ports in the future. */
460 * XXX: This could be optimized to operate on whole words instead
461 * of octets (128-bit CPU could do it in one iteration).
464 * Octets after prefix can be omitted because of
465 * previous initialization to zeros.
467 for (i = 0; (i < addr_len) && masklength; i++)
477 * XXX: This assumes MSB of octet is on the left side.
478 * This should be true for all architectures or solved
481 mask_data[i] = (uint8_t)~((1 << (8 - masklength)) - 1);
488 for (i=1; i <= masklength ; i++)
490 aca->mask |= (1U << (32 - i));
493 /* now mask off the host portion of the ip address
494 * (i.e. save on the network portion of the address).
496 aca->addr = aca->addr & aca->mask;
497 #endif /* def HAVE_RFC2553 */
502 #endif /* def FEATURE_ACL */
505 /*********************************************************************
507 * Function : connect_port_is_forbidden
509 * Description : Check to see if CONNECT requests to the destination
510 * port of this request are forbidden. The check is
511 * independend of the actual request method.
514 * 1 : csp = Current client state (buffers, headers, etc...)
516 * Returns : True if yes, false otherwise.
518 *********************************************************************/
519 int connect_port_is_forbidden(const struct client_state *csp)
521 return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
522 !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
527 /*********************************************************************
529 * Function : block_url
531 * Description : Called from `chat'. Check to see if we need to block this.
534 * 1 : csp = Current client state (buffers, headers, etc...)
536 * Returns : NULL => unblocked, else HTTP block response
538 *********************************************************************/
539 struct http_response *block_url(struct client_state *csp)
541 struct http_response *rsp;
542 const char *new_content_type = NULL;
545 * If it's not blocked, don't block it ;-)
547 if ((csp->action->flags & ACTION_BLOCK) == 0)
551 if (csp->action->flags & ACTION_REDIRECT)
553 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
556 * Else, prepare a response
558 if (NULL == (rsp = alloc_http_response()))
560 return cgi_error_memory();
564 * If it's an image-url, send back an image or redirect
565 * as specified by the relevant +image action
567 #ifdef FEATURE_IMAGE_BLOCKING
568 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
572 /* determine HOW images should be blocked */
573 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
575 if (csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
577 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
580 /* and handle accordingly: */
581 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
583 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
584 rsp->body = bindup(image_pattern_data, image_pattern_length);
585 if (rsp->body == NULL)
587 free_http_response(rsp);
588 return cgi_error_memory();
590 rsp->content_length = image_pattern_length;
592 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
594 free_http_response(rsp);
595 return cgi_error_memory();
598 else if (0 == strcmpic(p, "blank"))
600 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
601 rsp->body = bindup(image_blank_data, image_blank_length);
602 if (rsp->body == NULL)
604 free_http_response(rsp);
605 return cgi_error_memory();
607 rsp->content_length = image_blank_length;
609 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
611 free_http_response(rsp);
612 return cgi_error_memory();
617 rsp->status = strdup_or_die("302 Local Redirect from Privoxy");
619 if (enlist_unique_header(rsp->headers, "Location", p))
621 free_http_response(rsp);
622 return cgi_error_memory();
628 #endif /* def FEATURE_IMAGE_BLOCKING */
629 if (csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
632 * Send empty document.
634 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
637 rsp->body = strdup_or_die(" ");
638 rsp->content_length = 1;
640 if (csp->config->feature_flags & RUNTIME_FEATURE_EMPTY_DOC_RETURNS_OK)
643 * Workaround for firefox bug 492459
644 * https://bugzilla.mozilla.org/show_bug.cgi?id=492459
645 * Return a 200 OK status for pages blocked with +handle-as-empty-document
646 * if the "handle-as-empty-doc-returns-ok" runtime config option is set.
648 rsp->status = strdup_or_die("200 Request blocked by Privoxy");
652 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
655 if (new_content_type != 0)
657 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
658 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
660 free_http_response(rsp);
661 return cgi_error_memory();
668 * Else, generate an HTML "blocked" message:
672 struct map * exports;
674 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
676 exports = default_exports(csp, NULL);
679 free_http_response(rsp);
680 return cgi_error_memory();
683 #ifdef FEATURE_FORCE_LOAD
684 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
686 * Export the force conditional block killer if
688 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
689 * - Privoxy is configured to enforce blocks, or
690 * - it's a CONNECT request and enforcing wouldn't work anyway.
692 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
693 || (0 == strcmpic(csp->http->gpc, "connect")))
694 #endif /* ndef FEATURE_FORCE_LOAD */
696 err = map_block_killer(exports, "force-support");
699 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
700 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
701 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
702 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
705 const char *block_reason;
706 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
708 block_reason = csp->action->string[ACTION_STRING_BLOCK];
712 assert(connect_port_is_forbidden(csp));
713 block_reason = "Forbidden CONNECT port.";
715 err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
720 free_http_response(rsp);
721 return cgi_error_memory();
724 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
727 free_http_response(rsp);
728 return cgi_error_memory();
731 rsp->crunch_reason = BLOCKED;
733 return finish_http_response(csp, rsp);
739 /*********************************************************************
741 * Function : trust_url FIXME: I should be called distrust_url
743 * Description : Calls is_untrusted_url to determine if the URL is trusted
744 * and if not, returns a HTTP 403 response with a reject message.
747 * 1 : csp = Current client state (buffers, headers, etc...)
749 * Returns : NULL => trusted, else http_response.
751 *********************************************************************/
752 struct http_response *trust_url(struct client_state *csp)
754 struct http_response *rsp;
755 struct map * exports;
756 char buf[BUFFER_SIZE];
758 struct pattern_spec **tl;
759 struct pattern_spec *t;
763 * Don't bother to work on trusted URLs
765 if (!is_untrusted_url(csp))
771 * Else, prepare a response:
773 if (NULL == (rsp = alloc_http_response()))
775 return cgi_error_memory();
778 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
779 exports = default_exports(csp, NULL);
782 free_http_response(rsp);
783 return cgi_error_memory();
787 * Export the protocol, host, port, and referrer information
789 err = map(exports, "hostport", 1, csp->http->hostport, 1);
790 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
791 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
793 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
795 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
799 if (!err) err = map(exports, "referrer", 1, "none set", 1);
805 free_http_response(rsp);
806 return cgi_error_memory();
810 * Export the trust list
812 p = strdup_or_die("");
813 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
815 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
816 string_append(&p, buf);
818 err = map(exports, "trusted-referrers", 1, p, 0);
823 free_http_response(rsp);
824 return cgi_error_memory();
828 * Export the trust info, if available
830 if (csp->config->trust_info->first)
832 struct list_entry *l;
834 p = strdup_or_die("");
835 for (l = csp->config->trust_info->first; l ; l = l->next)
837 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
838 string_append(&p, buf);
840 err = map(exports, "trust-info", 1, p, 0);
844 err = map_block_killer(exports, "have-trust-info");
850 free_http_response(rsp);
851 return cgi_error_memory();
855 * Export the force conditional block killer if
857 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
858 * - Privoxy is configured to enforce blocks, or
859 * - it's a CONNECT request and enforcing wouldn't work anyway.
861 #ifdef FEATURE_FORCE_LOAD
862 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
863 || (0 == strcmpic(csp->http->gpc, "connect")))
865 err = map_block_killer(exports, "force-support");
869 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
871 #else /* ifndef FEATURE_FORCE_LOAD */
872 err = map_block_killer(exports, "force-support");
873 #endif /* ndef FEATURE_FORCE_LOAD */
878 free_http_response(rsp);
879 return cgi_error_memory();
885 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
888 free_http_response(rsp);
889 return cgi_error_memory();
891 rsp->crunch_reason = UNTRUSTED;
893 return finish_http_response(csp, rsp);
895 #endif /* def FEATURE_TRUST */
898 /*********************************************************************
900 * Function : compile_dynamic_pcrs_job_list
902 * Description : Compiles a dynamic pcrs job list (one with variables
903 * resolved at request time)
906 * 1 : csp = Current client state (buffers, headers, etc...)
907 * 2 : b = The filter list to compile
909 * Returns : NULL in case of errors, otherwise the
912 *********************************************************************/
913 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
915 struct list_entry *pattern;
916 pcrs_job *job_list = NULL;
917 pcrs_job *dummy = NULL;
918 pcrs_job *lastjob = NULL;
921 const struct pcrs_variable variables[] =
923 {"url", csp->http->url, 1},
924 {"path", csp->http->path, 1},
925 {"host", csp->http->host, 1},
926 {"origin", csp->ip_addr_str, 1},
927 {"listen-address", csp->listen_addr_str, 1},
931 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
933 assert(pattern->str != NULL);
935 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
938 log_error(LOG_LEVEL_ERROR,
939 "Compiling dynamic pcrs job '%s' for '%s' failed with error code %d: %s",
940 pattern->str, b->name, error, pcrs_strerror(error));
945 if (error == PCRS_WARN_TRUNCATION)
947 log_error(LOG_LEVEL_ERROR,
948 "At least one of the variables in \'%s\' had to "
949 "be truncated before compilation", pattern->str);
951 if (job_list == NULL)
957 lastjob->next = dummy;
967 /*********************************************************************
969 * Function : rewrite_url
971 * Description : Rewrites a URL with a single pcrs command
972 * and returns the result if it differs from the
973 * original and isn't obviously invalid.
976 * 1 : old_url = URL to rewrite.
977 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
980 * Returns : NULL if the pcrs_command didn't change the url, or
981 * the result of the modification.
983 *********************************************************************/
984 char *rewrite_url(char *old_url, const char *pcrs_command)
986 char *new_url = NULL;
990 assert(pcrs_command);
992 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
996 log_error(LOG_LEVEL_REDIRECTS,
997 "pcrs command \"%s\" didn't change \"%s\".",
998 pcrs_command, old_url);
1003 log_error(LOG_LEVEL_REDIRECTS,
1004 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1005 pcrs_command, old_url, pcrs_strerror(hits));
1008 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1010 log_error(LOG_LEVEL_ERROR,
1011 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1012 "but the result doesn't look like a valid URL and will be ignored.",
1013 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1018 log_error(LOG_LEVEL_REDIRECTS,
1019 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1020 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1028 #ifdef FEATURE_FAST_REDIRECTS
1029 /*********************************************************************
1031 * Function : get_last_url
1033 * Description : Search for the last URL inside a string.
1034 * If the string already is a URL, it will
1035 * be the first URL found.
1038 * 1 : subject = the string to check
1039 * 2 : redirect_mode = +fast-redirect{} mode
1041 * Returns : NULL if no URL was found, or
1042 * the last URL found.
1044 *********************************************************************/
1045 char *get_last_url(char *subject, const char *redirect_mode)
1047 char *new_url = NULL;
1051 assert(redirect_mode);
1053 subject = strdup(subject);
1054 if (subject == NULL)
1056 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1060 if (0 == strcmpic(redirect_mode, "check-decoded-url") && strchr(subject, '%'))
1062 char *url_segment = NULL;
1063 char **url_segments;
1064 size_t max_segments;
1067 log_error(LOG_LEVEL_REDIRECTS,
1068 "Checking \"%s\" for encoded redirects.", subject);
1071 * Check each parameter in the URL separately.
1072 * Sectionize the URL at "?" and "&",
1073 * go backwards through the segments, URL-decode them
1074 * and look for a URL in the decoded result.
1075 * Stop the search after the first match.
1077 * XXX: This estimate is guaranteed to be high enough as we
1078 * let ssplit() ignore empty fields, but also a bit wasteful.
1080 max_segments = strlen(subject) / 2;
1081 url_segments = malloc(max_segments * sizeof(char *));
1083 if (NULL == url_segments)
1085 log_error(LOG_LEVEL_ERROR,
1086 "Out of memory while decoding URL: %s", subject);
1091 segments = ssplit(subject, "?&", url_segments, max_segments);
1093 while (segments-- > 0)
1095 char *dtoken = url_decode(url_segments[segments]);
1098 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", url_segments[segments]);
1101 url_segment = strstr(dtoken, "http://");
1102 if (NULL == url_segment)
1104 url_segment = strstr(dtoken, "https://");
1106 if (NULL != url_segment)
1108 url_segment = strdup_or_die(url_segment);
1115 freez(url_segments);
1117 if (url_segment == NULL)
1121 subject = url_segment;
1125 /* Look for a URL inside this one, without decoding anything. */
1126 log_error(LOG_LEVEL_REDIRECTS,
1127 "Checking \"%s\" for unencoded redirects.", subject);
1131 * Find the last URL encoded in the request
1134 while ((tmp = strstr(tmp, "http://")) != NULL)
1138 tmp = (new_url != NULL) ? new_url : subject;
1139 while ((tmp = strstr(tmp, "https://")) != NULL)
1144 if ((new_url != NULL)
1145 && ( (new_url != subject)
1146 || (0 == strncmpic(subject, "http://", 7))
1147 || (0 == strncmpic(subject, "https://", 8))
1151 * Return new URL if we found a redirect
1152 * or if the subject already was a URL.
1154 * The second case makes sure that we can
1155 * chain get_last_url after another redirection check
1156 * (like rewrite_url) without losing earlier redirects.
1158 new_url = strdup(new_url);
1167 #endif /* def FEATURE_FAST_REDIRECTS */
1170 /*********************************************************************
1172 * Function : redirect_url
1174 * Description : Checks if Privoxy should answer the request with
1175 * a HTTP redirect and generates the redirect if
1179 * 1 : csp = Current client state (buffers, headers, etc...)
1181 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1183 *********************************************************************/
1184 struct http_response *redirect_url(struct client_state *csp)
1186 struct http_response *rsp;
1187 #ifdef FEATURE_FAST_REDIRECTS
1189 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1190 * as compile-time option? The user can easily disable
1191 * it in his action file.
1193 char * redirect_mode;
1194 #endif /* def FEATURE_FAST_REDIRECTS */
1195 char *old_url = NULL;
1196 char *new_url = NULL;
1197 char *redirection_string;
1199 if ((csp->action->flags & ACTION_REDIRECT))
1201 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1204 * If the redirection string begins with 's',
1205 * assume it's a pcrs command, otherwise treat it as
1206 * properly formatted URL and use it for the redirection
1209 * According to (the now obsolete) RFC 2616 section 14.30
1210 * the URL has to be absolute and if the user tries:
1211 * +redirect{sadly/this/will/be/parsed/as/pcrs_command.html}
1212 * she would get undefined results anyway.
1214 * RFC 7231 7.1.2 actually allows relative references,
1215 * but those start with a leading slash (RFC 3986 4.2) and
1216 * thus can't be mistaken for pcrs commands either.
1219 if (*redirection_string == 's')
1221 old_url = csp->http->url;
1222 new_url = rewrite_url(old_url, redirection_string);
1226 log_error(LOG_LEVEL_REDIRECTS,
1227 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1228 redirection_string);
1229 new_url = strdup(redirection_string);
1233 #ifdef FEATURE_FAST_REDIRECTS
1234 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1236 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1239 * If it exists, use the previously rewritten URL as input
1240 * otherwise just use the old path.
1242 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1243 new_url = get_last_url(old_url, redirect_mode);
1248 * Disable redirect checkers, so that they
1249 * will be only run more than once if the user
1250 * also enables them through tags.
1252 * From a performance point of view
1253 * it doesn't matter, but the duplicated
1254 * log messages are annoying.
1256 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1257 #endif /* def FEATURE_FAST_REDIRECTS */
1258 csp->action->flags &= ~ACTION_REDIRECT;
1260 /* Did any redirect action trigger? */
1263 if (url_requires_percent_encoding(new_url))
1266 log_error(LOG_LEVEL_REDIRECTS, "Percent-encoding redirect URL: %N",
1267 strlen(new_url), new_url);
1268 encoded_url = percent_encode_url(new_url);
1270 if (encoded_url == NULL)
1272 return cgi_error_memory();
1274 new_url = encoded_url;
1275 assert(FALSE == url_requires_percent_encoding(new_url));
1278 if (0 == strcmpic(new_url, csp->http->url))
1280 log_error(LOG_LEVEL_ERROR,
1281 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1282 csp->http->url, new_url);
1287 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1289 if (NULL == (rsp = alloc_http_response()))
1292 return cgi_error_memory();
1295 rsp->status = strdup_or_die("302 Local Redirect from Privoxy");
1296 if (enlist_unique_header(rsp->headers, "Location", new_url))
1299 free_http_response(rsp);
1300 return cgi_error_memory();
1302 rsp->crunch_reason = REDIRECTED;
1305 return finish_http_response(csp, rsp);
1309 /* Only reached if no redirect is required */
1315 #ifdef FEATURE_IMAGE_BLOCKING
1316 /*********************************************************************
1318 * Function : is_imageurl
1320 * Description : Given a URL, decide whether it should be treated
1321 * as image URL or not.
1324 * 1 : csp = Current client state (buffers, headers, etc...)
1326 * Returns : True (nonzero) if URL is an image URL, false (0)
1329 *********************************************************************/
1330 int is_imageurl(const struct client_state *csp)
1332 return ((csp->action->flags & ACTION_IMAGE) != 0);
1335 #endif /* def FEATURE_IMAGE_BLOCKING */
1338 #ifdef FEATURE_TRUST
1339 /*********************************************************************
1341 * Function : is_untrusted_url
1343 * Description : Should we "distrust" this URL (and block it)?
1345 * Yes if it matches a line in the trustfile, or if the
1346 * referrer matches a line starting with "+" in the
1351 * 1 : csp = Current client state (buffers, headers, etc...)
1353 * Returns : 0 => trusted, 1 => untrusted
1355 *********************************************************************/
1356 int is_untrusted_url(const struct client_state *csp)
1358 struct file_list *fl;
1359 struct block_spec *b;
1360 struct pattern_spec **trusted_url;
1361 struct http_request rhttp[1];
1362 const char * referer;
1366 * If we don't have a trustlist, we trust everybody
1368 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1373 memset(rhttp, '\0', sizeof(*rhttp));
1376 * Do we trust the request URL itself?
1378 for (b = b->next; b ; b = b->next)
1380 if (url_match(b->url, csp->http))
1386 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1388 /* no referrer was supplied */
1394 * If not, do we maybe trust its referrer?
1396 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
1402 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1404 if (url_match(*trusted_url, rhttp))
1406 /* if the URL's referrer is from a trusted referrer, then
1407 * add the target spec to the trustfile as an unblocked
1408 * domain and return 0 (which means it's OK).
1413 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1417 char * new_entry = strdup_or_die("~");
1419 string_append(&new_entry, csp->http->hostport);
1421 path = csp->http->path;
1422 if ( (path[0] == '/')
1424 && ((path_end = strchr(path + 2, '/')) != NULL))
1426 /* since this path points into a user's home space
1427 * be sure to include this spec in the trustfile.
1429 long path_len = path_end - path; /* save offset */
1430 path = strdup(path); /* Copy string */
1433 path_end = path + path_len; /* regenerate ptr to new buffer */
1434 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1436 string_join(&new_entry, path);
1440 * Give a reason for generating this entry.
1442 string_append(&new_entry, " # Trusted referrer was: ");
1443 string_append(&new_entry, referer);
1445 if (new_entry != NULL)
1447 if (-1 == fprintf(fp, "%s\n", new_entry))
1449 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1450 new_entry, csp->config->trustfile);
1456 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1457 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1464 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1465 csp->http->hostport, csp->config->trustfile);
1473 #endif /* def FEATURE_TRUST */
1476 /*********************************************************************
1478 * Function : get_filter
1480 * Description : Get a filter with a given name and type.
1481 * Note that taggers are filters, too.
1484 * 1 : csp = Current client state (buffers, headers, etc...)
1485 * 2 : requested_name = Name of the content filter to get
1486 * 3 : requested_type = Type of the filter to tagger to lookup
1488 * Returns : A pointer to the requested filter
1489 * or NULL if the filter wasn't found
1491 *********************************************************************/
1492 struct re_filterfile_spec *get_filter(const struct client_state *csp,
1493 const char *requested_name,
1494 enum filter_type requested_type)
1497 struct re_filterfile_spec *b;
1498 struct file_list *fl;
1500 for (i = 0; i < MAX_AF_FILES; i++)
1503 if ((NULL == fl) || (NULL == fl->f))
1506 * Either there are no filter files left or this
1507 * filter file just contains no valid filters.
1509 * Continue to be sure we don't miss valid filter
1510 * files that are chained after empty or invalid ones.
1515 for (b = fl->f; b != NULL; b = b->next)
1517 if (b->type != requested_type)
1519 /* The callers isn't interested in this filter type. */
1522 if (strcmp(b->name, requested_name) == 0)
1524 /* The requested filter has been found. Abort search. */
1530 /* No filter with the given name and type exists. */
1536 /*********************************************************************
1538 * Function : pcrs_filter_response
1540 * Description : Execute all text substitutions from all applying
1541 * +filter actions on the text buffer that's been
1542 * accumulated in csp->iob->buf.
1545 * 1 : csp = Current client state (buffers, headers, etc...)
1547 * Returns : a pointer to the (newly allocated) modified buffer.
1548 * or NULL if there were no hits or something went wrong
1550 *********************************************************************/
1551 static char *pcrs_filter_response(struct client_state *csp)
1554 size_t size, prev_size;
1560 struct re_filterfile_spec *b;
1561 struct list_entry *filtername;
1566 if (csp->iob->cur >= csp->iob->eod)
1571 if (filters_available(csp) == FALSE)
1573 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1574 "content filtering enabled, but no content filters available.");
1578 size = (size_t)(csp->iob->eod - csp->iob->cur);
1579 old = csp->iob->cur;
1582 * For all applying +filter actions, look if a filter by that
1583 * name exists and if yes, execute it's pcrs_joblist on the
1586 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1587 filtername != NULL; filtername = filtername->next)
1589 int current_hits = 0; /* Number of hits caused by this filter */
1590 int job_number = 0; /* Which job we're currently executing */
1591 int job_hits = 0; /* How many hits the current job caused */
1594 b = get_filter(csp, filtername->str, FT_CONTENT_FILTER);
1600 joblist = b->joblist;
1602 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1604 if (NULL == joblist)
1606 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1611 /* Apply all jobs from the joblist */
1612 for (job = joblist; NULL != job; job = job->next)
1615 job_hits = pcrs_execute(job, old, size, &new, &size);
1620 * That went well. Continue filtering
1621 * and use the result of this job as
1622 * input for the next one.
1624 current_hits += job_hits;
1625 if (old != csp->iob->cur)
1634 * This job caused an unexpected error. Inform the user
1635 * and skip the rest of the jobs in this filter. We could
1636 * continue with the next job, but usually the jobs
1637 * depend on each other or are similar enough to
1638 * fail for the same reason.
1640 * At the moment our pcrs expects the error codes of pcre 3.4,
1641 * but newer pcre versions can return additional error codes.
1642 * As a result pcrs_strerror()'s error message might be
1643 * "Unknown error ...", therefore we print the numerical value
1646 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1647 * should we use LOG_LEVEL_RE_FILTER instead?
1649 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1650 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1655 if (b->dynamic) pcrs_free_joblist(joblist);
1657 log_error(LOG_LEVEL_RE_FILTER,
1658 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
1659 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1661 hits += current_hits;
1665 * If there were no hits, destroy our copy and let
1666 * chat() use the original in csp->iob
1674 csp->flags |= CSP_FLAG_MODIFIED;
1675 csp->content_length = size;
1676 clear_iob(csp->iob);
1683 #ifdef FEATURE_EXTERNAL_FILTERS
1684 /*********************************************************************
1686 * Function : get_external_filter
1688 * Description : Lookup the code to execute for an external filter.
1689 * Masks the misuse of the re_filterfile_spec.
1692 * 1 : csp = Current client state (buffers, headers, etc...)
1693 * 2 : name = Name of the content filter to get
1695 * Returns : A pointer to the requested code
1696 * or NULL if the filter wasn't found
1698 *********************************************************************/
1699 static const char *get_external_filter(const struct client_state *csp,
1702 struct re_filterfile_spec *external_filter;
1704 external_filter = get_filter(csp, name, FT_EXTERNAL_CONTENT_FILTER);
1705 if (external_filter == NULL)
1707 log_error(LOG_LEVEL_FATAL,
1708 "Didn't find stuff to execute for external filter: %s",
1712 return external_filter->patterns->first->str;
1717 /*********************************************************************
1719 * Function : set_privoxy_variables
1721 * Description : Sets a couple of privoxy-specific environment variables
1724 * 1 : csp = Current client state (buffers, headers, etc...)
1728 *********************************************************************/
1729 static void set_privoxy_variables(const struct client_state *csp)
1736 { "PRIVOXY_URL", csp->http->url },
1737 { "PRIVOXY_PATH", csp->http->path },
1738 { "PRIVOXY_HOST", csp->http->host },
1739 { "PRIVOXY_ORIGIN", csp->ip_addr_str },
1740 { "PRIVOXY_LISTEN_ADDRESS", csp->listen_addr_str },
1743 for (i = 0; i < SZ(env); i++)
1745 if (setenv(env[i].name, env[i].value, 1))
1747 log_error(LOG_LEVEL_ERROR, "Failed to set %s=%s: %E",
1748 env[i].name, env[i].value);
1754 /*********************************************************************
1756 * Function : execute_external_filter
1758 * Description : Pipe content into external filter and return the output
1761 * 1 : csp = Current client state (buffers, headers, etc...)
1762 * 2 : name = Name of the external filter to execute
1763 * 3 : content = The original content to filter
1764 * 4 : size = The size of the content buffer
1766 * Returns : a pointer to the (newly allocated) modified buffer.
1767 * or NULL if there were no hits or something went wrong
1769 *********************************************************************/
1770 static char *execute_external_filter(const struct client_state *csp,
1771 const char *name, char *content, size_t *size)
1774 char file_name[FILENAME_MAX];
1776 char *filter_output;
1780 const char *external_filter;
1782 if (csp->config->temporary_directory == NULL)
1784 log_error(LOG_LEVEL_ERROR,
1785 "No temporary-directory configured. Can't execute filter: %s",
1790 external_filter = get_external_filter(csp, name);
1792 if (sizeof(file_name) < snprintf(file_name, sizeof(file_name),
1793 "%s/privoxy-XXXXXXXX", csp->config->temporary_directory))
1795 log_error(LOG_LEVEL_ERROR, "temporary-directory path too long");
1799 fd = mkstemp(file_name);
1802 log_error(LOG_LEVEL_ERROR, "mkstemp() failed to create %s: %E", file_name);
1806 fp = fdopen(fd, "w");
1809 log_error(LOG_LEVEL_ERROR, "fdopen() failed: %E");
1815 * The size may be zero if a previous filter discarded everything.
1817 * This isn't necessary unintentional, so we just don't try
1818 * to fwrite() nothing and let the user deal with the rest.
1820 if ((*size != 0) && fwrite(content, *size, 1, fp) != 1)
1822 log_error(LOG_LEVEL_ERROR, "fwrite(..., %d, 1, ..) failed: %E", *size);
1829 if (sizeof(cmd) < snprintf(cmd, sizeof(cmd), "%s < %s", external_filter, file_name))
1831 log_error(LOG_LEVEL_ERROR,
1832 "temporary-directory or external filter path too long");
1837 log_error(LOG_LEVEL_RE_FILTER, "Executing '%s': %s", name, cmd);
1840 * The locking is necessary to prevent other threads
1841 * from overwriting the environment variables before
1842 * the popen fork. Afterwards this no longer matters.
1844 privoxy_mutex_lock(&external_filter_mutex);
1845 set_privoxy_variables(csp);
1846 fp = popen(cmd, "r");
1847 privoxy_mutex_unlock(&external_filter_mutex);
1850 log_error(LOG_LEVEL_ERROR, "popen(\"%s\", \"r\") failed: %E", cmd);
1855 /* Allocate at least one byte */
1856 filter_output = malloc_or_die(*size + 1);
1859 while (!feof(fp) && !ferror(fp))
1862 /* Could be bigger ... */
1863 enum { READ_LENGTH = 2048 };
1865 if (new_size + READ_LENGTH >= *size)
1869 /* Could be considered wasteful if the content is 'large'. */
1870 *size += (*size >= READ_LENGTH) ? *size : READ_LENGTH;
1872 p = realloc(filter_output, *size);
1875 log_error(LOG_LEVEL_ERROR, "Out of memory while reading "
1876 "external filter output. Using what we got so far.");
1881 assert(new_size + READ_LENGTH < *size);
1882 len = fread(&filter_output[new_size], 1, READ_LENGTH, fp);
1892 log_error(LOG_LEVEL_ERROR, "Executing %s failed: %E", cmd);
1896 log_error(LOG_LEVEL_RE_FILTER,
1897 "Executing '%s' resulted in return value %d. "
1898 "Read %d of up to %d bytes.", name, (ret >> 8), new_size, *size);
1904 return filter_output;
1907 #endif /* def FEATURE_EXTERNAL_FILTERS */
1910 /*********************************************************************
1912 * Function : gif_deanimate_response
1914 * Description : Deanimate the GIF image that has been accumulated in
1915 * csp->iob->buf, set csp->content_length to the modified
1916 * size and raise the CSP_FLAG_MODIFIED flag.
1919 * 1 : csp = Current client state (buffers, headers, etc...)
1921 * Returns : a pointer to the (newly allocated) modified buffer.
1922 * or NULL in case something went wrong.
1924 *********************************************************************/
1926 char *gif_deanimate_response(struct client_state *csp)
1928 static char *gif_deanimate_response(struct client_state *csp)
1931 struct binbuffer *in, *out;
1935 size = (size_t)(csp->iob->eod - csp->iob->cur);
1937 in = zalloc_or_die(sizeof(*in));
1938 out = zalloc_or_die(sizeof(*out));
1940 in->buffer = csp->iob->cur;
1943 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1945 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1952 if ((int)size == out->offset)
1954 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1958 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1960 csp->content_length = out->offset;
1961 csp->flags |= CSP_FLAG_MODIFIED;
1971 /*********************************************************************
1973 * Function : get_filter_function
1975 * Description : Decides which content filter function has
1976 * to be applied (if any). Only considers functions
1977 * for internal filters which are mutually-exclusive.
1980 * 1 : csp = Current client state (buffers, headers, etc...)
1982 * Returns : The content filter function to run, or
1983 * NULL if no content filter is active
1985 *********************************************************************/
1986 static filter_function_ptr get_filter_function(const struct client_state *csp)
1988 filter_function_ptr filter_function = NULL;
1991 * Choose the applying filter function based on
1992 * the content type and action settings.
1994 if ((csp->content_type & CT_TEXT) &&
1995 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
1997 filter_function = pcrs_filter_response;
1999 else if ((csp->content_type & CT_GIF) &&
2000 (csp->action->flags & ACTION_DEANIMATE))
2002 filter_function = gif_deanimate_response;
2005 return filter_function;
2009 /*********************************************************************
2011 * Function : remove_chunked_transfer_coding
2013 * Description : In-situ remove the "chunked" transfer coding as defined
2014 * in RFC 7230 4.1 from a buffer. XXX: The implementation
2015 * is neither complete nor compliant (TODO #129).
2018 * 1 : buffer = Pointer to the text buffer
2019 * 2 : size = In: Number of bytes to be processed,
2020 * Out: Number of bytes after de-chunking.
2021 * (undefined in case of errors)
2023 * Returns : JB_ERR_OK for success,
2024 * JB_ERR_PARSE otherwise
2026 *********************************************************************/
2028 extern jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2030 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2034 unsigned int chunksize = 0;
2035 char *from_p, *to_p;
2036 const char *end_of_buffer = buffer + *size;
2040 log_error(LOG_LEVEL_FATAL, "Invalid chunked input. Buffer is empty.");
2041 return JB_ERR_PARSE;
2045 from_p = to_p = buffer;
2047 if (sscanf(buffer, "%x", &chunksize) != 1)
2049 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2050 return JB_ERR_PARSE;
2053 while (chunksize > 0U)
2056 * If the chunk-size is valid, we should have at least
2057 * chunk-size bytes of chunk-data and five bytes of
2058 * meta data (chunk-size, CRLF, CRLF) left in the buffer.
2060 if (chunksize + 5 >= *size - newsize)
2062 log_error(LOG_LEVEL_ERROR,
2063 "Chunk size %u exceeds buffered data left. "
2064 "Already digested %u of %u buffered bytes.",
2065 chunksize, (unsigned int)newsize, (unsigned int)*size);
2066 return JB_ERR_PARSE;
2070 * Skip the chunk-size, the optional chunk-ext and the CRLF
2071 * that is supposed to be located directly before the start
2074 if (NULL == (from_p = strstr(from_p, "\r\n")))
2076 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2077 return JB_ERR_PARSE;
2082 * The previous strstr() does not enforce chunk-validity
2083 * and is sattisfied as long a CRLF is left in the buffer.
2085 * Make sure the bytes we consider chunk-data are within
2088 if (from_p + chunksize >= end_of_buffer)
2090 log_error(LOG_LEVEL_ERROR,
2091 "End of chunk is beyond the end of the buffer.");
2092 return JB_ERR_PARSE;
2095 memmove(to_p, from_p, (size_t) chunksize);
2096 newsize += chunksize;
2097 to_p = buffer + newsize;
2098 from_p += chunksize;
2101 * Not merging this check with the previous one allows us
2102 * to keep chunks without trailing CRLF. It's not clear
2103 * if we actually have to care about those, though.
2105 if (from_p + 2 >= end_of_buffer)
2107 log_error(LOG_LEVEL_ERROR, "Not enough room for trailing CRLF.");
2108 return JB_ERR_PARSE;
2111 if (sscanf(from_p, "%x", &chunksize) != 1)
2113 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
2118 /* XXX: Should get its own loglevel. */
2119 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
2128 /*********************************************************************
2130 * Function : prepare_for_filtering
2132 * Description : If necessary, de-chunks and decompresses
2133 * the content so it can get filterd.
2136 * 1 : csp = Current client state (buffers, headers, etc...)
2138 * Returns : JB_ERR_OK for success,
2139 * JB_ERR_PARSE otherwise
2141 *********************************************************************/
2142 static jb_err prepare_for_filtering(struct client_state *csp)
2144 jb_err err = JB_ERR_OK;
2147 * If the body has a "chunked" transfer-encoding,
2148 * get rid of it, adjusting size and iob->eod
2150 if (csp->flags & CSP_FLAG_CHUNKED)
2152 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2154 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2155 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2156 if (JB_ERR_OK == err)
2158 csp->iob->eod = csp->iob->cur + size;
2159 csp->flags |= CSP_FLAG_MODIFIED;
2163 return JB_ERR_PARSE;
2169 * If the body has a supported transfer-encoding,
2170 * decompress it, adjusting size and iob->eod.
2172 if (csp->content_type & (CT_GZIP|CT_DEFLATE))
2174 if (0 == csp->iob->eod - csp->iob->cur)
2176 /* Nothing left after de-chunking. */
2180 err = decompress_iob(csp);
2182 if (JB_ERR_OK == err)
2184 csp->flags |= CSP_FLAG_MODIFIED;
2185 csp->content_type &= ~CT_TABOO;
2190 * Unset CT_GZIP and CT_DEFLATE to remember not
2191 * to modify the Content-Encoding header later.
2193 csp->content_type &= ~CT_GZIP;
2194 csp->content_type &= ~CT_DEFLATE;
2203 /*********************************************************************
2205 * Function : execute_content_filters
2207 * Description : Executes a given content filter.
2210 * 1 : csp = Current client state (buffers, headers, etc...)
2212 * Returns : Pointer to the modified buffer, or
2213 * NULL if filtering failed or wasn't necessary.
2215 *********************************************************************/
2216 char *execute_content_filters(struct client_state *csp)
2219 filter_function_ptr content_filter;
2221 assert(content_filters_enabled(csp->action));
2223 if (0 == csp->iob->eod - csp->iob->cur)
2226 * No content (probably status code 301, 302 ...),
2227 * no filtering necessary.
2232 if (JB_ERR_OK != prepare_for_filtering(csp))
2235 * failed to de-chunk or decompress.
2240 if (0 == csp->iob->eod - csp->iob->cur)
2243 * Clown alarm: chunked and/or compressed nothing delivered.
2248 content_filter = get_filter_function(csp);
2249 content = (content_filter != NULL) ? (*content_filter)(csp) : NULL;
2251 #ifdef FEATURE_EXTERNAL_FILTERS
2252 if ((csp->content_type & CT_TEXT) &&
2253 !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]))
2255 struct list_entry *filtername;
2256 size_t size = (size_t)csp->content_length;
2258 if (content == NULL)
2260 content = csp->iob->cur;
2261 size = (size_t)(csp->iob->eod - csp->iob->cur);
2264 for (filtername = csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]->first;
2265 filtername ; filtername = filtername->next)
2267 char *result = execute_external_filter(csp, filtername->str, content, &size);
2270 if (content != csp->iob->cur)
2277 csp->flags |= CSP_FLAG_MODIFIED;
2278 csp->content_length = size;
2280 #endif /* def FEATURE_EXTERNAL_FILTERS */
2287 /*********************************************************************
2289 * Function : get_url_actions
2291 * Description : Gets the actions for this URL.
2294 * 1 : csp = Current client state (buffers, headers, etc...)
2295 * 2 : http = http_request request for blocked URLs
2299 *********************************************************************/
2300 void get_url_actions(struct client_state *csp, struct http_request *http)
2302 struct file_list *fl;
2303 struct url_actions *b;
2306 init_current_action(csp->action);
2308 for (i = 0; i < MAX_AF_FILES; i++)
2310 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2315 #ifdef FEATURE_CLIENT_TAGS
2316 apply_url_actions(csp->action, http, csp->client_tags, b);
2318 apply_url_actions(csp->action, http, b);
2325 /*********************************************************************
2327 * Function : apply_url_actions
2329 * Description : Applies a list of URL actions.
2332 * 1 : action = Destination.
2333 * 2 : http = Current URL
2334 * 3 : client_tags = list of client tags
2335 * 4 : b = list of URL actions to apply
2339 *********************************************************************/
2340 static void apply_url_actions(struct current_action_spec *action,
2341 struct http_request *http,
2342 #ifdef FEATURE_CLIENT_TAGS
2343 const struct list *client_tags,
2345 struct url_actions *b)
2349 /* Should never happen */
2353 for (b = b->next; NULL != b; b = b->next)
2355 if (url_match(b->url, http))
2357 merge_current_action(action, b->action);
2359 #ifdef FEATURE_CLIENT_TAGS
2360 if (client_tag_match(b->url, client_tags))
2362 merge_current_action(action, b->action);
2369 /*********************************************************************
2371 * Function : get_forward_override_settings
2373 * Description : Returns forward settings as specified with the
2374 * forward-override{} action. forward-override accepts
2375 * forward lines similar to the one used in the
2376 * configuration file, but without the URL pattern.
2382 * in the configuration file can be replaced with
2383 * the action section:
2385 * {+forward-override{forward .}}
2389 * 1 : csp = Current client state (buffers, headers, etc...)
2391 * Returns : Pointer to forwarding structure in case of success.
2392 * Invalid syntax is fatal.
2394 *********************************************************************/
2395 static const struct forward_spec *get_forward_override_settings(struct client_state *csp)
2397 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2398 char forward_settings[BUFFER_SIZE];
2399 char *http_parent = NULL;
2400 /* variable names were chosen for consistency reasons. */
2401 struct forward_spec *fwd = NULL;
2405 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2406 /* Should be enforced by load_one_actions_file() */
2407 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2409 /* Create a copy ssplit can modify */
2410 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2412 if (NULL != csp->fwd)
2415 * XXX: Currently necessary to prevent memory
2416 * leaks when the show-url-info cgi page is visited.
2418 unload_forward_spec(csp->fwd);
2422 * allocate a new forward node, valid only for
2423 * the lifetime of this request. Save its location
2424 * in csp as well, so sweep() can free it later on.
2426 fwd = csp->fwd = zalloc_or_die(sizeof(*fwd));
2428 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec));
2429 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2431 fwd->type = SOCKS_NONE;
2433 /* Parse the parent HTTP proxy host:port */
2434 http_parent = vec[1];
2437 else if ((vec_count == 2) && !strcasecmp(vec[0], "forward-webserver"))
2439 fwd->type = FORWARD_WEBSERVER;
2441 /* Parse the parent HTTP server host:port */
2442 http_parent = vec[1];
2445 else if (vec_count == 3)
2447 char *socks_proxy = NULL;
2449 if (!strcasecmp(vec[0], "forward-socks4"))
2451 fwd->type = SOCKS_4;
2452 socks_proxy = vec[1];
2454 else if (!strcasecmp(vec[0], "forward-socks4a"))
2456 fwd->type = SOCKS_4A;
2457 socks_proxy = vec[1];
2459 else if (!strcasecmp(vec[0], "forward-socks5"))
2461 fwd->type = SOCKS_5;
2462 socks_proxy = vec[1];
2464 else if (!strcasecmp(vec[0], "forward-socks5t"))
2466 fwd->type = SOCKS_5T;
2467 socks_proxy = vec[1];
2470 if (NULL != socks_proxy)
2472 /* Parse the SOCKS proxy host[:port] */
2473 fwd->gateway_port = 1080;
2474 parse_forwarder_address(socks_proxy,
2475 &fwd->gateway_host, &fwd->gateway_port);
2477 http_parent = vec[2];
2481 if (NULL == http_parent)
2483 log_error(LOG_LEVEL_FATAL,
2484 "Invalid forward-override syntax in: %s", forward_override_line);
2485 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2488 /* Parse http forwarding settings */
2489 if (strcmp(http_parent, ".") != 0)
2491 fwd->forward_port = 8000;
2492 parse_forwarder_address(http_parent,
2493 &fwd->forward_host, &fwd->forward_port);
2496 assert (NULL != fwd);
2498 log_error(LOG_LEVEL_CONNECT,
2499 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2505 /*********************************************************************
2507 * Function : forward_url
2509 * Description : Should we forward this to another proxy?
2512 * 1 : csp = Current client state (buffers, headers, etc...)
2513 * 2 : http = http_request request for current URL
2515 * Returns : Pointer to forwarding information.
2517 *********************************************************************/
2518 const struct forward_spec *forward_url(struct client_state *csp,
2519 const struct http_request *http)
2521 static const struct forward_spec fwd_default[1]; /* Zero'ed due to being static. */
2522 struct forward_spec *fwd = csp->config->forward;
2524 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2526 return get_forward_override_settings(csp);
2536 if (url_match(fwd->url, http))
2547 /*********************************************************************
2549 * Function : direct_response
2551 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2552 * request and if so, return a HTTP 501 to the client.
2554 * FIXME: I have a stupid name and I should handle the
2555 * requests properly. Still, what we do here is rfc-
2556 * compliant, whereas ignoring or forwarding are not.
2559 * 1 : csp = Current client state (buffers, headers, etc...)
2561 * Returns : http_response if , NULL if nonmatch or handler fail
2563 *********************************************************************/
2564 struct http_response *direct_response(struct client_state *csp)
2566 struct http_response *rsp;
2567 struct list_entry *p;
2569 if ((0 == strcmpic(csp->http->gpc, "trace"))
2570 || (0 == strcmpic(csp->http->gpc, "options")))
2572 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2574 if (!strncmpic(p->str, "Max-Forwards:", 13))
2576 unsigned int max_forwards;
2579 * If it's a Max-Forwards value of zero,
2580 * we have to intercept the request.
2582 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2585 * FIXME: We could handle at least TRACE here,
2586 * but that would require a verbatim copy of
2587 * the request which we don't have anymore
2589 log_error(LOG_LEVEL_HEADER,
2590 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2593 /* Get mem for response or fail*/
2594 if (NULL == (rsp = alloc_http_response()))
2596 return cgi_error_memory();
2599 rsp->status = strdup_or_die("501 Not Implemented");
2601 rsp->crunch_reason = UNSUPPORTED;
2603 return(finish_http_response(csp, rsp));
2612 /*********************************************************************
2614 * Function : content_requires_filtering
2616 * Description : Checks whether there are any content filters
2617 * enabled for the current request and if they
2618 * can actually be applied..
2621 * 1 : csp = Current client state (buffers, headers, etc...)
2623 * Returns : TRUE for yes, FALSE otherwise
2625 *********************************************************************/
2626 int content_requires_filtering(struct client_state *csp)
2628 if ((csp->content_type & CT_TABOO)
2629 && !(csp->action->flags & ACTION_FORCE_TEXT_MODE))
2635 * Are we enabling text mode by force?
2637 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2640 * Do we really have to?
2642 if (csp->content_type & CT_TEXT)
2644 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2648 csp->content_type |= CT_TEXT;
2649 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2653 if (!(csp->content_type & CT_DECLARED))
2656 * The server didn't bother to declare a MIME-Type.
2657 * Assume it's text that can be filtered.
2659 * This also regulary happens with 304 responses,
2660 * therefore logging anything here would cause
2663 csp->content_type |= CT_TEXT;
2667 * Choose the applying filter function based on
2668 * the content type and action settings.
2670 if ((csp->content_type & CT_TEXT) &&
2671 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER]) ||
2672 !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER])))
2676 else if ((csp->content_type & CT_GIF) &&
2677 (csp->action->flags & ACTION_DEANIMATE))
2687 /*********************************************************************
2689 * Function : content_filters_enabled
2691 * Description : Checks whether there are any content filters
2692 * enabled for the current request.
2695 * 1 : action = Action spec to check.
2697 * Returns : TRUE for yes, FALSE otherwise
2699 *********************************************************************/
2700 int content_filters_enabled(const struct current_action_spec *action)
2702 return ((action->flags & ACTION_DEANIMATE) ||
2703 !list_is_empty(action->multi[ACTION_MULTI_FILTER]) ||
2704 !list_is_empty(action->multi[ACTION_MULTI_EXTERNAL_FILTER]));
2708 /*********************************************************************
2710 * Function : filters_available
2712 * Description : Checks whether there are any filters available.
2715 * 1 : csp = Current client state (buffers, headers, etc...)
2717 * Returns : TRUE for yes, FALSE otherwise.
2719 *********************************************************************/
2720 int filters_available(const struct client_state *csp)
2723 for (i = 0; i < MAX_AF_FILES; i++)
2725 const struct file_list *fl = csp->rlist[i];
2726 if ((NULL != fl) && (NULL != fl->f))