1 /*********************************************************************
3 * File : $Source: /cvsroot/ijbswa/current/filters.c,v $
5 * Purpose : Declares functions to parse/crunch headers and pages.
7 * Copyright : Written by and Copyright (C) 2001-2020 the
8 * Privoxy team. https://www.privoxy.org/
10 * Based on the Internet Junkbuster originally written
11 * by and Copyright (C) 1997 Anonymous Coders and
12 * Junkbusters Corporation. http://www.junkbusters.com
14 * This program is free software; you can redistribute it
15 * and/or modify it under the terms of the GNU General
16 * Public License as published by the Free Software
17 * Foundation; either version 2 of the License, or (at
18 * your option) any later version.
20 * This program is distributed in the hope that it will
21 * be useful, but WITHOUT ANY WARRANTY; without even the
22 * implied warranty of MERCHANTABILITY or FITNESS FOR A
23 * PARTICULAR PURPOSE. See the GNU General Public
24 * License for more details.
26 * The GNU General Public License should be included with
27 * this file. If not, you can view it at
28 * http://www.gnu.org/copyleft/gpl.html
29 * or write to the Free Software Foundation, Inc., 59
30 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
32 *********************************************************************/
38 #include <sys/types.h>
46 #include <netinet/in.h>
49 #endif /* ndef _WIN32 */
57 #include "jbsockets.h"
63 #include "deanimate.h"
66 #ifdef FEATURE_CLIENT_TAGS
67 #include "client-tags.h"
69 #ifdef FEATURE_HTTPS_INSPECTION
77 typedef char *(*filter_function_ptr)();
78 static filter_function_ptr get_filter_function(const struct client_state *csp);
79 static jb_err prepare_for_filtering(struct client_state *csp);
80 static void apply_url_actions(struct current_action_spec *action,
81 struct http_request *http,
82 #ifdef FEATURE_CLIENT_TAGS
83 const struct list *client_tags,
85 struct url_actions *b);
87 #ifdef FEATURE_EXTENDED_STATISTICS
88 static void increment_block_reason_counter(const char *block_reason);
93 /*********************************************************************
95 * Function : sockaddr_storage_to_ip
97 * Description : Access internal structure of sockaddr_storage
100 * 1 : addr = socket address
101 * 2 : ip = IP address as array of octets in network order
102 * (it points into addr)
103 * 3 : len = length of IP address in octets
104 * 4 : port = port number in network order;
108 *********************************************************************/
109 static void sockaddr_storage_to_ip(const struct sockaddr_storage *addr,
110 uint8_t **ip, unsigned int *len,
113 assert(NULL != addr);
114 assert(addr->ss_family == AF_INET || addr->ss_family == AF_INET6);
116 switch (addr->ss_family)
126 &(((struct sockaddr_in *)addr)->sin_addr.s_addr);
130 *port = &((struct sockaddr_in *)addr)->sin_port;
141 *ip = ((struct sockaddr_in6 *)addr)->sin6_addr.s6_addr;
145 *port = &((struct sockaddr_in6 *)addr)->sin6_port;
153 /*********************************************************************
155 * Function : match_sockaddr
157 * Description : Check whether address matches network (IP address and port)
160 * 1 : network = socket address of subnework
161 * 2 : netmask = network mask as socket address
162 * 3 : address = checked socket address against given network
164 * Returns : 0 = doesn't match; 1 = does match
166 *********************************************************************/
167 static int match_sockaddr(const struct sockaddr_storage *network,
168 const struct sockaddr_storage *netmask,
169 const struct sockaddr_storage *address)
171 uint8_t *network_addr, *netmask_addr, *address_addr;
172 unsigned int addr_len;
173 in_port_t *network_port, *netmask_port, *address_port;
176 if (network->ss_family != netmask->ss_family)
178 /* This should never happen */
179 assert(network->ss_family == netmask->ss_family);
180 log_error(LOG_LEVEL_FATAL, "Network and netmask differ in family.");
183 sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
184 sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
185 sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
187 /* Check for family */
188 if ((network->ss_family == AF_INET) && (address->ss_family == AF_INET6)
189 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr))
191 /* Map AF_INET6 V4MAPPED address into AF_INET */
195 else if ((network->ss_family == AF_INET6) && (address->ss_family == AF_INET)
196 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr))
198 /* Map AF_INET6 V4MAPPED network into AF_INET */
204 /* XXX: Port check is signaled in netmask */
205 if (*netmask_port && *network_port != *address_port)
210 /* TODO: Optimize by checking by words instead of octets */
211 for (i = 0; (i < addr_len) && netmask_addr[i]; i++)
213 if ((network_addr[i] & netmask_addr[i]) !=
214 (address_addr[i] & netmask_addr[i]))
222 #endif /* def HAVE_RFC2553 */
225 /*********************************************************************
227 * Function : block_acl
229 * Description : Block this request?
230 * Decide yes or no based on ACL file.
233 * 1 : dst = The proxy or gateway address this is going to.
234 * Or NULL to check all possible targets.
235 * 2 : csp = Current client state (buffers, headers, etc...)
236 * Also includes the client IP address.
238 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
240 *********************************************************************/
241 int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
243 struct access_control_list *acl = csp->config->acl;
245 /* if not using an access control list, then permit the connection */
251 /* search the list */
256 match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
258 (csp->ip_addr_long & acl->src->mask) == acl->src->addr
264 /* Just want to check if they have any access */
265 if (acl->action == ACL_PERMIT)
277 * XXX: An undefined acl->dst is full of zeros and should be
278 * considered a wildcard address. sockaddr_storage_to_ip()
279 * fails on such destinations because of unknown sa_familly
280 * (glibc only?). However this test is not portable.
282 * So, we signal the acl->dst is wildcard in wildcard_dst.
285 match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
287 ((dst->addr & acl->dst->mask) == acl->dst->addr)
288 && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
292 if (acl->action == ACL_PERMIT)
310 /*********************************************************************
312 * Function : acl_addr
314 * Description : Called from `load_config' to parse an ACL address.
317 * 1 : aspec = String specifying ACL address.
318 * 2 : aca = struct access_control_addr to fill in.
320 * Returns : 0 => Ok, everything else is an error.
322 *********************************************************************/
323 int acl_addr(const char *aspec, struct access_control_addr *aca)
327 struct addrinfo hints, *result;
329 in_port_t *mask_port;
330 unsigned int addr_len;
333 #endif /* def HAVE_RFC2553 */
335 char *acl_spec = NULL;
338 /* XXX: Depend on ai_family */
346 * Use a temporary acl spec copy so we can log
347 * the unmodified original in case of parse errors.
349 acl_spec = strdup_or_die(aspec);
351 if ((p = strchr(acl_spec, '/')) != NULL)
354 if (privoxy_isdigit(*p) == 0)
359 masklength = atoi(p);
362 if ((masklength < 0) ||
374 if ((*acl_spec == '[') && (NULL != (p = strchr(acl_spec, ']'))))
377 memmove(acl_spec, acl_spec + 1, (size_t)(p - acl_spec));
386 p = strchr(acl_spec, ':');
396 memset(&hints, 0, sizeof(struct addrinfo));
397 hints.ai_family = AF_UNSPEC;
398 hints.ai_socktype = SOCK_STREAM;
400 i = getaddrinfo(acl_spec, p, &hints, &result);
404 log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s",
405 acl_spec, p, gai_strerror(i));
411 /* TODO: Allow multihomed hostnames */
412 memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
413 freeaddrinfo(result);
419 port = strtol(p, &endptr, 10);
421 if (port <= 0 || port > 65535 || *endptr != '\0')
428 aca->port = (unsigned long)port;
430 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
433 if (aca->addr == INADDR_NONE)
435 /* XXX: This will be logged as parse error. */
438 #endif /* def HAVE_RFC2553 */
440 /* build the netmask */
442 /* Clip masklength according to current family. */
443 if ((aca->addr.ss_family == AF_INET) && (masklength > 32))
448 aca->mask.ss_family = aca->addr.ss_family;
449 sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port);
453 /* ACL contains a port number, check ports in the future. */
458 * XXX: This could be optimized to operate on whole words instead
459 * of octets (128-bit CPU could do it in one iteration).
462 * Octets after prefix can be omitted because of
463 * previous initialization to zeros.
465 for (i = 0; (i < addr_len) && masklength; i++)
475 * XXX: This assumes MSB of octet is on the left side.
476 * This should be true for all architectures or solved
479 mask_data[i] = (uint8_t)~((1 << (8 - masklength)) - 1);
486 for (i=1; i <= masklength ; i++)
488 aca->mask |= (1U << (32 - i));
491 /* now mask off the host portion of the ip address
492 * (i.e. save on the network portion of the address).
494 aca->addr = aca->addr & aca->mask;
495 #endif /* def HAVE_RFC2553 */
500 #endif /* def FEATURE_ACL */
503 /*********************************************************************
505 * Function : connect_port_is_forbidden
507 * Description : Check to see if CONNECT requests to the destination
508 * port of this request are forbidden. The check is
509 * independent of the actual request method.
512 * 1 : csp = Current client state (buffers, headers, etc...)
514 * Returns : True if yes, false otherwise.
516 *********************************************************************/
517 int connect_port_is_forbidden(const struct client_state *csp)
519 return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
520 !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
525 /*********************************************************************
527 * Function : block_url
529 * Description : Called from `chat'. Check to see if we need to block this.
532 * 1 : csp = Current client state (buffers, headers, etc...)
534 * Returns : NULL => unblocked, else HTTP block response
536 *********************************************************************/
537 struct http_response *block_url(struct client_state *csp)
539 struct http_response *rsp;
540 const char *new_content_type = NULL;
543 * If it's not blocked, don't block it ;-)
545 if ((csp->action->flags & ACTION_BLOCK) == 0)
549 if (csp->action->flags & ACTION_REDIRECT)
551 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
554 * Else, prepare a response
556 if (NULL == (rsp = alloc_http_response()))
558 return cgi_error_memory();
561 #ifdef FEATURE_EXTENDED_STATISTICS
562 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
564 increment_block_reason_counter(csp->action->string[ACTION_STRING_BLOCK]);
569 * If it's an image-url, send back an image or redirect
570 * as specified by the relevant +image action
572 #ifdef FEATURE_IMAGE_BLOCKING
573 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
577 /* determine HOW images should be blocked */
578 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
580 if (csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
582 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
585 /* and handle accordingly: */
586 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
588 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
589 rsp->body = bindup(image_pattern_data, image_pattern_length);
590 if (rsp->body == NULL)
592 free_http_response(rsp);
593 return cgi_error_memory();
595 rsp->content_length = image_pattern_length;
597 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
599 free_http_response(rsp);
600 return cgi_error_memory();
603 else if (0 == strcmpic(p, "blank"))
605 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
606 rsp->body = bindup(image_blank_data, image_blank_length);
607 if (rsp->body == NULL)
609 free_http_response(rsp);
610 return cgi_error_memory();
612 rsp->content_length = image_blank_length;
614 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
616 free_http_response(rsp);
617 return cgi_error_memory();
622 rsp->status = strdup_or_die("302 Local Redirect from Privoxy");
624 if (enlist_unique_header(rsp->headers, "Location", p))
626 free_http_response(rsp);
627 return cgi_error_memory();
633 #endif /* def FEATURE_IMAGE_BLOCKING */
634 if (csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
637 * Send empty document.
639 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
642 rsp->body = strdup_or_die(" ");
643 rsp->content_length = 1;
645 if (csp->config->feature_flags & RUNTIME_FEATURE_EMPTY_DOC_RETURNS_OK)
648 * Workaround for firefox bug 492459
649 * https://bugzilla.mozilla.org/show_bug.cgi?id=492459
650 * Return a 200 OK status for pages blocked with +handle-as-empty-document
651 * if the "handle-as-empty-doc-returns-ok" runtime config option is set.
653 rsp->status = strdup_or_die("200 Request blocked by Privoxy");
657 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
660 if (new_content_type != 0)
662 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
663 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
665 free_http_response(rsp);
666 return cgi_error_memory();
673 * Else, generate an HTML "blocked" message:
677 struct map * exports;
679 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
681 exports = default_exports(csp, NULL);
684 free_http_response(rsp);
685 return cgi_error_memory();
688 #ifdef FEATURE_FORCE_LOAD
689 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
691 * Export the force conditional block killer if
693 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
694 * - Privoxy is configured to enforce blocks, or
695 * - it's a CONNECT request and enforcing wouldn't work anyway.
697 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
698 || (0 == strcmpic(csp->http->gpc, "connect")))
699 #endif /* ndef FEATURE_FORCE_LOAD */
701 err = map_block_killer(exports, "force-support");
704 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
705 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
706 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
707 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
710 const char *block_reason;
711 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
713 block_reason = csp->action->string[ACTION_STRING_BLOCK];
717 assert(connect_port_is_forbidden(csp));
718 block_reason = "Forbidden CONNECT port.";
720 err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
725 free_http_response(rsp);
726 return cgi_error_memory();
729 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
732 free_http_response(rsp);
733 return cgi_error_memory();
736 rsp->crunch_reason = BLOCKED;
738 return finish_http_response(csp, rsp);
744 /*********************************************************************
746 * Function : trust_url FIXME: I should be called distrust_url
748 * Description : Calls is_untrusted_url to determine if the URL is trusted
749 * and if not, returns a HTTP 403 response with a reject message.
752 * 1 : csp = Current client state (buffers, headers, etc...)
754 * Returns : NULL => trusted, else http_response.
756 *********************************************************************/
757 struct http_response *trust_url(struct client_state *csp)
759 struct http_response *rsp;
760 struct map * exports;
761 char buf[BUFFER_SIZE];
763 struct pattern_spec **tl;
764 struct pattern_spec *t;
768 * Don't bother to work on trusted URLs
770 if (!is_untrusted_url(csp))
776 * Else, prepare a response:
778 if (NULL == (rsp = alloc_http_response()))
780 return cgi_error_memory();
783 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
784 exports = default_exports(csp, NULL);
787 free_http_response(rsp);
788 return cgi_error_memory();
792 * Export the protocol, host, port, and referrer information
794 err = map(exports, "hostport", 1, csp->http->hostport, 1);
795 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
796 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
798 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
800 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
804 if (!err) err = map(exports, "referrer", 1, "none set", 1);
810 free_http_response(rsp);
811 return cgi_error_memory();
815 * Export the trust list
817 p = strdup_or_die("");
818 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
820 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
821 string_append(&p, buf);
823 err = map(exports, "trusted-referrers", 1, p, 0);
828 free_http_response(rsp);
829 return cgi_error_memory();
833 * Export the trust info, if available
835 if (csp->config->trust_info->first)
837 struct list_entry *l;
839 p = strdup_or_die("");
840 for (l = csp->config->trust_info->first; l ; l = l->next)
842 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
843 string_append(&p, buf);
845 err = map(exports, "trust-info", 1, p, 0);
849 err = map_block_killer(exports, "have-trust-info");
855 free_http_response(rsp);
856 return cgi_error_memory();
860 * Export the force conditional block killer if
862 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
863 * - Privoxy is configured to enforce blocks, or
864 * - it's a CONNECT request and enforcing wouldn't work anyway.
866 #ifdef FEATURE_FORCE_LOAD
867 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
868 || (0 == strcmpic(csp->http->gpc, "connect")))
870 err = map_block_killer(exports, "force-support");
874 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
876 #else /* ifndef FEATURE_FORCE_LOAD */
877 err = map_block_killer(exports, "force-support");
878 #endif /* ndef FEATURE_FORCE_LOAD */
883 free_http_response(rsp);
884 return cgi_error_memory();
890 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
893 free_http_response(rsp);
894 return cgi_error_memory();
896 rsp->crunch_reason = UNTRUSTED;
898 return finish_http_response(csp, rsp);
900 #endif /* def FEATURE_TRUST */
903 /*********************************************************************
905 * Function : compile_dynamic_pcrs_job_list
907 * Description : Compiles a dynamic pcrs job list (one with variables
908 * resolved at request time)
911 * 1 : csp = Current client state (buffers, headers, etc...)
912 * 2 : b = The filter list to compile
914 * Returns : NULL in case of errors, otherwise the
917 *********************************************************************/
918 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
920 struct list_entry *pattern;
921 pcrs_job *job_list = NULL;
922 pcrs_job *dummy = NULL;
923 pcrs_job *lastjob = NULL;
926 const struct pcrs_variable variables[] =
928 {"url", csp->http->url, 1},
929 {"path", csp->http->path, 1},
930 {"host", csp->http->host, 1},
931 {"origin", csp->ip_addr_str, 1},
932 {"listen-address", csp->listen_addr_str, 1},
936 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
938 assert(pattern->str != NULL);
940 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
943 log_error(LOG_LEVEL_ERROR,
944 "Compiling dynamic pcrs job '%s' for '%s' failed with error code %d: %s",
945 pattern->str, b->name, error, pcrs_strerror(error));
950 if (error == PCRS_WARN_TRUNCATION)
952 log_error(LOG_LEVEL_ERROR,
953 "At least one of the variables in \'%s\' had to "
954 "be truncated before compilation", pattern->str);
956 if (job_list == NULL)
962 lastjob->next = dummy;
972 /*********************************************************************
974 * Function : rewrite_url
976 * Description : Rewrites a URL with a single pcrs command
977 * and returns the result if it differs from the
978 * original and isn't obviously invalid.
981 * 1 : old_url = URL to rewrite.
982 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
985 * Returns : NULL if the pcrs_command didn't change the url, or
986 * the result of the modification.
988 *********************************************************************/
989 char *rewrite_url(char *old_url, const char *pcrs_command)
991 char *new_url = NULL;
995 assert(pcrs_command);
997 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1001 log_error(LOG_LEVEL_REDIRECTS,
1002 "pcrs command \"%s\" didn't change \"%s\".",
1003 pcrs_command, old_url);
1008 log_error(LOG_LEVEL_REDIRECTS,
1009 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1010 pcrs_command, old_url, pcrs_strerror(hits));
1013 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1015 log_error(LOG_LEVEL_ERROR,
1016 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1017 "but the result doesn't look like a valid URL and will be ignored.",
1018 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1023 log_error(LOG_LEVEL_REDIRECTS,
1024 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1025 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1033 #ifdef FEATURE_FAST_REDIRECTS
1034 /*********************************************************************
1036 * Function : get_last_url
1038 * Description : Search for the last URL inside a string.
1039 * If the string already is a URL, it will
1040 * be the first URL found.
1043 * 1 : subject = the string to check
1044 * 2 : redirect_mode = +fast-redirect{} mode
1046 * Returns : NULL if no URL was found, or
1047 * the last URL found.
1049 *********************************************************************/
1050 static char *get_last_url(char *subject, const char *redirect_mode)
1052 char *new_url = NULL;
1056 assert(redirect_mode);
1058 subject = strdup(subject);
1059 if (subject == NULL)
1061 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1065 if (0 == strcmpic(redirect_mode, "check-decoded-url") && strchr(subject, '%'))
1067 char *url_segment = NULL;
1068 char **url_segments;
1069 size_t max_segments;
1072 log_error(LOG_LEVEL_REDIRECTS,
1073 "Checking \"%s\" for encoded redirects.", subject);
1076 * Check each parameter in the URL separately.
1077 * Sectionize the URL at "?" and "&",
1078 * go backwards through the segments, URL-decode them
1079 * and look for a URL in the decoded result.
1080 * Stop the search after the first match.
1082 * XXX: This estimate is guaranteed to be high enough as we
1083 * let ssplit() ignore empty fields, but also a bit wasteful.
1085 max_segments = strlen(subject) / 2;
1086 url_segments = malloc(max_segments * sizeof(char *));
1088 if (NULL == url_segments)
1090 log_error(LOG_LEVEL_ERROR,
1091 "Out of memory while decoding URL: %s", subject);
1096 segments = ssplit(subject, "?&", url_segments, max_segments);
1098 while (segments-- > 0)
1100 char *dtoken = url_decode(url_segments[segments]);
1103 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", url_segments[segments]);
1106 url_segment = strstr(dtoken, "http://");
1107 if (NULL == url_segment)
1109 url_segment = strstr(dtoken, "https://");
1111 if (NULL != url_segment)
1113 url_segment = strdup_or_die(url_segment);
1120 freez(url_segments);
1122 if (url_segment == NULL)
1126 subject = url_segment;
1130 /* Look for a URL inside this one, without decoding anything. */
1131 log_error(LOG_LEVEL_REDIRECTS,
1132 "Checking \"%s\" for unencoded redirects.", subject);
1136 * Find the last URL encoded in the request
1139 while ((tmp = strstr(tmp, "http://")) != NULL)
1143 tmp = (new_url != NULL) ? new_url : subject;
1144 while ((tmp = strstr(tmp, "https://")) != NULL)
1149 if ((new_url != NULL)
1150 && ( (new_url != subject)
1151 || (0 == strncmpic(subject, "http://", 7))
1152 || (0 == strncmpic(subject, "https://", 8))
1156 * Return new URL if we found a redirect
1157 * or if the subject already was a URL.
1159 * The second case makes sure that we can
1160 * chain get_last_url after another redirection check
1161 * (like rewrite_url) without losing earlier redirects.
1163 new_url = strdup(new_url);
1172 #endif /* def FEATURE_FAST_REDIRECTS */
1175 /*********************************************************************
1177 * Function : redirect_url
1179 * Description : Checks if Privoxy should answer the request with
1180 * a HTTP redirect and generates the redirect if
1184 * 1 : csp = Current client state (buffers, headers, etc...)
1186 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1188 *********************************************************************/
1189 struct http_response *redirect_url(struct client_state *csp)
1191 struct http_response *rsp;
1192 #ifdef FEATURE_FAST_REDIRECTS
1194 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1195 * as compile-time option? The user can easily disable
1196 * it in his action file.
1198 char * redirect_mode;
1199 #endif /* def FEATURE_FAST_REDIRECTS */
1200 char *old_url = NULL;
1201 char *new_url = NULL;
1202 char *redirection_string;
1204 if ((csp->action->flags & ACTION_REDIRECT))
1206 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1209 * If the redirection string begins with 's',
1210 * assume it's a pcrs command, otherwise treat it as
1211 * properly formatted URL and use it for the redirection
1214 * According to (the now obsolete) RFC 2616 section 14.30
1215 * the URL has to be absolute and if the user tries:
1216 * +redirect{sadly/this/will/be/parsed/as/pcrs_command.html}
1217 * she would get undefined results anyway.
1219 * RFC 7231 7.1.2 actually allows relative references,
1220 * but those start with a leading slash (RFC 3986 4.2) and
1221 * thus can't be mistaken for pcrs commands either.
1224 if (*redirection_string == 's')
1226 #ifdef FEATURE_HTTPS_INSPECTION
1227 if (client_use_ssl(csp))
1231 old_url = strdup_or_die("https://");
1232 err = string_append(&old_url, csp->http->hostport);
1233 if (!err) err = string_append(&old_url, csp->http->path);
1236 log_error(LOG_LEVEL_FATAL,
1237 "Failed to rebuild URL 'https://%s%s'",
1238 csp->http->hostport, csp->http->path);
1244 old_url = csp->http->url;
1246 new_url = rewrite_url(old_url, redirection_string);
1247 #ifdef FEATURE_HTTPS_INSPECTION
1248 if (client_use_ssl(csp))
1256 log_error(LOG_LEVEL_REDIRECTS,
1257 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1258 redirection_string);
1259 new_url = strdup(redirection_string);
1263 #ifdef FEATURE_FAST_REDIRECTS
1264 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1266 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1269 * If it exists, use the previously rewritten URL as input
1270 * otherwise just use the old path.
1272 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1273 new_url = get_last_url(old_url, redirect_mode);
1278 * Disable redirect checkers, so that they
1279 * will be only run more than once if the user
1280 * also enables them through tags.
1282 * From a performance point of view
1283 * it doesn't matter, but the duplicated
1284 * log messages are annoying.
1286 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1287 #endif /* def FEATURE_FAST_REDIRECTS */
1288 csp->action->flags &= ~ACTION_REDIRECT;
1290 /* Did any redirect action trigger? */
1293 if (url_requires_percent_encoding(new_url))
1296 log_error(LOG_LEVEL_REDIRECTS, "Percent-encoding redirect URL: %N",
1297 strlen(new_url), new_url);
1298 encoded_url = percent_encode_url(new_url);
1300 if (encoded_url == NULL)
1302 return cgi_error_memory();
1304 new_url = encoded_url;
1305 assert(FALSE == url_requires_percent_encoding(new_url));
1308 if (0 == strcmpic(new_url, csp->http->url))
1310 log_error(LOG_LEVEL_ERROR,
1311 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1312 csp->http->url, new_url);
1317 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1319 if (NULL == (rsp = alloc_http_response()))
1322 return cgi_error_memory();
1325 rsp->status = strdup_or_die("302 Local Redirect from Privoxy");
1326 if (enlist_unique_header(rsp->headers, "Location", new_url))
1329 free_http_response(rsp);
1330 return cgi_error_memory();
1332 rsp->crunch_reason = REDIRECTED;
1335 return finish_http_response(csp, rsp);
1339 /* Only reached if no redirect is required */
1345 #ifdef FEATURE_IMAGE_BLOCKING
1346 /*********************************************************************
1348 * Function : is_imageurl
1350 * Description : Given a URL, decide whether it should be treated
1351 * as image URL or not.
1354 * 1 : csp = Current client state (buffers, headers, etc...)
1356 * Returns : True (nonzero) if URL is an image URL, false (0)
1359 *********************************************************************/
1360 int is_imageurl(const struct client_state *csp)
1362 return ((csp->action->flags & ACTION_IMAGE) != 0);
1365 #endif /* def FEATURE_IMAGE_BLOCKING */
1368 #ifdef FEATURE_TRUST
1369 /*********************************************************************
1371 * Function : is_untrusted_url
1373 * Description : Should we "distrust" this URL (and block it)?
1375 * Yes if it matches a line in the trustfile, or if the
1376 * referrer matches a line starting with "+" in the
1381 * 1 : csp = Current client state (buffers, headers, etc...)
1383 * Returns : 0 => trusted, 1 => untrusted
1385 *********************************************************************/
1386 int is_untrusted_url(const struct client_state *csp)
1388 struct file_list *fl;
1389 struct block_spec *b;
1390 struct pattern_spec **trusted_url;
1391 struct http_request rhttp[1];
1392 const char * referer;
1396 * If we don't have a trustlist, we trust everybody
1398 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1403 memset(rhttp, '\0', sizeof(*rhttp));
1406 * Do we trust the request URL itself?
1408 for (b = b->next; b ; b = b->next)
1410 if (url_match(b->url, csp->http))
1416 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1418 /* no referrer was supplied */
1424 * If not, do we maybe trust its referrer?
1426 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
1432 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1434 if (url_match(*trusted_url, rhttp))
1436 /* if the URL's referrer is from a trusted referrer, then
1437 * add the target spec to the trustfile as an unblocked
1438 * domain and return 0 (which means it's OK).
1443 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1447 char * new_entry = strdup_or_die("~");
1449 string_append(&new_entry, csp->http->hostport);
1451 path = csp->http->path;
1452 if ( (path[0] == '/')
1454 && ((path_end = strchr(path + 2, '/')) != NULL))
1456 /* since this path points into a user's home space
1457 * be sure to include this spec in the trustfile.
1459 long path_len = path_end - path; /* save offset */
1460 path = strdup(path); /* Copy string */
1463 path_end = path + path_len; /* regenerate ptr to new buffer */
1464 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1466 string_join(&new_entry, path);
1470 * Give a reason for generating this entry.
1472 string_append(&new_entry, " # Trusted referrer was: ");
1473 string_append(&new_entry, referer);
1475 if (new_entry != NULL)
1477 if (-1 == fprintf(fp, "%s\n", new_entry))
1479 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1480 new_entry, csp->config->trustfile);
1486 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1487 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1494 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1495 csp->http->hostport, csp->config->trustfile);
1503 #endif /* def FEATURE_TRUST */
1506 /*********************************************************************
1508 * Function : get_filter
1510 * Description : Get a filter with a given name and type.
1511 * Note that taggers are filters, too.
1514 * 1 : csp = Current client state (buffers, headers, etc...)
1515 * 2 : requested_name = Name of the content filter to get
1516 * 3 : requested_type = Type of the filter to tagger to lookup
1518 * Returns : A pointer to the requested filter
1519 * or NULL if the filter wasn't found
1521 *********************************************************************/
1522 struct re_filterfile_spec *get_filter(const struct client_state *csp,
1523 const char *requested_name,
1524 enum filter_type requested_type)
1527 struct re_filterfile_spec *b;
1528 struct file_list *fl;
1530 for (i = 0; i < MAX_AF_FILES; i++)
1533 if ((NULL == fl) || (NULL == fl->f))
1536 * Either there are no filter files left or this
1537 * filter file just contains no valid filters.
1539 * Continue to be sure we don't miss valid filter
1540 * files that are chained after empty or invalid ones.
1545 for (b = fl->f; b != NULL; b = b->next)
1547 if (b->type != requested_type)
1549 /* The callers isn't interested in this filter type. */
1552 if (strcmp(b->name, requested_name) == 0)
1554 /* The requested filter has been found. Abort search. */
1560 /* No filter with the given name and type exists. */
1566 /*********************************************************************
1568 * Function : pcrs_filter_response
1570 * Description : Execute all text substitutions from all applying
1571 * +filter actions on the text buffer that's been
1572 * accumulated in csp->iob->buf.
1575 * 1 : csp = Current client state (buffers, headers, etc...)
1577 * Returns : a pointer to the (newly allocated) modified buffer.
1578 * or NULL if there were no hits or something went wrong
1580 *********************************************************************/
1581 static char *pcrs_filter_response(struct client_state *csp)
1584 size_t size, prev_size;
1590 struct re_filterfile_spec *b;
1591 struct list_entry *filtername;
1596 if (csp->iob->cur >= csp->iob->eod)
1601 if (filters_available(csp) == FALSE)
1603 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1604 "content filtering enabled, but no content filters available.");
1608 size = (size_t)(csp->iob->eod - csp->iob->cur);
1609 old = csp->iob->cur;
1612 * For all applying +filter actions, look if a filter by that
1613 * name exists and if yes, execute it's pcrs_joblist on the
1616 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1617 filtername != NULL; filtername = filtername->next)
1619 int current_hits = 0; /* Number of hits caused by this filter */
1620 int job_number = 0; /* Which job we're currently executing */
1621 int job_hits = 0; /* How many hits the current job caused */
1624 b = get_filter(csp, filtername->str, FT_CONTENT_FILTER);
1630 joblist = b->joblist;
1632 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1634 if (NULL == joblist)
1636 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1641 /* Apply all jobs from the joblist */
1642 for (job = joblist; NULL != job; job = job->next)
1645 job_hits = pcrs_execute(job, old, size, &new, &size);
1650 * That went well. Continue filtering
1651 * and use the result of this job as
1652 * input for the next one.
1654 current_hits += job_hits;
1655 if (old != csp->iob->cur)
1664 * This job caused an unexpected error. Inform the user
1665 * and skip the rest of the jobs in this filter. We could
1666 * continue with the next job, but usually the jobs
1667 * depend on each other or are similar enough to
1668 * fail for the same reason.
1670 * At the moment our pcrs expects the error codes of pcre 3.4,
1671 * but newer pcre versions can return additional error codes.
1672 * As a result pcrs_strerror()'s error message might be
1673 * "Unknown error ...", therefore we print the numerical value
1676 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1677 * should we use LOG_LEVEL_RE_FILTER instead?
1679 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1680 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1685 if (b->dynamic) pcrs_free_joblist(joblist);
1687 log_error(LOG_LEVEL_RE_FILTER,
1688 "filtering %s%s (size %lu) with \'%s\' produced %d hits (new size %lu).",
1689 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1690 #ifdef FEATURE_EXTENDED_STATISTICS
1691 update_filter_statistics(b->name, current_hits);
1693 hits += current_hits;
1697 * If there were no hits, destroy our copy and let
1698 * chat() use the original in csp->iob
1702 if (old != csp->iob->cur && old != new)
1710 csp->flags |= CSP_FLAG_MODIFIED;
1711 csp->content_length = size;
1712 clear_iob(csp->iob);
1719 #ifdef FEATURE_EXTERNAL_FILTERS
1720 /*********************************************************************
1722 * Function : get_external_filter
1724 * Description : Lookup the code to execute for an external filter.
1725 * Masks the misuse of the re_filterfile_spec.
1728 * 1 : csp = Current client state (buffers, headers, etc...)
1729 * 2 : name = Name of the content filter to get
1731 * Returns : A pointer to the requested code
1732 * or NULL if the filter wasn't found
1734 *********************************************************************/
1735 static const char *get_external_filter(const struct client_state *csp,
1738 struct re_filterfile_spec *external_filter;
1740 external_filter = get_filter(csp, name, FT_EXTERNAL_CONTENT_FILTER);
1741 if (external_filter == NULL)
1743 log_error(LOG_LEVEL_FATAL,
1744 "Didn't find stuff to execute for external filter: %s",
1748 return external_filter->patterns->first->str;
1753 /*********************************************************************
1755 * Function : set_privoxy_variables
1757 * Description : Sets a couple of privoxy-specific environment variables
1760 * 1 : csp = Current client state (buffers, headers, etc...)
1764 *********************************************************************/
1765 static void set_privoxy_variables(const struct client_state *csp)
1772 { "PRIVOXY_URL", csp->http->url },
1773 { "PRIVOXY_PATH", csp->http->path },
1774 { "PRIVOXY_HOST", csp->http->host },
1775 { "PRIVOXY_ORIGIN", csp->ip_addr_str },
1776 { "PRIVOXY_LISTEN_ADDRESS", csp->listen_addr_str },
1779 for (i = 0; i < SZ(env); i++)
1781 if (setenv(env[i].name, env[i].value, 1))
1783 log_error(LOG_LEVEL_ERROR, "Failed to set %s=%s: %E",
1784 env[i].name, env[i].value);
1790 /*********************************************************************
1792 * Function : execute_external_filter
1794 * Description : Pipe content into external filter and return the output
1797 * 1 : csp = Current client state (buffers, headers, etc...)
1798 * 2 : name = Name of the external filter to execute
1799 * 3 : content = The original content to filter
1800 * 4 : size = The size of the content buffer
1802 * Returns : a pointer to the (newly allocated) modified buffer.
1803 * or NULL if there were no hits or something went wrong
1805 *********************************************************************/
1806 static char *execute_external_filter(const struct client_state *csp,
1807 const char *name, char *content, size_t *size)
1810 char file_name[FILENAME_MAX];
1812 char *filter_output;
1816 const char *external_filter;
1818 if (csp->config->temporary_directory == NULL)
1820 log_error(LOG_LEVEL_ERROR,
1821 "No temporary-directory configured. Can't execute filter: %s",
1826 external_filter = get_external_filter(csp, name);
1828 if (sizeof(file_name) < snprintf(file_name, sizeof(file_name),
1829 "%s/privoxy-XXXXXXXX", csp->config->temporary_directory))
1831 log_error(LOG_LEVEL_ERROR, "temporary-directory path too long");
1835 fd = mkstemp(file_name);
1838 log_error(LOG_LEVEL_ERROR, "mkstemp() failed to create %s: %E", file_name);
1842 fp = fdopen(fd, "w");
1845 log_error(LOG_LEVEL_ERROR, "fdopen() failed: %E");
1851 * The size may be zero if a previous filter discarded everything.
1853 * This isn't necessary unintentional, so we just don't try
1854 * to fwrite() nothing and let the user deal with the rest.
1856 if ((*size != 0) && fwrite(content, *size, 1, fp) != 1)
1858 log_error(LOG_LEVEL_ERROR, "fwrite(..., %lu, 1, ..) failed: %E", *size);
1865 if (sizeof(cmd) < snprintf(cmd, sizeof(cmd), "%s < %s", external_filter, file_name))
1867 log_error(LOG_LEVEL_ERROR,
1868 "temporary-directory or external filter path too long");
1873 log_error(LOG_LEVEL_RE_FILTER, "Executing '%s': %s", name, cmd);
1876 * The locking is necessary to prevent other threads
1877 * from overwriting the environment variables before
1878 * the popen fork. Afterwards this no longer matters.
1880 privoxy_mutex_lock(&external_filter_mutex);
1881 set_privoxy_variables(csp);
1882 fp = popen(cmd, "r");
1883 privoxy_mutex_unlock(&external_filter_mutex);
1886 log_error(LOG_LEVEL_ERROR, "popen(\"%s\", \"r\") failed: %E", cmd);
1891 /* Allocate at least one byte */
1892 filter_output = malloc_or_die(*size + 1);
1895 while (!feof(fp) && !ferror(fp))
1898 /* Could be bigger ... */
1899 enum { READ_LENGTH = 2048 };
1901 if (new_size + READ_LENGTH >= *size)
1905 /* Could be considered wasteful if the content is 'large'. */
1906 *size += (*size >= READ_LENGTH) ? *size : READ_LENGTH;
1908 p = realloc(filter_output, *size);
1911 log_error(LOG_LEVEL_ERROR, "Out of memory while reading "
1912 "external filter output. Using what we got so far.");
1917 assert(new_size + READ_LENGTH < *size);
1918 len = fread(&filter_output[new_size], 1, READ_LENGTH, fp);
1928 log_error(LOG_LEVEL_ERROR, "Executing %s failed: %E", cmd);
1932 log_error(LOG_LEVEL_RE_FILTER,
1933 "Executing '%s' resulted in return value %d. "
1934 "Read %lu of up to %lu bytes.", name, (ret >> 8), new_size, *size);
1940 return filter_output;
1943 #endif /* def FEATURE_EXTERNAL_FILTERS */
1946 /*********************************************************************
1948 * Function : gif_deanimate_response
1950 * Description : Deanimate the GIF image that has been accumulated in
1951 * csp->iob->buf, set csp->content_length to the modified
1952 * size and raise the CSP_FLAG_MODIFIED flag.
1955 * 1 : csp = Current client state (buffers, headers, etc...)
1957 * Returns : a pointer to the (newly allocated) modified buffer.
1958 * or NULL in case something went wrong.
1960 *********************************************************************/
1962 char *gif_deanimate_response(struct client_state *csp)
1964 static char *gif_deanimate_response(struct client_state *csp)
1967 struct binbuffer *in, *out;
1971 size = (size_t)(csp->iob->eod - csp->iob->cur);
1973 in = zalloc_or_die(sizeof(*in));
1974 out = zalloc_or_die(sizeof(*out));
1976 in->buffer = csp->iob->cur;
1979 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1981 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1988 if ((int)size == out->offset)
1990 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1994 log_error(LOG_LEVEL_DEANIMATE,
1995 "Success! GIF shrunk from %lu bytes to %lu.", size, out->offset);
1997 csp->content_length = out->offset;
1998 csp->flags |= CSP_FLAG_MODIFIED;
2008 /*********************************************************************
2010 * Function : get_filter_function
2012 * Description : Decides which content filter function has
2013 * to be applied (if any). Only considers functions
2014 * for internal filters which are mutually-exclusive.
2017 * 1 : csp = Current client state (buffers, headers, etc...)
2019 * Returns : The content filter function to run, or
2020 * NULL if no content filter is active
2022 *********************************************************************/
2023 static filter_function_ptr get_filter_function(const struct client_state *csp)
2025 filter_function_ptr filter_function = NULL;
2028 * Choose the applying filter function based on
2029 * the content type and action settings.
2031 if ((csp->content_type & CT_TEXT) &&
2032 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2034 filter_function = pcrs_filter_response;
2036 else if ((csp->content_type & CT_GIF) &&
2037 (csp->action->flags & ACTION_DEANIMATE))
2039 filter_function = gif_deanimate_response;
2042 return filter_function;
2046 /*********************************************************************
2048 * Function : remove_chunked_transfer_coding
2050 * Description : In-situ remove the "chunked" transfer coding as defined
2051 * in RFC 7230 4.1 from a buffer. XXX: The implementation
2052 * is neither complete nor compliant (TODO #129).
2055 * 1 : buffer = Pointer to the text buffer
2056 * 2 : size = In: Number of bytes to be processed,
2057 * Out: Number of bytes after de-chunking.
2058 * (undefined in case of errors)
2060 * Returns : JB_ERR_OK for success,
2061 * JB_ERR_PARSE otherwise
2063 *********************************************************************/
2065 extern jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2067 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2071 unsigned int chunksize = 0;
2072 char *from_p, *to_p;
2073 const char *end_of_buffer = buffer + *size;
2077 log_error(LOG_LEVEL_FATAL, "Invalid chunked input. Buffer is empty.");
2078 return JB_ERR_PARSE;
2082 from_p = to_p = buffer;
2084 if (sscanf(buffer, "%x", &chunksize) != 1)
2086 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2087 return JB_ERR_PARSE;
2090 while (chunksize > 0U)
2093 * If the chunk-size is valid, we should have at least
2094 * chunk-size bytes of chunk-data and five bytes of
2095 * meta data (chunk-size, CRLF, CRLF) left in the buffer.
2097 if (chunksize + 5 >= *size - newsize)
2099 log_error(LOG_LEVEL_ERROR,
2100 "Chunk size %u exceeds buffered data left. "
2101 "Already digested %lu of %lu buffered bytes.",
2102 chunksize, newsize, *size);
2103 return JB_ERR_PARSE;
2107 * Skip the chunk-size, the optional chunk-ext and the CRLF
2108 * that is supposed to be located directly before the start
2111 if (NULL == (from_p = strstr(from_p, "\r\n")))
2113 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2114 return JB_ERR_PARSE;
2119 * The previous strstr() does not enforce chunk-validity
2120 * and is sattisfied as long a CRLF is left in the buffer.
2122 * Make sure the bytes we consider chunk-data are within
2125 if (from_p + chunksize >= end_of_buffer)
2127 log_error(LOG_LEVEL_ERROR,
2128 "End of chunk is beyond the end of the buffer.");
2129 return JB_ERR_PARSE;
2132 memmove(to_p, from_p, (size_t) chunksize);
2133 newsize += chunksize;
2134 to_p = buffer + newsize;
2135 from_p += chunksize;
2138 * Not merging this check with the previous one allows us
2139 * to keep chunks without trailing CRLF. It's not clear
2140 * if we actually have to care about those, though.
2142 if (from_p + 2 >= end_of_buffer)
2144 log_error(LOG_LEVEL_ERROR, "Not enough room for trailing CRLF.");
2145 return JB_ERR_PARSE;
2148 if (sscanf(from_p, "%x", &chunksize) != 1)
2150 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
2155 /* XXX: Should get its own loglevel. */
2156 log_error(LOG_LEVEL_RE_FILTER,
2157 "De-chunking successful. Shrunk from %lu to %lu", *size, newsize);
2166 /*********************************************************************
2168 * Function : prepare_for_filtering
2170 * Description : If necessary, de-chunks and decompresses
2171 * the content so it can get filterd.
2174 * 1 : csp = Current client state (buffers, headers, etc...)
2176 * Returns : JB_ERR_OK for success,
2177 * JB_ERR_PARSE otherwise
2179 *********************************************************************/
2180 static jb_err prepare_for_filtering(struct client_state *csp)
2182 jb_err err = JB_ERR_OK;
2185 * If the body has a "chunked" transfer-encoding,
2186 * get rid of it, adjusting size and iob->eod
2188 if (csp->flags & CSP_FLAG_CHUNKED)
2190 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2192 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2193 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2194 if (JB_ERR_OK == err)
2196 csp->iob->eod = csp->iob->cur + size;
2197 csp->flags |= CSP_FLAG_MODIFIED;
2201 return JB_ERR_PARSE;
2207 * If the body has a supported transfer-encoding,
2208 * decompress it, adjusting size and iob->eod.
2210 if ((csp->content_type & (CT_GZIP|CT_DEFLATE))
2211 #ifdef FEATURE_BROTLI
2212 || (csp->content_type & CT_BROTLI)
2216 if (0 == csp->iob->eod - csp->iob->cur)
2218 /* Nothing left after de-chunking. */
2222 err = decompress_iob(csp);
2224 if (JB_ERR_OK == err)
2226 csp->flags |= CSP_FLAG_MODIFIED;
2227 csp->content_type &= ~CT_TABOO;
2232 * Unset content types to remember not to
2233 * modify the Content-Encoding header later.
2235 csp->content_type &= ~CT_GZIP;
2236 csp->content_type &= ~CT_DEFLATE;
2237 #ifdef FEATURE_BROTLI
2238 csp->content_type &= ~CT_BROTLI;
2248 /*********************************************************************
2250 * Function : execute_content_filters
2252 * Description : Executes a given content filter.
2255 * 1 : csp = Current client state (buffers, headers, etc...)
2257 * Returns : Pointer to the modified buffer, or
2258 * NULL if filtering failed or wasn't necessary.
2260 *********************************************************************/
2261 char *execute_content_filters(struct client_state *csp)
2264 filter_function_ptr content_filter;
2266 assert(content_filters_enabled(csp->action));
2268 if (0 == csp->iob->eod - csp->iob->cur)
2271 * No content (probably status code 301, 302 ...),
2272 * no filtering necessary.
2277 if (JB_ERR_OK != prepare_for_filtering(csp))
2280 * failed to de-chunk or decompress.
2285 if (0 == csp->iob->eod - csp->iob->cur)
2288 * Clown alarm: chunked and/or compressed nothing delivered.
2293 content_filter = get_filter_function(csp);
2294 content = (content_filter != NULL) ? (*content_filter)(csp) : NULL;
2296 #ifdef FEATURE_EXTERNAL_FILTERS
2297 if ((csp->content_type & CT_TEXT) &&
2298 !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]))
2300 struct list_entry *filtername;
2301 size_t size = (size_t)csp->content_length;
2303 if (content == NULL)
2305 content = csp->iob->cur;
2306 size = (size_t)(csp->iob->eod - csp->iob->cur);
2309 for (filtername = csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]->first;
2310 filtername ; filtername = filtername->next)
2312 char *result = execute_external_filter(csp, filtername->str, content, &size);
2315 if (content != csp->iob->cur)
2322 csp->flags |= CSP_FLAG_MODIFIED;
2323 csp->content_length = size;
2325 #endif /* def FEATURE_EXTERNAL_FILTERS */
2332 /*********************************************************************
2334 * Function : get_url_actions
2336 * Description : Gets the actions for this URL.
2339 * 1 : csp = Current client state (buffers, headers, etc...)
2340 * 2 : http = http_request request for blocked URLs
2344 *********************************************************************/
2345 void get_url_actions(struct client_state *csp, struct http_request *http)
2347 struct file_list *fl;
2348 struct url_actions *b;
2351 init_current_action(csp->action);
2353 for (i = 0; i < MAX_AF_FILES; i++)
2355 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2360 #ifdef FEATURE_CLIENT_TAGS
2361 apply_url_actions(csp->action, http, csp->client_tags, b);
2363 apply_url_actions(csp->action, http, b);
2370 /*********************************************************************
2372 * Function : apply_url_actions
2374 * Description : Applies a list of URL actions.
2377 * 1 : action = Destination.
2378 * 2 : http = Current URL
2379 * 3 : client_tags = list of client tags
2380 * 4 : b = list of URL actions to apply
2384 *********************************************************************/
2385 static void apply_url_actions(struct current_action_spec *action,
2386 struct http_request *http,
2387 #ifdef FEATURE_CLIENT_TAGS
2388 const struct list *client_tags,
2390 struct url_actions *b)
2394 /* Should never happen */
2398 for (b = b->next; NULL != b; b = b->next)
2400 if (url_match(b->url, http))
2402 merge_current_action(action, b->action);
2404 #ifdef FEATURE_CLIENT_TAGS
2405 if (client_tag_match(b->url, client_tags))
2407 merge_current_action(action, b->action);
2414 /*********************************************************************
2416 * Function : get_forward_override_settings
2418 * Description : Returns forward settings as specified with the
2419 * forward-override{} action. forward-override accepts
2420 * forward lines similar to the one used in the
2421 * configuration file, but without the URL pattern.
2427 * in the configuration file can be replaced with
2428 * the action section:
2430 * {+forward-override{forward .}}
2434 * 1 : csp = Current client state (buffers, headers, etc...)
2436 * Returns : Pointer to forwarding structure in case of success.
2437 * Invalid syntax is fatal.
2439 *********************************************************************/
2440 static const struct forward_spec *get_forward_override_settings(struct client_state *csp)
2442 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2443 char forward_settings[BUFFER_SIZE];
2444 char *http_parent = NULL;
2445 /* variable names were chosen for consistency reasons. */
2446 struct forward_spec *fwd = NULL;
2450 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2451 /* Should be enforced by load_one_actions_file() */
2452 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2454 /* Create a copy ssplit can modify */
2455 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2457 if (NULL != csp->fwd)
2460 * XXX: Currently necessary to prevent memory
2461 * leaks when the show-url-info cgi page is visited.
2463 unload_forward_spec(csp->fwd);
2467 * allocate a new forward node, valid only for
2468 * the lifetime of this request. Save its location
2469 * in csp as well, so sweep() can free it later on.
2471 fwd = csp->fwd = zalloc_or_die(sizeof(*fwd));
2473 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec));
2474 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2476 fwd->type = SOCKS_NONE;
2478 /* Parse the parent HTTP proxy host:port */
2479 http_parent = vec[1];
2482 else if ((vec_count == 2) && !strcasecmp(vec[0], "forward-webserver"))
2484 fwd->type = FORWARD_WEBSERVER;
2486 /* Parse the parent HTTP server host:port */
2487 http_parent = vec[1];
2490 else if (vec_count == 3)
2492 char *socks_proxy = NULL;
2494 if (!strcasecmp(vec[0], "forward-socks4"))
2496 fwd->type = SOCKS_4;
2497 socks_proxy = vec[1];
2499 else if (!strcasecmp(vec[0], "forward-socks4a"))
2501 fwd->type = SOCKS_4A;
2502 socks_proxy = vec[1];
2504 else if (!strcasecmp(vec[0], "forward-socks5"))
2506 fwd->type = SOCKS_5;
2507 socks_proxy = vec[1];
2509 else if (!strcasecmp(vec[0], "forward-socks5t"))
2511 fwd->type = SOCKS_5T;
2512 socks_proxy = vec[1];
2515 if (NULL != socks_proxy)
2517 /* Parse the SOCKS proxy [user:pass@]host[:port] */
2518 fwd->gateway_port = 1080;
2519 parse_forwarder_address(socks_proxy,
2520 &fwd->gateway_host, &fwd->gateway_port,
2521 &fwd->auth_username, &fwd->auth_password);
2523 http_parent = vec[2];
2527 if (NULL == http_parent)
2529 log_error(LOG_LEVEL_FATAL,
2530 "Invalid forward-override syntax in: %s", forward_override_line);
2531 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2534 /* Parse http forwarding settings */
2535 if (strcmp(http_parent, ".") != 0)
2537 fwd->forward_port = 8000;
2538 parse_forwarder_address(http_parent,
2539 &fwd->forward_host, &fwd->forward_port,
2543 assert (NULL != fwd);
2545 log_error(LOG_LEVEL_CONNECT,
2546 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2552 /*********************************************************************
2554 * Function : forward_url
2556 * Description : Should we forward this to another proxy?
2559 * 1 : csp = Current client state (buffers, headers, etc...)
2560 * 2 : http = http_request request for current URL
2562 * Returns : Pointer to forwarding information.
2564 *********************************************************************/
2565 const struct forward_spec *forward_url(struct client_state *csp,
2566 const struct http_request *http)
2568 static const struct forward_spec fwd_default[1]; /* Zero'ed due to being static. */
2569 struct forward_spec *fwd = csp->config->forward;
2571 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2573 return get_forward_override_settings(csp);
2583 if (url_match(fwd->url, http))
2594 /*********************************************************************
2596 * Function : direct_response
2598 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2599 * request and if so, return a HTTP 501 to the client.
2601 * FIXME: I have a stupid name and I should handle the
2602 * requests properly. Still, what we do here is rfc-
2603 * compliant, whereas ignoring or forwarding are not.
2606 * 1 : csp = Current client state (buffers, headers, etc...)
2608 * Returns : http_response if , NULL if nonmatch or handler fail
2610 *********************************************************************/
2611 struct http_response *direct_response(struct client_state *csp)
2613 struct http_response *rsp;
2614 struct list_entry *p;
2616 if ((0 == strcmpic(csp->http->gpc, "trace"))
2617 || (0 == strcmpic(csp->http->gpc, "options")))
2619 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2621 if (!strncmpic(p->str, "Max-Forwards:", 13))
2623 unsigned int max_forwards;
2626 * If it's a Max-Forwards value of zero,
2627 * we have to intercept the request.
2629 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2632 * FIXME: We could handle at least TRACE here,
2633 * but that would require a verbatim copy of
2634 * the request which we don't have anymore
2636 log_error(LOG_LEVEL_HEADER,
2637 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2640 /* Get mem for response or fail*/
2641 if (NULL == (rsp = alloc_http_response()))
2643 return cgi_error_memory();
2646 rsp->status = strdup_or_die("501 Not Implemented");
2648 rsp->crunch_reason = UNSUPPORTED;
2650 return(finish_http_response(csp, rsp));
2659 /*********************************************************************
2661 * Function : content_requires_filtering
2663 * Description : Checks whether there are any content filters
2664 * enabled for the current request and if they
2665 * can actually be applied..
2668 * 1 : csp = Current client state (buffers, headers, etc...)
2670 * Returns : TRUE for yes, FALSE otherwise
2672 *********************************************************************/
2673 int content_requires_filtering(struct client_state *csp)
2675 if ((csp->content_type & CT_TABOO)
2676 && !(csp->action->flags & ACTION_FORCE_TEXT_MODE))
2682 * Are we enabling text mode by force?
2684 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2687 * Do we really have to?
2689 if (csp->content_type & CT_TEXT)
2691 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2695 csp->content_type |= CT_TEXT;
2696 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2700 if (!(csp->content_type & CT_DECLARED))
2703 * The server didn't bother to declare a MIME-Type.
2704 * Assume it's text that can be filtered.
2706 * This also regularly happens with 304 responses,
2707 * therefore logging anything here would cause
2710 csp->content_type |= CT_TEXT;
2714 * Choose the applying filter function based on
2715 * the content type and action settings.
2717 if ((csp->content_type & CT_TEXT) &&
2718 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER]) ||
2719 !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER])))
2723 else if ((csp->content_type & CT_GIF) &&
2724 (csp->action->flags & ACTION_DEANIMATE))
2734 /*********************************************************************
2736 * Function : content_filters_enabled
2738 * Description : Checks whether there are any content filters
2739 * enabled for the current request.
2742 * 1 : action = Action spec to check.
2744 * Returns : TRUE for yes, FALSE otherwise
2746 *********************************************************************/
2747 int content_filters_enabled(const struct current_action_spec *action)
2749 return ((action->flags & ACTION_DEANIMATE) ||
2750 !list_is_empty(action->multi[ACTION_MULTI_FILTER]) ||
2751 !list_is_empty(action->multi[ACTION_MULTI_EXTERNAL_FILTER]));
2755 /*********************************************************************
2757 * Function : filters_available
2759 * Description : Checks whether there are any filters available.
2762 * 1 : csp = Current client state (buffers, headers, etc...)
2764 * Returns : TRUE for yes, FALSE otherwise.
2766 *********************************************************************/
2767 int filters_available(const struct client_state *csp)
2770 for (i = 0; i < MAX_AF_FILES; i++)
2772 const struct file_list *fl = csp->rlist[i];
2773 if ((NULL != fl) && (NULL != fl->f))
2781 #ifdef FEATURE_EXTENDED_STATISTICS
2783 struct filter_statistics_entry
2786 unsigned long long executions;
2787 unsigned long long response_bodies_modified;
2788 unsigned long long hits;
2790 struct filter_statistics_entry *next;
2793 static struct filter_statistics_entry *filter_statistics = NULL;
2796 /*********************************************************************
2798 * Function : register_filter_for_statistics
2800 * Description : Registers a filter so we can gather statistics for
2801 * it unless the filter has already been registered
2805 * 1 : filter = Name of the filter to register
2809 *********************************************************************/
2810 void register_filter_for_statistics(const char *filter)
2812 struct filter_statistics_entry *entry;
2814 privoxy_mutex_lock(&filter_statistics_mutex);
2816 if (filter_statistics == NULL)
2818 filter_statistics = zalloc_or_die(sizeof(struct filter_statistics_entry));
2819 entry = filter_statistics;
2820 entry->filter = strdup_or_die(filter);
2821 privoxy_mutex_unlock(&filter_statistics_mutex);
2824 entry = filter_statistics;
2825 while (entry != NULL)
2827 if (!strcmp(entry->filter, filter))
2829 /* Already registered, nothing to do. */
2832 if (entry->next == NULL)
2834 entry->next = zalloc_or_die(sizeof(struct filter_statistics_entry));
2835 entry->next->filter = strdup_or_die(filter);
2838 entry = entry->next;
2841 privoxy_mutex_unlock(&filter_statistics_mutex);
2846 /*********************************************************************
2848 * Function : update_filter_statistics
2850 * Description : Updates the statistics for a filter.
2853 * 1 : filter = Name of the filter to update
2854 * 2 : hits = Hit count.
2858 *********************************************************************/
2859 void update_filter_statistics(const char *filter, int hits)
2861 struct filter_statistics_entry *entry;
2863 privoxy_mutex_lock(&filter_statistics_mutex);
2865 entry = filter_statistics;
2866 while (entry != NULL)
2868 if (!strcmp(entry->filter, filter))
2870 entry->executions++;
2873 entry->response_bodies_modified++;
2874 entry->hits += (unsigned)hits;
2878 entry = entry->next;
2881 privoxy_mutex_unlock(&filter_statistics_mutex);
2886 /*********************************************************************
2888 * Function : get_filter_statistics
2890 * Description : Gets the statistics for a filter.
2893 * 1 : filter = Name of the filter to get statistics for.
2894 * 2 : executions = Storage for the execution count.
2895 * 3 : response_bodies_modified = Storage for the number
2896 * of modified response bodies.
2897 * 4 : hits = Storage for the number of hits.
2901 *********************************************************************/
2902 void get_filter_statistics(const char *filter, unsigned long long *executions,
2903 unsigned long long *response_bodies_modified,
2904 unsigned long long *hits)
2906 struct filter_statistics_entry *entry;
2908 privoxy_mutex_lock(&filter_statistics_mutex);
2910 entry = filter_statistics;
2911 while (entry != NULL)
2913 if (!strcmp(entry->filter, filter))
2915 *executions = entry->executions;
2916 *response_bodies_modified = entry->response_bodies_modified;
2917 *hits = entry->hits;
2920 entry = entry->next;
2923 privoxy_mutex_unlock(&filter_statistics_mutex);
2928 struct block_statistics_entry
2931 unsigned long long count;
2933 struct block_statistics_entry *next;
2936 static struct block_statistics_entry *block_statistics = NULL;
2938 /*********************************************************************
2940 * Function : register_block_reason_for_statistics
2942 * Description : Registers a block reason so we can gather statistics
2943 * for it unless the block reason has already been
2944 * registered before.
2947 * 1 : block_reason = Block reason to register
2951 *********************************************************************/
2952 void register_block_reason_for_statistics(const char *block_reason)
2954 struct block_statistics_entry *entry;
2956 privoxy_mutex_lock(&block_statistics_mutex);
2958 if (block_statistics == NULL)
2960 block_statistics = zalloc_or_die(sizeof(struct block_statistics_entry));
2961 entry = block_statistics;
2962 entry->block_reason = strdup_or_die(block_reason);
2963 privoxy_mutex_unlock(&block_statistics_mutex);
2966 entry = block_statistics;
2967 while (entry != NULL)
2969 if (!strcmp(entry->block_reason, block_reason))
2971 /* Already registered, nothing to do. */
2974 if (entry->next == NULL)
2976 entry->next = zalloc_or_die(sizeof(struct block_statistics_entry));
2977 entry->next->block_reason = strdup_or_die(block_reason);
2980 entry = entry->next;
2983 privoxy_mutex_unlock(&block_statistics_mutex);
2988 /*********************************************************************
2990 * Function : increment_block_reason_counter
2992 * Description : Updates the counter for a block reason.
2995 * 1 : block_reason = Block reason to count
2999 *********************************************************************/
3000 static void increment_block_reason_counter(const char *block_reason)
3002 struct block_statistics_entry *entry;
3004 privoxy_mutex_lock(&block_statistics_mutex);
3006 entry = block_statistics;
3007 while (entry != NULL)
3009 if (!strcmp(entry->block_reason, block_reason))
3014 entry = entry->next;
3017 privoxy_mutex_unlock(&block_statistics_mutex);
3022 /*********************************************************************
3024 * Function : get_block_reason_count
3026 * Description : Gets number of times a block reason was used.
3029 * 1 : block_reason = Block reason to get statistics for.
3030 * 2 : count = Storage for the number of times the block
3035 *********************************************************************/
3036 void get_block_reason_count(const char *block_reason, unsigned long long *count)
3038 struct block_statistics_entry *entry;
3040 privoxy_mutex_lock(&block_statistics_mutex);
3042 entry = block_statistics;
3043 while (entry != NULL)
3045 if (!strcmp(entry->block_reason, block_reason))
3047 *count = entry->count;
3050 entry = entry->next;
3053 privoxy_mutex_unlock(&block_statistics_mutex);
3057 #endif /* def FEATURE_EXTENDED_STATISTICS */