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 *new_url = NULL;
1201 char *redirection_string;
1203 if ((csp->action->flags & ACTION_REDIRECT))
1205 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1208 * If the redirection string begins with 's',
1209 * assume it's a pcrs command, otherwise treat it as
1210 * properly formatted URL and use it for the redirection
1213 * According to (the now obsolete) RFC 2616 section 14.30
1214 * the URL has to be absolute and if the user tries:
1215 * +redirect{sadly/this/will/be/parsed/as/pcrs_command.html}
1216 * she would get undefined results anyway.
1218 * RFC 7231 7.1.2 actually allows relative references,
1219 * but those start with a leading slash (RFC 3986 4.2) and
1220 * thus can't be mistaken for pcrs commands either.
1223 if (*redirection_string == 's')
1225 char *requested_url;
1227 #ifdef FEATURE_HTTPS_INSPECTION
1228 if (client_use_ssl(csp))
1232 requested_url = strdup_or_die("https://");
1233 err = string_append(&requested_url, csp->http->hostport);
1234 if (!err) err = string_append(&requested_url, csp->http->path);
1237 log_error(LOG_LEVEL_FATAL,
1238 "Failed to rebuild URL 'https://%s%s'",
1239 csp->http->hostport, csp->http->path);
1245 requested_url = csp->http->url;
1247 new_url = rewrite_url(requested_url, redirection_string);
1248 #ifdef FEATURE_HTTPS_INSPECTION
1249 if (client_use_ssl(csp))
1251 freez(requested_url);
1257 log_error(LOG_LEVEL_REDIRECTS,
1258 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1259 redirection_string);
1260 new_url = strdup(redirection_string);
1264 #ifdef FEATURE_FAST_REDIRECTS
1265 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1269 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1272 * If it exists, use the previously rewritten URL as input
1273 * otherwise just use the old path.
1275 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1276 new_url = get_last_url(old_url, redirect_mode);
1281 * Disable redirect checkers, so that they
1282 * will be only run more than once if the user
1283 * also enables them through tags.
1285 * From a performance point of view
1286 * it doesn't matter, but the duplicated
1287 * log messages are annoying.
1289 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1290 #endif /* def FEATURE_FAST_REDIRECTS */
1291 csp->action->flags &= ~ACTION_REDIRECT;
1293 /* Did any redirect action trigger? */
1296 if (url_requires_percent_encoding(new_url))
1299 log_error(LOG_LEVEL_REDIRECTS, "Percent-encoding redirect URL: %N",
1300 strlen(new_url), new_url);
1301 encoded_url = percent_encode_url(new_url);
1303 if (encoded_url == NULL)
1305 return cgi_error_memory();
1307 new_url = encoded_url;
1308 assert(FALSE == url_requires_percent_encoding(new_url));
1311 if (0 == strcmpic(new_url, csp->http->url))
1313 log_error(LOG_LEVEL_ERROR,
1314 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1315 csp->http->url, new_url);
1320 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1322 if (NULL == (rsp = alloc_http_response()))
1325 return cgi_error_memory();
1328 rsp->status = strdup_or_die("302 Local Redirect from Privoxy");
1329 if (enlist_unique_header(rsp->headers, "Location", new_url))
1332 free_http_response(rsp);
1333 return cgi_error_memory();
1335 rsp->crunch_reason = REDIRECTED;
1338 return finish_http_response(csp, rsp);
1342 /* Only reached if no redirect is required */
1348 #ifdef FEATURE_IMAGE_BLOCKING
1349 /*********************************************************************
1351 * Function : is_imageurl
1353 * Description : Given a URL, decide whether it should be treated
1354 * as image URL or not.
1357 * 1 : csp = Current client state (buffers, headers, etc...)
1359 * Returns : True (nonzero) if URL is an image URL, false (0)
1362 *********************************************************************/
1363 int is_imageurl(const struct client_state *csp)
1365 return ((csp->action->flags & ACTION_IMAGE) != 0);
1368 #endif /* def FEATURE_IMAGE_BLOCKING */
1371 #ifdef FEATURE_TRUST
1372 /*********************************************************************
1374 * Function : is_untrusted_url
1376 * Description : Should we "distrust" this URL (and block it)?
1378 * Yes if it matches a line in the trustfile, or if the
1379 * referrer matches a line starting with "+" in the
1384 * 1 : csp = Current client state (buffers, headers, etc...)
1386 * Returns : 0 => trusted, 1 => untrusted
1388 *********************************************************************/
1389 int is_untrusted_url(const struct client_state *csp)
1391 struct file_list *fl;
1392 struct block_spec *b;
1393 struct pattern_spec **trusted_url;
1394 struct http_request rhttp[1];
1395 const char * referer;
1399 * If we don't have a trustlist, we trust everybody
1401 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1406 memset(rhttp, '\0', sizeof(*rhttp));
1409 * Do we trust the request URL itself?
1411 for (b = b->next; b ; b = b->next)
1413 if (url_match(b->url, csp->http))
1419 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1421 /* no referrer was supplied */
1427 * If not, do we maybe trust its referrer?
1429 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
1435 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1437 if (url_match(*trusted_url, rhttp))
1439 /* if the URL's referrer is from a trusted referrer, then
1440 * add the target spec to the trustfile as an unblocked
1441 * domain and return 0 (which means it's OK).
1446 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1450 char * new_entry = strdup_or_die("~");
1452 string_append(&new_entry, csp->http->hostport);
1454 path = csp->http->path;
1455 if ( (path[0] == '/')
1457 && ((path_end = strchr(path + 2, '/')) != NULL))
1459 /* since this path points into a user's home space
1460 * be sure to include this spec in the trustfile.
1462 long path_len = path_end - path; /* save offset */
1463 path = strdup(path); /* Copy string */
1466 path_end = path + path_len; /* regenerate ptr to new buffer */
1467 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1469 string_join(&new_entry, path);
1473 * Give a reason for generating this entry.
1475 string_append(&new_entry, " # Trusted referrer was: ");
1476 string_append(&new_entry, referer);
1478 if (new_entry != NULL)
1480 if (-1 == fprintf(fp, "%s\n", new_entry))
1482 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1483 new_entry, csp->config->trustfile);
1489 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1490 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1497 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1498 csp->http->hostport, csp->config->trustfile);
1506 #endif /* def FEATURE_TRUST */
1509 /*********************************************************************
1511 * Function : get_filter
1513 * Description : Get a filter with a given name and type.
1514 * Note that taggers are filters, too.
1517 * 1 : csp = Current client state (buffers, headers, etc...)
1518 * 2 : requested_name = Name of the content filter to get
1519 * 3 : requested_type = Type of the filter to tagger to lookup
1521 * Returns : A pointer to the requested filter
1522 * or NULL if the filter wasn't found
1524 *********************************************************************/
1525 struct re_filterfile_spec *get_filter(const struct client_state *csp,
1526 const char *requested_name,
1527 enum filter_type requested_type)
1530 struct re_filterfile_spec *b;
1531 struct file_list *fl;
1533 for (i = 0; i < MAX_AF_FILES; i++)
1536 if ((NULL == fl) || (NULL == fl->f))
1539 * Either there are no filter files left or this
1540 * filter file just contains no valid filters.
1542 * Continue to be sure we don't miss valid filter
1543 * files that are chained after empty or invalid ones.
1548 for (b = fl->f; b != NULL; b = b->next)
1550 if (b->type != requested_type)
1552 /* The callers isn't interested in this filter type. */
1555 if (strcmp(b->name, requested_name) == 0)
1557 /* The requested filter has been found. Abort search. */
1563 /* No filter with the given name and type exists. */
1569 /*********************************************************************
1571 * Function : pcrs_filter_response
1573 * Description : Execute all text substitutions from all applying
1574 * +filter actions on the text buffer that's been
1575 * accumulated in csp->iob->buf.
1578 * 1 : csp = Current client state (buffers, headers, etc...)
1580 * Returns : a pointer to the (newly allocated) modified buffer.
1581 * or NULL if there were no hits or something went wrong
1583 *********************************************************************/
1584 static char *pcrs_filter_response(struct client_state *csp)
1587 size_t size, prev_size;
1593 struct re_filterfile_spec *b;
1594 struct list_entry *filtername;
1599 if (csp->iob->cur >= csp->iob->eod)
1604 if (filters_available(csp) == FALSE)
1606 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1607 "content filtering enabled, but no content filters available.");
1611 size = (size_t)(csp->iob->eod - csp->iob->cur);
1612 old = csp->iob->cur;
1615 * For all applying +filter actions, look if a filter by that
1616 * name exists and if yes, execute it's pcrs_joblist on the
1619 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1620 filtername != NULL; filtername = filtername->next)
1622 int current_hits = 0; /* Number of hits caused by this filter */
1623 int job_number = 0; /* Which job we're currently executing */
1624 int job_hits = 0; /* How many hits the current job caused */
1627 b = get_filter(csp, filtername->str, FT_CONTENT_FILTER);
1633 joblist = b->joblist;
1635 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1637 if (NULL == joblist)
1639 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1644 /* Apply all jobs from the joblist */
1645 for (job = joblist; NULL != job; job = job->next)
1648 job_hits = pcrs_execute(job, old, size, &new, &size);
1653 * That went well. Continue filtering
1654 * and use the result of this job as
1655 * input for the next one.
1657 current_hits += job_hits;
1658 if (old != csp->iob->cur)
1667 * This job caused an unexpected error. Inform the user
1668 * and skip the rest of the jobs in this filter. We could
1669 * continue with the next job, but usually the jobs
1670 * depend on each other or are similar enough to
1671 * fail for the same reason.
1673 * At the moment our pcrs expects the error codes of pcre 3.4,
1674 * but newer pcre versions can return additional error codes.
1675 * As a result pcrs_strerror()'s error message might be
1676 * "Unknown error ...", therefore we print the numerical value
1679 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1680 * should we use LOG_LEVEL_RE_FILTER instead?
1682 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1683 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1688 if (b->dynamic) pcrs_free_joblist(joblist);
1690 log_error(LOG_LEVEL_RE_FILTER,
1691 "filtering %s%s (size %lu) with \'%s\' produced %d hits (new size %lu).",
1692 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1693 #ifdef FEATURE_EXTENDED_STATISTICS
1694 update_filter_statistics(b->name, current_hits);
1696 hits += current_hits;
1700 * If there were no hits, destroy our copy and let
1701 * chat() use the original in csp->iob
1705 if (old != csp->iob->cur && old != new)
1713 csp->flags |= CSP_FLAG_MODIFIED;
1714 csp->content_length = size;
1715 clear_iob(csp->iob);
1722 #ifdef FEATURE_EXTERNAL_FILTERS
1723 /*********************************************************************
1725 * Function : get_external_filter
1727 * Description : Lookup the code to execute for an external filter.
1728 * Masks the misuse of the re_filterfile_spec.
1731 * 1 : csp = Current client state (buffers, headers, etc...)
1732 * 2 : name = Name of the content filter to get
1734 * Returns : A pointer to the requested code
1735 * or NULL if the filter wasn't found
1737 *********************************************************************/
1738 static const char *get_external_filter(const struct client_state *csp,
1741 struct re_filterfile_spec *external_filter;
1743 external_filter = get_filter(csp, name, FT_EXTERNAL_CONTENT_FILTER);
1744 if (external_filter == NULL)
1746 log_error(LOG_LEVEL_FATAL,
1747 "Didn't find stuff to execute for external filter: %s",
1751 return external_filter->patterns->first->str;
1756 /*********************************************************************
1758 * Function : set_privoxy_variables
1760 * Description : Sets a couple of privoxy-specific environment variables
1763 * 1 : csp = Current client state (buffers, headers, etc...)
1767 *********************************************************************/
1768 static void set_privoxy_variables(const struct client_state *csp)
1775 { "PRIVOXY_URL", csp->http->url },
1776 { "PRIVOXY_PATH", csp->http->path },
1777 { "PRIVOXY_HOST", csp->http->host },
1778 { "PRIVOXY_ORIGIN", csp->ip_addr_str },
1779 { "PRIVOXY_LISTEN_ADDRESS", csp->listen_addr_str },
1782 for (i = 0; i < SZ(env); i++)
1784 if (setenv(env[i].name, env[i].value, 1))
1786 log_error(LOG_LEVEL_ERROR, "Failed to set %s=%s: %E",
1787 env[i].name, env[i].value);
1793 /*********************************************************************
1795 * Function : execute_external_filter
1797 * Description : Pipe content into external filter and return the output
1800 * 1 : csp = Current client state (buffers, headers, etc...)
1801 * 2 : name = Name of the external filter to execute
1802 * 3 : content = The original content to filter
1803 * 4 : size = The size of the content buffer
1805 * Returns : a pointer to the (newly allocated) modified buffer.
1806 * or NULL if there were no hits or something went wrong
1808 *********************************************************************/
1809 static char *execute_external_filter(const struct client_state *csp,
1810 const char *name, char *content, size_t *size)
1813 char file_name[FILENAME_MAX];
1815 char *filter_output;
1819 const char *external_filter;
1821 if (csp->config->temporary_directory == NULL)
1823 log_error(LOG_LEVEL_ERROR,
1824 "No temporary-directory configured. Can't execute filter: %s",
1829 external_filter = get_external_filter(csp, name);
1831 if (sizeof(file_name) < snprintf(file_name, sizeof(file_name),
1832 "%s/privoxy-XXXXXXXX", csp->config->temporary_directory))
1834 log_error(LOG_LEVEL_ERROR, "temporary-directory path too long");
1838 fd = mkstemp(file_name);
1841 log_error(LOG_LEVEL_ERROR, "mkstemp() failed to create %s: %E", file_name);
1845 fp = fdopen(fd, "w");
1848 log_error(LOG_LEVEL_ERROR, "fdopen() failed: %E");
1854 * The size may be zero if a previous filter discarded everything.
1856 * This isn't necessary unintentional, so we just don't try
1857 * to fwrite() nothing and let the user deal with the rest.
1859 if ((*size != 0) && fwrite(content, *size, 1, fp) != 1)
1861 log_error(LOG_LEVEL_ERROR, "fwrite(..., %lu, 1, ..) failed: %E", *size);
1868 if (sizeof(cmd) < snprintf(cmd, sizeof(cmd), "%s < %s", external_filter, file_name))
1870 log_error(LOG_LEVEL_ERROR,
1871 "temporary-directory or external filter path too long");
1876 log_error(LOG_LEVEL_RE_FILTER, "Executing '%s': %s", name, cmd);
1879 * The locking is necessary to prevent other threads
1880 * from overwriting the environment variables before
1881 * the popen fork. Afterwards this no longer matters.
1883 privoxy_mutex_lock(&external_filter_mutex);
1884 set_privoxy_variables(csp);
1885 fp = popen(cmd, "r");
1886 privoxy_mutex_unlock(&external_filter_mutex);
1889 log_error(LOG_LEVEL_ERROR, "popen(\"%s\", \"r\") failed: %E", cmd);
1894 /* Allocate at least one byte */
1895 filter_output = malloc_or_die(*size + 1);
1898 while (!feof(fp) && !ferror(fp))
1901 /* Could be bigger ... */
1902 enum { READ_LENGTH = 2048 };
1904 if (new_size + READ_LENGTH >= *size)
1908 /* Could be considered wasteful if the content is 'large'. */
1909 *size += (*size >= READ_LENGTH) ? *size : READ_LENGTH;
1911 p = realloc(filter_output, *size);
1914 log_error(LOG_LEVEL_ERROR, "Out of memory while reading "
1915 "external filter output. Using what we got so far.");
1920 assert(new_size + READ_LENGTH < *size);
1921 len = fread(&filter_output[new_size], 1, READ_LENGTH, fp);
1931 log_error(LOG_LEVEL_ERROR, "Executing %s failed: %E", cmd);
1935 log_error(LOG_LEVEL_RE_FILTER,
1936 "Executing '%s' resulted in return value %d. "
1937 "Read %lu of up to %lu bytes.", name, (ret >> 8), new_size, *size);
1943 return filter_output;
1946 #endif /* def FEATURE_EXTERNAL_FILTERS */
1949 /*********************************************************************
1951 * Function : gif_deanimate_response
1953 * Description : Deanimate the GIF image that has been accumulated in
1954 * csp->iob->buf, set csp->content_length to the modified
1955 * size and raise the CSP_FLAG_MODIFIED flag.
1958 * 1 : csp = Current client state (buffers, headers, etc...)
1960 * Returns : a pointer to the (newly allocated) modified buffer.
1961 * or NULL in case something went wrong.
1963 *********************************************************************/
1965 char *gif_deanimate_response(struct client_state *csp)
1967 static char *gif_deanimate_response(struct client_state *csp)
1970 struct binbuffer *in, *out;
1974 size = (size_t)(csp->iob->eod - csp->iob->cur);
1976 in = zalloc_or_die(sizeof(*in));
1977 out = zalloc_or_die(sizeof(*out));
1979 in->buffer = csp->iob->cur;
1982 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1984 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1991 if ((int)size == out->offset)
1993 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1997 log_error(LOG_LEVEL_DEANIMATE,
1998 "Success! GIF shrunk from %lu bytes to %lu.", size, out->offset);
2000 csp->content_length = out->offset;
2001 csp->flags |= CSP_FLAG_MODIFIED;
2011 /*********************************************************************
2013 * Function : get_filter_function
2015 * Description : Decides which content filter function has
2016 * to be applied (if any). Only considers functions
2017 * for internal filters which are mutually-exclusive.
2020 * 1 : csp = Current client state (buffers, headers, etc...)
2022 * Returns : The content filter function to run, or
2023 * NULL if no content filter is active
2025 *********************************************************************/
2026 static filter_function_ptr get_filter_function(const struct client_state *csp)
2028 filter_function_ptr filter_function = NULL;
2031 * Choose the applying filter function based on
2032 * the content type and action settings.
2034 if ((csp->content_type & CT_TEXT) &&
2035 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2037 filter_function = pcrs_filter_response;
2039 else if ((csp->content_type & CT_GIF) &&
2040 (csp->action->flags & ACTION_DEANIMATE))
2042 filter_function = gif_deanimate_response;
2045 return filter_function;
2049 /*********************************************************************
2051 * Function : remove_chunked_transfer_coding
2053 * Description : In-situ remove the "chunked" transfer coding as defined
2054 * in RFC 7230 4.1 from a buffer. XXX: The implementation
2055 * is neither complete nor compliant (TODO #129).
2058 * 1 : buffer = Pointer to the text buffer
2059 * 2 : size = In: Number of bytes to be processed,
2060 * Out: Number of bytes after de-chunking.
2061 * (undefined in case of errors)
2063 * Returns : JB_ERR_OK for success,
2064 * JB_ERR_PARSE otherwise
2066 *********************************************************************/
2068 extern jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2070 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2074 unsigned int chunksize = 0;
2075 char *from_p, *to_p;
2076 const char *end_of_buffer = buffer + *size;
2080 log_error(LOG_LEVEL_FATAL, "Invalid chunked input. Buffer is empty.");
2081 return JB_ERR_PARSE;
2085 from_p = to_p = buffer;
2087 if (sscanf(buffer, "%x", &chunksize) != 1)
2089 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2090 return JB_ERR_PARSE;
2093 while (chunksize > 0U)
2096 * If the chunk-size is valid, we should have at least
2097 * chunk-size bytes of chunk-data and five bytes of
2098 * meta data (chunk-size, CRLF, CRLF) left in the buffer.
2100 if (chunksize + 5 >= *size - newsize)
2102 log_error(LOG_LEVEL_ERROR,
2103 "Chunk size %u exceeds buffered data left. "
2104 "Already digested %lu of %lu buffered bytes.",
2105 chunksize, newsize, *size);
2106 return JB_ERR_PARSE;
2110 * Skip the chunk-size, the optional chunk-ext and the CRLF
2111 * that is supposed to be located directly before the start
2114 if (NULL == (from_p = strstr(from_p, "\r\n")))
2116 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2117 return JB_ERR_PARSE;
2122 * The previous strstr() does not enforce chunk-validity
2123 * and is sattisfied as long a CRLF is left in the buffer.
2125 * Make sure the bytes we consider chunk-data are within
2128 if (from_p + chunksize >= end_of_buffer)
2130 log_error(LOG_LEVEL_ERROR,
2131 "End of chunk is beyond the end of the buffer.");
2132 return JB_ERR_PARSE;
2135 memmove(to_p, from_p, (size_t) chunksize);
2136 newsize += chunksize;
2137 to_p = buffer + newsize;
2138 from_p += chunksize;
2141 * Not merging this check with the previous one allows us
2142 * to keep chunks without trailing CRLF. It's not clear
2143 * if we actually have to care about those, though.
2145 if (from_p + 2 >= end_of_buffer)
2147 log_error(LOG_LEVEL_ERROR, "Not enough room for trailing CRLF.");
2148 return JB_ERR_PARSE;
2151 if (sscanf(from_p, "%x", &chunksize) != 1)
2153 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
2158 /* XXX: Should get its own loglevel. */
2159 log_error(LOG_LEVEL_RE_FILTER,
2160 "De-chunking successful. Shrunk from %lu to %lu", *size, newsize);
2169 /*********************************************************************
2171 * Function : prepare_for_filtering
2173 * Description : If necessary, de-chunks and decompresses
2174 * the content so it can get filterd.
2177 * 1 : csp = Current client state (buffers, headers, etc...)
2179 * Returns : JB_ERR_OK for success,
2180 * JB_ERR_PARSE otherwise
2182 *********************************************************************/
2183 static jb_err prepare_for_filtering(struct client_state *csp)
2185 jb_err err = JB_ERR_OK;
2188 * If the body has a "chunked" transfer-encoding,
2189 * get rid of it, adjusting size and iob->eod
2191 if (csp->flags & CSP_FLAG_CHUNKED)
2193 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2195 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2196 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2197 if (JB_ERR_OK == err)
2199 csp->iob->eod = csp->iob->cur + size;
2200 csp->flags |= CSP_FLAG_MODIFIED;
2204 return JB_ERR_PARSE;
2210 * If the body has a supported transfer-encoding,
2211 * decompress it, adjusting size and iob->eod.
2213 if ((csp->content_type & (CT_GZIP|CT_DEFLATE))
2214 #ifdef FEATURE_BROTLI
2215 || (csp->content_type & CT_BROTLI)
2219 if (0 == csp->iob->eod - csp->iob->cur)
2221 /* Nothing left after de-chunking. */
2225 err = decompress_iob(csp);
2227 if (JB_ERR_OK == err)
2229 csp->flags |= CSP_FLAG_MODIFIED;
2230 csp->content_type &= ~CT_TABOO;
2235 * Unset content types to remember not to
2236 * modify the Content-Encoding header later.
2238 csp->content_type &= ~CT_GZIP;
2239 csp->content_type &= ~CT_DEFLATE;
2240 #ifdef FEATURE_BROTLI
2241 csp->content_type &= ~CT_BROTLI;
2251 /*********************************************************************
2253 * Function : execute_content_filters
2255 * Description : Executes a given content filter.
2258 * 1 : csp = Current client state (buffers, headers, etc...)
2260 * Returns : Pointer to the modified buffer, or
2261 * NULL if filtering failed or wasn't necessary.
2263 *********************************************************************/
2264 char *execute_content_filters(struct client_state *csp)
2267 filter_function_ptr content_filter;
2269 assert(content_filters_enabled(csp->action));
2271 if (0 == csp->iob->eod - csp->iob->cur)
2274 * No content (probably status code 301, 302 ...),
2275 * no filtering necessary.
2280 if (JB_ERR_OK != prepare_for_filtering(csp))
2283 * failed to de-chunk or decompress.
2288 if (0 == csp->iob->eod - csp->iob->cur)
2291 * Clown alarm: chunked and/or compressed nothing delivered.
2296 content_filter = get_filter_function(csp);
2297 content = (content_filter != NULL) ? (*content_filter)(csp) : NULL;
2299 #ifdef FEATURE_EXTERNAL_FILTERS
2300 if ((csp->content_type & CT_TEXT) &&
2301 !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]))
2303 struct list_entry *filtername;
2304 size_t size = (size_t)csp->content_length;
2306 if (content == NULL)
2308 content = csp->iob->cur;
2309 size = (size_t)(csp->iob->eod - csp->iob->cur);
2312 for (filtername = csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]->first;
2313 filtername ; filtername = filtername->next)
2315 char *result = execute_external_filter(csp, filtername->str, content, &size);
2318 if (content != csp->iob->cur)
2325 csp->flags |= CSP_FLAG_MODIFIED;
2326 csp->content_length = size;
2328 #endif /* def FEATURE_EXTERNAL_FILTERS */
2335 /*********************************************************************
2337 * Function : get_url_actions
2339 * Description : Gets the actions for this URL.
2342 * 1 : csp = Current client state (buffers, headers, etc...)
2343 * 2 : http = http_request request for blocked URLs
2347 *********************************************************************/
2348 void get_url_actions(struct client_state *csp, struct http_request *http)
2350 struct file_list *fl;
2351 struct url_actions *b;
2354 init_current_action(csp->action);
2356 for (i = 0; i < MAX_AF_FILES; i++)
2358 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2363 #ifdef FEATURE_CLIENT_TAGS
2364 apply_url_actions(csp->action, http, csp->client_tags, b);
2366 apply_url_actions(csp->action, http, b);
2373 /*********************************************************************
2375 * Function : apply_url_actions
2377 * Description : Applies a list of URL actions.
2380 * 1 : action = Destination.
2381 * 2 : http = Current URL
2382 * 3 : client_tags = list of client tags
2383 * 4 : b = list of URL actions to apply
2387 *********************************************************************/
2388 static void apply_url_actions(struct current_action_spec *action,
2389 struct http_request *http,
2390 #ifdef FEATURE_CLIENT_TAGS
2391 const struct list *client_tags,
2393 struct url_actions *b)
2397 /* Should never happen */
2401 for (b = b->next; NULL != b; b = b->next)
2403 if (url_match(b->url, http))
2405 merge_current_action(action, b->action);
2407 #ifdef FEATURE_CLIENT_TAGS
2408 if (client_tag_match(b->url, client_tags))
2410 merge_current_action(action, b->action);
2417 /*********************************************************************
2419 * Function : get_forward_override_settings
2421 * Description : Returns forward settings as specified with the
2422 * forward-override{} action. forward-override accepts
2423 * forward lines similar to the one used in the
2424 * configuration file, but without the URL pattern.
2430 * in the configuration file can be replaced with
2431 * the action section:
2433 * {+forward-override{forward .}}
2437 * 1 : csp = Current client state (buffers, headers, etc...)
2439 * Returns : Pointer to forwarding structure in case of success.
2440 * Invalid syntax is fatal.
2442 *********************************************************************/
2443 static const struct forward_spec *get_forward_override_settings(struct client_state *csp)
2445 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2446 char forward_settings[BUFFER_SIZE];
2447 char *http_parent = NULL;
2448 /* variable names were chosen for consistency reasons. */
2449 struct forward_spec *fwd = NULL;
2453 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2454 /* Should be enforced by load_one_actions_file() */
2455 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2457 /* Create a copy ssplit can modify */
2458 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2460 if (NULL != csp->fwd)
2463 * XXX: Currently necessary to prevent memory
2464 * leaks when the show-url-info cgi page is visited.
2466 unload_forward_spec(csp->fwd);
2470 * allocate a new forward node, valid only for
2471 * the lifetime of this request. Save its location
2472 * in csp as well, so sweep() can free it later on.
2474 fwd = csp->fwd = zalloc_or_die(sizeof(*fwd));
2476 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec));
2477 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2479 fwd->type = SOCKS_NONE;
2481 /* Parse the parent HTTP proxy host:port */
2482 http_parent = vec[1];
2485 else if ((vec_count == 2) && !strcasecmp(vec[0], "forward-webserver"))
2487 fwd->type = FORWARD_WEBSERVER;
2489 /* Parse the parent HTTP server host:port */
2490 http_parent = vec[1];
2493 else if (vec_count == 3)
2495 char *socks_proxy = NULL;
2497 if (!strcasecmp(vec[0], "forward-socks4"))
2499 fwd->type = SOCKS_4;
2500 socks_proxy = vec[1];
2502 else if (!strcasecmp(vec[0], "forward-socks4a"))
2504 fwd->type = SOCKS_4A;
2505 socks_proxy = vec[1];
2507 else if (!strcasecmp(vec[0], "forward-socks5"))
2509 fwd->type = SOCKS_5;
2510 socks_proxy = vec[1];
2512 else if (!strcasecmp(vec[0], "forward-socks5t"))
2514 fwd->type = SOCKS_5T;
2515 socks_proxy = vec[1];
2518 if (NULL != socks_proxy)
2520 /* Parse the SOCKS proxy [user:pass@]host[:port] */
2521 fwd->gateway_port = 1080;
2522 parse_forwarder_address(socks_proxy,
2523 &fwd->gateway_host, &fwd->gateway_port,
2524 &fwd->auth_username, &fwd->auth_password);
2526 http_parent = vec[2];
2530 if (NULL == http_parent)
2532 log_error(LOG_LEVEL_FATAL,
2533 "Invalid forward-override syntax in: %s", forward_override_line);
2534 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2537 /* Parse http forwarding settings */
2538 if (strcmp(http_parent, ".") != 0)
2540 fwd->forward_port = 8000;
2541 parse_forwarder_address(http_parent,
2542 &fwd->forward_host, &fwd->forward_port,
2546 assert (NULL != fwd);
2548 log_error(LOG_LEVEL_CONNECT,
2549 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2555 /*********************************************************************
2557 * Function : forward_url
2559 * Description : Should we forward this to another proxy?
2562 * 1 : csp = Current client state (buffers, headers, etc...)
2563 * 2 : http = http_request request for current URL
2565 * Returns : Pointer to forwarding information.
2567 *********************************************************************/
2568 const struct forward_spec *forward_url(struct client_state *csp,
2569 const struct http_request *http)
2571 static const struct forward_spec fwd_default[1]; /* Zero'ed due to being static. */
2572 struct forward_spec *fwd = csp->config->forward;
2574 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2576 return get_forward_override_settings(csp);
2586 if (url_match(fwd->url, http))
2597 /*********************************************************************
2599 * Function : direct_response
2601 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2602 * request and if so, return a HTTP 501 to the client.
2604 * FIXME: I have a stupid name and I should handle the
2605 * requests properly. Still, what we do here is rfc-
2606 * compliant, whereas ignoring or forwarding are not.
2609 * 1 : csp = Current client state (buffers, headers, etc...)
2611 * Returns : http_response if , NULL if nonmatch or handler fail
2613 *********************************************************************/
2614 struct http_response *direct_response(struct client_state *csp)
2616 struct http_response *rsp;
2617 struct list_entry *p;
2619 if ((0 == strcmpic(csp->http->gpc, "trace"))
2620 || (0 == strcmpic(csp->http->gpc, "options")))
2622 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2624 if (!strncmpic(p->str, "Max-Forwards:", 13))
2626 unsigned int max_forwards;
2629 * If it's a Max-Forwards value of zero,
2630 * we have to intercept the request.
2632 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2635 * FIXME: We could handle at least TRACE here,
2636 * but that would require a verbatim copy of
2637 * the request which we don't have anymore
2639 log_error(LOG_LEVEL_HEADER,
2640 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2643 /* Get mem for response or fail*/
2644 if (NULL == (rsp = alloc_http_response()))
2646 return cgi_error_memory();
2649 rsp->status = strdup_or_die("501 Not Implemented");
2651 rsp->crunch_reason = UNSUPPORTED;
2653 return(finish_http_response(csp, rsp));
2662 /*********************************************************************
2664 * Function : content_requires_filtering
2666 * Description : Checks whether there are any content filters
2667 * enabled for the current request and if they
2668 * can actually be applied..
2671 * 1 : csp = Current client state (buffers, headers, etc...)
2673 * Returns : TRUE for yes, FALSE otherwise
2675 *********************************************************************/
2676 int content_requires_filtering(struct client_state *csp)
2678 if ((csp->content_type & CT_TABOO)
2679 && !(csp->action->flags & ACTION_FORCE_TEXT_MODE))
2685 * Are we enabling text mode by force?
2687 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2690 * Do we really have to?
2692 if (csp->content_type & CT_TEXT)
2694 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2698 csp->content_type |= CT_TEXT;
2699 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2703 if (!(csp->content_type & CT_DECLARED))
2706 * The server didn't bother to declare a MIME-Type.
2707 * Assume it's text that can be filtered.
2709 * This also regularly happens with 304 responses,
2710 * therefore logging anything here would cause
2713 csp->content_type |= CT_TEXT;
2717 * Choose the applying filter function based on
2718 * the content type and action settings.
2720 if ((csp->content_type & CT_TEXT) &&
2721 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER]) ||
2722 !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER])))
2726 else if ((csp->content_type & CT_GIF) &&
2727 (csp->action->flags & ACTION_DEANIMATE))
2737 /*********************************************************************
2739 * Function : content_filters_enabled
2741 * Description : Checks whether there are any content filters
2742 * enabled for the current request.
2745 * 1 : action = Action spec to check.
2747 * Returns : TRUE for yes, FALSE otherwise
2749 *********************************************************************/
2750 int content_filters_enabled(const struct current_action_spec *action)
2752 return ((action->flags & ACTION_DEANIMATE) ||
2753 !list_is_empty(action->multi[ACTION_MULTI_FILTER]) ||
2754 !list_is_empty(action->multi[ACTION_MULTI_EXTERNAL_FILTER]));
2758 /*********************************************************************
2760 * Function : filters_available
2762 * Description : Checks whether there are any filters available.
2765 * 1 : csp = Current client state (buffers, headers, etc...)
2767 * Returns : TRUE for yes, FALSE otherwise.
2769 *********************************************************************/
2770 int filters_available(const struct client_state *csp)
2773 for (i = 0; i < MAX_AF_FILES; i++)
2775 const struct file_list *fl = csp->rlist[i];
2776 if ((NULL != fl) && (NULL != fl->f))
2784 #ifdef FEATURE_EXTENDED_STATISTICS
2786 struct filter_statistics_entry
2789 unsigned long long executions;
2790 unsigned long long response_bodies_modified;
2791 unsigned long long hits;
2793 struct filter_statistics_entry *next;
2796 static struct filter_statistics_entry *filter_statistics = NULL;
2799 /*********************************************************************
2801 * Function : register_filter_for_statistics
2803 * Description : Registers a filter so we can gather statistics for
2804 * it unless the filter has already been registered
2808 * 1 : filter = Name of the filter to register
2812 *********************************************************************/
2813 void register_filter_for_statistics(const char *filter)
2815 struct filter_statistics_entry *entry;
2817 privoxy_mutex_lock(&filter_statistics_mutex);
2819 if (filter_statistics == NULL)
2821 filter_statistics = zalloc_or_die(sizeof(struct filter_statistics_entry));
2822 entry = filter_statistics;
2823 entry->filter = strdup_or_die(filter);
2824 privoxy_mutex_unlock(&filter_statistics_mutex);
2827 entry = filter_statistics;
2828 while (entry != NULL)
2830 if (!strcmp(entry->filter, filter))
2832 /* Already registered, nothing to do. */
2835 if (entry->next == NULL)
2837 entry->next = zalloc_or_die(sizeof(struct filter_statistics_entry));
2838 entry->next->filter = strdup_or_die(filter);
2841 entry = entry->next;
2844 privoxy_mutex_unlock(&filter_statistics_mutex);
2849 /*********************************************************************
2851 * Function : update_filter_statistics
2853 * Description : Updates the statistics for a filter.
2856 * 1 : filter = Name of the filter to update
2857 * 2 : hits = Hit count.
2861 *********************************************************************/
2862 void update_filter_statistics(const char *filter, int hits)
2864 struct filter_statistics_entry *entry;
2866 privoxy_mutex_lock(&filter_statistics_mutex);
2868 entry = filter_statistics;
2869 while (entry != NULL)
2871 if (!strcmp(entry->filter, filter))
2873 entry->executions++;
2876 entry->response_bodies_modified++;
2877 entry->hits += (unsigned)hits;
2881 entry = entry->next;
2884 privoxy_mutex_unlock(&filter_statistics_mutex);
2889 /*********************************************************************
2891 * Function : get_filter_statistics
2893 * Description : Gets the statistics for a filter.
2896 * 1 : filter = Name of the filter to get statistics for.
2897 * 2 : executions = Storage for the execution count.
2898 * 3 : response_bodies_modified = Storage for the number
2899 * of modified response bodies.
2900 * 4 : hits = Storage for the number of hits.
2904 *********************************************************************/
2905 void get_filter_statistics(const char *filter, unsigned long long *executions,
2906 unsigned long long *response_bodies_modified,
2907 unsigned long long *hits)
2909 struct filter_statistics_entry *entry;
2911 privoxy_mutex_lock(&filter_statistics_mutex);
2913 entry = filter_statistics;
2914 while (entry != NULL)
2916 if (!strcmp(entry->filter, filter))
2918 *executions = entry->executions;
2919 *response_bodies_modified = entry->response_bodies_modified;
2920 *hits = entry->hits;
2923 entry = entry->next;
2926 privoxy_mutex_unlock(&filter_statistics_mutex);
2931 struct block_statistics_entry
2934 unsigned long long count;
2936 struct block_statistics_entry *next;
2939 static struct block_statistics_entry *block_statistics = NULL;
2941 /*********************************************************************
2943 * Function : register_block_reason_for_statistics
2945 * Description : Registers a block reason so we can gather statistics
2946 * for it unless the block reason has already been
2947 * registered before.
2950 * 1 : block_reason = Block reason to register
2954 *********************************************************************/
2955 void register_block_reason_for_statistics(const char *block_reason)
2957 struct block_statistics_entry *entry;
2959 privoxy_mutex_lock(&block_statistics_mutex);
2961 if (block_statistics == NULL)
2963 block_statistics = zalloc_or_die(sizeof(struct block_statistics_entry));
2964 entry = block_statistics;
2965 entry->block_reason = strdup_or_die(block_reason);
2966 privoxy_mutex_unlock(&block_statistics_mutex);
2969 entry = block_statistics;
2970 while (entry != NULL)
2972 if (!strcmp(entry->block_reason, block_reason))
2974 /* Already registered, nothing to do. */
2977 if (entry->next == NULL)
2979 entry->next = zalloc_or_die(sizeof(struct block_statistics_entry));
2980 entry->next->block_reason = strdup_or_die(block_reason);
2983 entry = entry->next;
2986 privoxy_mutex_unlock(&block_statistics_mutex);
2991 /*********************************************************************
2993 * Function : increment_block_reason_counter
2995 * Description : Updates the counter for a block reason.
2998 * 1 : block_reason = Block reason to count
3002 *********************************************************************/
3003 static void increment_block_reason_counter(const char *block_reason)
3005 struct block_statistics_entry *entry;
3007 privoxy_mutex_lock(&block_statistics_mutex);
3009 entry = block_statistics;
3010 while (entry != NULL)
3012 if (!strcmp(entry->block_reason, block_reason))
3017 entry = entry->next;
3020 privoxy_mutex_unlock(&block_statistics_mutex);
3025 /*********************************************************************
3027 * Function : get_block_reason_count
3029 * Description : Gets number of times a block reason was used.
3032 * 1 : block_reason = Block reason to get statistics for.
3033 * 2 : count = Storage for the number of times the block
3038 *********************************************************************/
3039 void get_block_reason_count(const char *block_reason, unsigned long long *count)
3041 struct block_statistics_entry *entry;
3043 privoxy_mutex_lock(&block_statistics_mutex);
3045 entry = block_statistics;
3046 while (entry != NULL)
3048 if (!strcmp(entry->block_reason, block_reason))
3050 *count = entry->count;
3053 entry = entry->next;
3056 privoxy_mutex_unlock(&block_statistics_mutex);
3060 #endif /* def FEATURE_EXTENDED_STATISTICS */