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 (requested_url != csp->http->url)
1251 assert(client_use_ssl(csp));
1252 freez(requested_url);
1258 log_error(LOG_LEVEL_REDIRECTS,
1259 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1260 redirection_string);
1261 new_url = strdup(redirection_string);
1265 #ifdef FEATURE_FAST_REDIRECTS
1266 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1270 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1273 * If it exists, use the previously rewritten URL as input
1274 * otherwise just use the old path.
1276 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1277 new_url = get_last_url(old_url, redirect_mode);
1280 #endif /* def FEATURE_FAST_REDIRECTS */
1282 /* Did any redirect action trigger? */
1285 if (url_requires_percent_encoding(new_url))
1288 log_error(LOG_LEVEL_REDIRECTS, "Percent-encoding redirect URL: %N",
1289 strlen(new_url), new_url);
1290 encoded_url = percent_encode_url(new_url);
1292 if (encoded_url == NULL)
1294 return cgi_error_memory();
1296 new_url = encoded_url;
1297 assert(FALSE == url_requires_percent_encoding(new_url));
1300 if (0 == strcmpic(new_url, csp->http->url))
1302 log_error(LOG_LEVEL_ERROR,
1303 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1304 csp->http->url, new_url);
1309 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1311 if (NULL == (rsp = alloc_http_response()))
1314 return cgi_error_memory();
1317 rsp->status = strdup_or_die("302 Local Redirect from Privoxy");
1318 if (enlist_unique_header(rsp->headers, "Location", new_url))
1321 free_http_response(rsp);
1322 return cgi_error_memory();
1324 rsp->crunch_reason = REDIRECTED;
1327 return finish_http_response(csp, rsp);
1331 /* Only reached if no redirect is required */
1337 #ifdef FEATURE_IMAGE_BLOCKING
1338 /*********************************************************************
1340 * Function : is_imageurl
1342 * Description : Given a URL, decide whether it should be treated
1343 * as image URL or not.
1346 * 1 : csp = Current client state (buffers, headers, etc...)
1348 * Returns : True (nonzero) if URL is an image URL, false (0)
1351 *********************************************************************/
1352 int is_imageurl(const struct client_state *csp)
1354 return ((csp->action->flags & ACTION_IMAGE) != 0);
1357 #endif /* def FEATURE_IMAGE_BLOCKING */
1360 #ifdef FEATURE_TRUST
1361 /*********************************************************************
1363 * Function : is_untrusted_url
1365 * Description : Should we "distrust" this URL (and block it)?
1367 * Yes if it matches a line in the trustfile, or if the
1368 * referrer matches a line starting with "+" in the
1373 * 1 : csp = Current client state (buffers, headers, etc...)
1375 * Returns : 0 => trusted, 1 => untrusted
1377 *********************************************************************/
1378 int is_untrusted_url(const struct client_state *csp)
1380 struct file_list *fl;
1381 struct block_spec *b;
1382 struct pattern_spec **trusted_url;
1383 struct http_request rhttp[1];
1384 const char * referer;
1388 * If we don't have a trustlist, we trust everybody
1390 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1395 memset(rhttp, '\0', sizeof(*rhttp));
1398 * Do we trust the request URL itself?
1400 for (b = b->next; b ; b = b->next)
1402 if (url_match(b->url, csp->http))
1408 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1410 /* no referrer was supplied */
1416 * If not, do we maybe trust its referrer?
1418 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
1424 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1426 if (url_match(*trusted_url, rhttp))
1428 /* if the URL's referrer is from a trusted referrer, then
1429 * add the target spec to the trustfile as an unblocked
1430 * domain and return 0 (which means it's OK).
1435 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1439 char * new_entry = strdup_or_die("~");
1441 string_append(&new_entry, csp->http->hostport);
1443 path = csp->http->path;
1444 if ( (path[0] == '/')
1446 && ((path_end = strchr(path + 2, '/')) != NULL))
1448 /* since this path points into a user's home space
1449 * be sure to include this spec in the trustfile.
1451 long path_len = path_end - path; /* save offset */
1452 path = strdup(path); /* Copy string */
1455 path_end = path + path_len; /* regenerate ptr to new buffer */
1456 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1458 string_join(&new_entry, path);
1462 * Give a reason for generating this entry.
1464 string_append(&new_entry, " # Trusted referrer was: ");
1465 string_append(&new_entry, referer);
1467 if (new_entry != NULL)
1469 if (-1 == fprintf(fp, "%s\n", new_entry))
1471 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1472 new_entry, csp->config->trustfile);
1478 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1479 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1486 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1487 csp->http->hostport, csp->config->trustfile);
1495 #endif /* def FEATURE_TRUST */
1498 /*********************************************************************
1500 * Function : get_filter
1502 * Description : Get a filter with a given name and type.
1503 * Note that taggers are filters, too.
1506 * 1 : csp = Current client state (buffers, headers, etc...)
1507 * 2 : requested_name = Name of the content filter to get
1508 * 3 : requested_type = Type of the filter to tagger to lookup
1510 * Returns : A pointer to the requested filter
1511 * or NULL if the filter wasn't found
1513 *********************************************************************/
1514 struct re_filterfile_spec *get_filter(const struct client_state *csp,
1515 const char *requested_name,
1516 enum filter_type requested_type)
1519 struct re_filterfile_spec *b;
1520 struct file_list *fl;
1522 for (i = 0; i < MAX_AF_FILES; i++)
1525 if ((NULL == fl) || (NULL == fl->f))
1528 * Either there are no filter files left or this
1529 * filter file just contains no valid filters.
1531 * Continue to be sure we don't miss valid filter
1532 * files that are chained after empty or invalid ones.
1537 for (b = fl->f; b != NULL; b = b->next)
1539 if (b->type != requested_type)
1541 /* The callers isn't interested in this filter type. */
1544 if (strcmp(b->name, requested_name) == 0)
1546 /* The requested filter has been found. Abort search. */
1552 /* No filter with the given name and type exists. */
1558 /*********************************************************************
1560 * Function : pcrs_filter_impl
1562 * Description : Execute all text substitutions from all applying
1563 * (based on filter_response_body value) +filter
1564 * or +client_body_filter actions on the given buffer.
1567 * 1 : csp = Current client state (buffers, headers, etc...)
1568 * 2 : filter_response_body = when TRUE execute +filter
1569 * actions; execute +client_body_filter actions otherwise
1570 * 3 : data = Target data
1571 * 4 : data_len = Target data len
1573 * Returns : a pointer to the (newly allocated) modified buffer.
1574 * or NULL if there were no hits or something went wrong
1576 *********************************************************************/
1577 static char *pcrs_filter_impl(const struct client_state *csp, int filter_response_body,
1578 const char *data, size_t *data_len)
1581 size_t size, prev_size;
1582 const int filters_idx =
1583 filter_response_body ? ACTION_MULTI_FILTER : ACTION_MULTI_CLIENT_BODY_FILTER;
1584 const enum filter_type filter_type =
1585 filter_response_body ? FT_CONTENT_FILTER : FT_CLIENT_BODY_FILTER;
1587 const char *old = NULL;
1591 struct re_filterfile_spec *b;
1592 struct list_entry *filtername;
1602 if (filters_available(csp) == FALSE)
1604 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1605 "content filtering enabled, but no content filters available.");
1613 * For all applying actions, look if a filter by that
1614 * name exists and if yes, execute it's pcrs_joblist on the
1617 for (filtername = csp->action->multi[filters_idx]->first;
1618 filtername != NULL; filtername = filtername->next)
1620 int current_hits = 0; /* Number of hits caused by this filter */
1621 int job_number = 0; /* Which job we're currently executing */
1622 int job_hits = 0; /* How many hits the current job caused */
1625 b = get_filter(csp, filtername->str, filter_type);
1631 joblist = b->joblist;
1633 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1635 if (NULL == joblist)
1637 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1642 /* Apply all jobs from the joblist */
1643 for (job = joblist; NULL != job; job = job->next)
1646 job_hits = pcrs_execute(job, old, size, &new, &size);
1651 * That went well. Continue filtering
1652 * and use the result of this job as
1653 * input for the next one.
1655 current_hits += job_hits;
1665 * This job caused an unexpected error. Inform the user
1666 * and skip the rest of the jobs in this filter. We could
1667 * continue with the next job, but usually the jobs
1668 * depend on each other or are similar enough to
1669 * fail for the same reason.
1671 * At the moment our pcrs expects the error codes of pcre 3.4,
1672 * but newer pcre versions can return additional error codes.
1673 * As a result pcrs_strerror()'s error message might be
1674 * "Unknown error ...", therefore we print the numerical value
1677 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1678 * should we use LOG_LEVEL_RE_FILTER instead?
1680 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1681 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1686 if (b->dynamic) pcrs_free_joblist(joblist);
1688 if (filter_response_body)
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);
1696 log_error(LOG_LEVEL_RE_FILTER, "filtering request body from client %s "
1697 "(size %lu) with \'%s\' produced %d hits (new size %lu).",
1698 csp->ip_addr_str, prev_size, b->name, current_hits, size);
1700 #ifdef FEATURE_EXTENDED_STATISTICS
1701 update_filter_statistics(b->name, current_hits);
1703 hits += current_hits;
1707 * If there were no hits, destroy our copy and let
1708 * chat() use the original content
1712 if (old != data && old != new)
1725 /*********************************************************************
1727 * Function : pcrs_filter_response_body
1729 * Description : Execute all text substitutions from all applying
1730 * +filter actions on the text buffer that's been
1731 * accumulated in csp->iob->buf.
1734 * 1 : csp = Current client state (buffers, headers, etc...)
1736 * Returns : a pointer to the (newly allocated) modified buffer.
1737 * or NULL if there were no hits or something went wrong
1739 *********************************************************************/
1740 static char *pcrs_filter_response_body(struct client_state *csp)
1742 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
1749 if (csp->iob->cur >= csp->iob->eod)
1754 new = pcrs_filter_impl(csp, TRUE, csp->iob->cur, &size);
1758 csp->flags |= CSP_FLAG_MODIFIED;
1759 csp->content_length = size;
1760 clear_iob(csp->iob);
1767 #ifdef FEATURE_EXTERNAL_FILTERS
1768 /*********************************************************************
1770 * Function : get_external_filter
1772 * Description : Lookup the code to execute for an external filter.
1773 * Masks the misuse of the re_filterfile_spec.
1776 * 1 : csp = Current client state (buffers, headers, etc...)
1777 * 2 : name = Name of the content filter to get
1779 * Returns : A pointer to the requested code
1780 * or NULL if the filter wasn't found
1782 *********************************************************************/
1783 static const char *get_external_filter(const struct client_state *csp,
1786 struct re_filterfile_spec *external_filter;
1788 external_filter = get_filter(csp, name, FT_EXTERNAL_CONTENT_FILTER);
1789 if (external_filter == NULL)
1791 log_error(LOG_LEVEL_FATAL,
1792 "Didn't find stuff to execute for external filter: %s",
1796 return external_filter->patterns->first->str;
1801 /*********************************************************************
1803 * Function : set_privoxy_variables
1805 * Description : Sets a couple of privoxy-specific environment variables
1808 * 1 : csp = Current client state (buffers, headers, etc...)
1812 *********************************************************************/
1813 static void set_privoxy_variables(const struct client_state *csp)
1820 { "PRIVOXY_URL", csp->http->url },
1821 { "PRIVOXY_PATH", csp->http->path },
1822 { "PRIVOXY_HOST", csp->http->host },
1823 { "PRIVOXY_ORIGIN", csp->ip_addr_str },
1824 { "PRIVOXY_LISTEN_ADDRESS", csp->listen_addr_str },
1827 for (i = 0; i < SZ(env); i++)
1829 if (setenv(env[i].name, env[i].value, 1))
1831 log_error(LOG_LEVEL_ERROR, "Failed to set %s=%s: %E",
1832 env[i].name, env[i].value);
1838 /*********************************************************************
1840 * Function : execute_external_filter
1842 * Description : Pipe content into external filter and return the output
1845 * 1 : csp = Current client state (buffers, headers, etc...)
1846 * 2 : name = Name of the external filter to execute
1847 * 3 : content = The original content to filter
1848 * 4 : size = The size of the content buffer
1850 * Returns : a pointer to the (newly allocated) modified buffer.
1851 * or NULL if there were no hits or something went wrong
1853 *********************************************************************/
1854 static char *execute_external_filter(const struct client_state *csp,
1855 const char *name, char *content, size_t *size)
1858 char file_name[FILENAME_MAX];
1860 char *filter_output;
1864 const char *external_filter;
1866 if (csp->config->temporary_directory == NULL)
1868 log_error(LOG_LEVEL_ERROR,
1869 "No temporary-directory configured. Can't execute filter: %s",
1874 external_filter = get_external_filter(csp, name);
1876 if (sizeof(file_name) < snprintf(file_name, sizeof(file_name),
1877 "%s/privoxy-XXXXXXXX", csp->config->temporary_directory))
1879 log_error(LOG_LEVEL_ERROR, "temporary-directory path too long");
1883 fd = mkstemp(file_name);
1886 log_error(LOG_LEVEL_ERROR, "mkstemp() failed to create %s: %E", file_name);
1890 fp = fdopen(fd, "w");
1893 log_error(LOG_LEVEL_ERROR, "fdopen() failed: %E");
1899 * The size may be zero if a previous filter discarded everything.
1901 * This isn't necessary unintentional, so we just don't try
1902 * to fwrite() nothing and let the user deal with the rest.
1904 if ((*size != 0) && fwrite(content, *size, 1, fp) != 1)
1906 log_error(LOG_LEVEL_ERROR, "fwrite(..., %lu, 1, ..) failed: %E", *size);
1913 if (sizeof(cmd) < snprintf(cmd, sizeof(cmd), "%s < %s", external_filter, file_name))
1915 log_error(LOG_LEVEL_ERROR,
1916 "temporary-directory or external filter path too long");
1921 log_error(LOG_LEVEL_RE_FILTER, "Executing '%s': %s", name, cmd);
1924 * The locking is necessary to prevent other threads
1925 * from overwriting the environment variables before
1926 * the popen fork. Afterwards this no longer matters.
1928 privoxy_mutex_lock(&external_filter_mutex);
1929 set_privoxy_variables(csp);
1930 fp = popen(cmd, "r");
1931 privoxy_mutex_unlock(&external_filter_mutex);
1934 log_error(LOG_LEVEL_ERROR, "popen(\"%s\", \"r\") failed: %E", cmd);
1939 /* Allocate at least one byte */
1940 filter_output = malloc_or_die(*size + 1);
1943 while (!feof(fp) && !ferror(fp))
1946 /* Could be bigger ... */
1947 enum { READ_LENGTH = 2048 };
1949 if (new_size + READ_LENGTH >= *size)
1953 /* Could be considered wasteful if the content is 'large'. */
1954 *size += (*size >= READ_LENGTH) ? *size : READ_LENGTH;
1956 p = realloc(filter_output, *size);
1959 log_error(LOG_LEVEL_ERROR, "Out of memory while reading "
1960 "external filter output. Using what we got so far.");
1965 assert(new_size + READ_LENGTH < *size);
1966 len = fread(&filter_output[new_size], 1, READ_LENGTH, fp);
1976 log_error(LOG_LEVEL_ERROR, "Executing %s failed: %E", cmd);
1980 log_error(LOG_LEVEL_RE_FILTER,
1981 "Executing '%s' resulted in return value %d. "
1982 "Read %lu of up to %lu bytes.", name, (ret >> 8), new_size, *size);
1988 return filter_output;
1991 #endif /* def FEATURE_EXTERNAL_FILTERS */
1994 /*********************************************************************
1996 * Function : pcrs_filter_request_body
1998 * Description : Execute all text substitutions from all applying
1999 * +client_body_filter actions on the given text buffer.
2002 * 1 : csp = Current client state (buffers, headers, etc...)
2003 * 2 : data = Target data
2004 * 3 : data_len = Target data len
2006 * Returns : a pointer to the (newly allocated) modified buffer.
2007 * or NULL if there were no hits or something went wrong
2009 *********************************************************************/
2010 static char *pcrs_filter_request_body(const struct client_state *csp, const char *data, size_t *data_len)
2012 return pcrs_filter_impl(csp, FALSE, data, data_len);
2016 /*********************************************************************
2018 * Function : gif_deanimate_response
2020 * Description : Deanimate the GIF image that has been accumulated in
2021 * csp->iob->buf, set csp->content_length to the modified
2022 * size and raise the CSP_FLAG_MODIFIED flag.
2025 * 1 : csp = Current client state (buffers, headers, etc...)
2027 * Returns : a pointer to the (newly allocated) modified buffer.
2028 * or NULL in case something went wrong.
2030 *********************************************************************/
2032 char *gif_deanimate_response(struct client_state *csp)
2034 static char *gif_deanimate_response(struct client_state *csp)
2037 struct binbuffer *in, *out;
2041 size = (size_t)(csp->iob->eod - csp->iob->cur);
2043 in = zalloc_or_die(sizeof(*in));
2044 out = zalloc_or_die(sizeof(*out));
2046 in->buffer = csp->iob->cur;
2049 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2051 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2058 if ((int)size == out->offset)
2060 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2064 log_error(LOG_LEVEL_DEANIMATE,
2065 "Success! GIF shrunk from %lu bytes to %lu.", size, out->offset);
2067 csp->content_length = out->offset;
2068 csp->flags |= CSP_FLAG_MODIFIED;
2078 /*********************************************************************
2080 * Function : get_filter_function
2082 * Description : Decides which content filter function has
2083 * to be applied (if any). Only considers functions
2084 * for internal filters which are mutually-exclusive.
2087 * 1 : csp = Current client state (buffers, headers, etc...)
2089 * Returns : The content filter function to run, or
2090 * NULL if no content filter is active
2092 *********************************************************************/
2093 static filter_function_ptr get_filter_function(const struct client_state *csp)
2095 filter_function_ptr filter_function = NULL;
2098 * Choose the applying filter function based on
2099 * the content type and action settings.
2101 if ((csp->content_type & CT_TEXT) &&
2102 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2104 filter_function = pcrs_filter_response_body;
2106 else if ((csp->content_type & CT_GIF) &&
2107 (csp->action->flags & ACTION_DEANIMATE))
2109 filter_function = gif_deanimate_response;
2112 return filter_function;
2116 /*********************************************************************
2118 * Function : get_bytes_to_next_chunk_start
2120 * Description : Returns the number of bytes to the start of the
2121 * next chunk in the buffer.
2124 * 1 : buffer = Pointer to the text buffer
2125 * 2 : size = Number of bytes in the buffer.
2126 * 3 : offset = Where to expect the beginning of the next chunk.
2128 * Returns : -1 if the size can't be determined or data is missing,
2129 * otherwise the number of bytes to the start of the next chunk
2130 * or 0 if the last chunk has been fully buffered.
2132 *********************************************************************/
2133 static int get_bytes_to_next_chunk_start(char *buffer, size_t size, size_t offset)
2137 unsigned int chunk_size = 0;
2140 if (size <= offset || size < 5)
2143 * Not enough bytes bufferd to figure
2144 * out the size of the next chunk.
2149 chunk_start = buffer + offset;
2151 p = strstr(chunk_start, "\r\n");
2155 * The line with the chunk-size hasn't been completely received
2156 * yet (or is invalid).
2158 log_error(LOG_LEVEL_RE_FILTER,
2159 "Not enough or invalid data in buffer in chunk size line.");
2163 if (sscanf(chunk_start, "%x", &chunk_size) != 1)
2165 /* XXX: Write test case to trigger this. */
2166 log_error(LOG_LEVEL_ERROR, "Failed to parse chunk size. "
2167 "Size: %lu, offset: %lu. Chunk size start: %N", size, offset,
2168 (size - offset), chunk_start);
2173 * To get to the start of the next chunk size we have to skip
2174 * the line with the current chunk size followed by "\r\n" followd
2175 * by the actual data and another "\r\n" following the data.
2177 bytes_to_skip = (int)(p - chunk_start) + 2 + (int)chunk_size + 2;
2179 if (bytes_to_skip <= 0)
2181 log_error(LOG_LEVEL_ERROR,
2182 "Failed to figure out chunk offset. %u and %d seem dubious.",
2183 chunk_size, bytes_to_skip);
2186 if (chunk_size == 0)
2188 if (bytes_to_skip <= (size - offset))
2194 log_error(LOG_LEVEL_INFO,
2195 "Last chunk detected but we're still missing data.");
2200 return bytes_to_skip;
2204 /*********************************************************************
2206 * Function : get_bytes_missing_from_chunked_data
2208 * Description : Figures out how many bytes of data we need to get
2209 * to the start of the next chunk of data (XXX: terminology).
2210 * Due to the nature of chunk-encoded data we can only see
2211 * how many data is missing according to the last chunk size
2215 * 1 : buffer = Pointer to the text buffer
2216 * 2 : size = Number of bytes in the buffer.
2217 * 3 : offset = Where to expect the beginning of the next chunk.
2219 * Returns : -1 if the data can't be parsed (yet),
2220 * 0 if the buffer is complete or a
2221 * number of bytes that is missing.
2223 *********************************************************************/
2224 int get_bytes_missing_from_chunked_data(char *buffer, size_t size, size_t offset)
2227 int last_valid_offset = -1;
2229 if (size < offset || size < 5)
2231 /* Not enough data buffered yet */
2237 ret = get_bytes_to_next_chunk_start(buffer, size, offset);
2240 return last_valid_offset;
2248 last_valid_offset = (int)offset;
2250 offset += (size_t)ret;
2251 } while (offset < size);
2258 /*********************************************************************
2260 * Function : chunked_data_is_complete
2262 * Description : Detects if a buffer with chunk-encoded data looks
2266 * 1 : buffer = Pointer to the text buffer
2267 * 2 : size = Number of bytes in the buffer.
2268 * 3 : offset = Where to expect the beginning of the
2269 * first complete chunk.
2271 * Returns : TRUE if it looks like the data is complete,
2274 *********************************************************************/
2275 int chunked_data_is_complete(char *buffer, size_t size, size_t offset)
2277 return (0 == get_bytes_missing_from_chunked_data(buffer, size, offset));
2282 /*********************************************************************
2284 * Function : remove_chunked_transfer_coding
2286 * Description : In-situ remove the "chunked" transfer coding as defined
2287 * in RFC 7230 4.1 from a buffer. XXX: The implementation
2288 * is neither complete nor compliant (TODO #129).
2291 * 1 : buffer = Pointer to the text buffer
2292 * 2 : size = In: Number of bytes to be processed,
2293 * Out: Number of bytes after de-chunking.
2294 * (undefined in case of errors)
2296 * Returns : JB_ERR_OK for success,
2297 * JB_ERR_PARSE otherwise
2299 *********************************************************************/
2301 extern jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2303 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2307 unsigned int chunksize = 0;
2308 char *from_p, *to_p;
2309 const char *end_of_buffer = buffer + *size;
2313 log_error(LOG_LEVEL_FATAL, "Invalid chunked input. Buffer is empty.");
2314 return JB_ERR_PARSE;
2318 from_p = to_p = buffer;
2322 * Refuse to de-chunk invalid or incomplete data unless we're fuzzing.
2324 if (!chunked_data_is_complete(buffer, *size, 0))
2326 log_error(LOG_LEVEL_ERROR,
2327 "Chunk-encoding appears to be invalid. Content can't be filtered.");
2328 return JB_ERR_PARSE;
2332 if (sscanf(buffer, "%x", &chunksize) != 1)
2334 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2335 return JB_ERR_PARSE;
2338 while (chunksize > 0U)
2341 * If the chunk-size is valid, we should have at least
2342 * chunk-size bytes of chunk-data and five bytes of
2343 * meta data (chunk-size, CRLF, CRLF) left in the buffer.
2345 if (chunksize + 5 >= *size - newsize)
2347 log_error(LOG_LEVEL_ERROR,
2348 "Chunk size %u exceeds buffered data left. "
2349 "Already digested %lu of %lu buffered bytes.",
2350 chunksize, newsize, *size);
2351 return JB_ERR_PARSE;
2355 * Skip the chunk-size, the optional chunk-ext and the CRLF
2356 * that is supposed to be located directly before the start
2359 if (NULL == (from_p = strstr(from_p, "\r\n")))
2361 log_error(LOG_LEVEL_ERROR,
2362 "Failed to strip \"chunked\" transfer coding. "
2363 "Line with chunk size doesn't seem to end properly.");
2364 return JB_ERR_PARSE;
2369 * The previous strstr() does not enforce chunk-validity
2370 * and is sattisfied as long a CRLF is left in the buffer.
2372 * Make sure the bytes we consider chunk-data are within
2375 if (from_p + chunksize >= end_of_buffer)
2377 log_error(LOG_LEVEL_ERROR,
2378 "Failed to decode content for filtering. "
2379 "One chunk end is beyond the end of the buffer.");
2380 return JB_ERR_PARSE;
2383 memmove(to_p, from_p, (size_t) chunksize);
2384 newsize += chunksize;
2385 to_p = buffer + newsize;
2386 from_p += chunksize;
2389 * Not merging this check with the previous one allows us
2390 * to keep chunks without trailing CRLF. It's not clear
2391 * if we actually have to care about those, though.
2393 if (from_p + 2 >= end_of_buffer)
2395 log_error(LOG_LEVEL_ERROR, "Not enough room for trailing CRLF.");
2396 return JB_ERR_PARSE;
2399 if (sscanf(from_p, "%x", &chunksize) != 1)
2401 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
2406 /* XXX: Should get its own loglevel. */
2407 log_error(LOG_LEVEL_RE_FILTER,
2408 "De-chunking successful. Shrunk from %lu to %lu", *size, newsize);
2417 /*********************************************************************
2419 * Function : prepare_for_filtering
2421 * Description : If necessary, de-chunks and decompresses
2422 * the content so it can get filterd.
2425 * 1 : csp = Current client state (buffers, headers, etc...)
2427 * Returns : JB_ERR_OK for success,
2428 * JB_ERR_PARSE otherwise
2430 *********************************************************************/
2431 static jb_err prepare_for_filtering(struct client_state *csp)
2433 jb_err err = JB_ERR_OK;
2436 * If the body has a "chunked" transfer-encoding,
2437 * get rid of it, adjusting size and iob->eod
2439 if (csp->flags & CSP_FLAG_CHUNKED)
2441 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2443 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2444 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2445 if (JB_ERR_OK == err)
2447 csp->iob->eod = csp->iob->cur + size;
2448 csp->flags |= CSP_FLAG_MODIFIED;
2452 return JB_ERR_PARSE;
2458 * If the body has a supported transfer-encoding,
2459 * decompress it, adjusting size and iob->eod.
2461 if ((csp->content_type & (CT_GZIP|CT_DEFLATE))
2462 #ifdef FEATURE_BROTLI
2463 || (csp->content_type & CT_BROTLI)
2467 if (0 == csp->iob->eod - csp->iob->cur)
2469 /* Nothing left after de-chunking. */
2473 err = decompress_iob(csp);
2475 if (JB_ERR_OK == err)
2477 csp->flags |= CSP_FLAG_MODIFIED;
2478 csp->content_type &= ~CT_TABOO;
2483 * Unset content types to remember not to
2484 * modify the Content-Encoding header later.
2486 csp->content_type &= ~CT_GZIP;
2487 csp->content_type &= ~CT_DEFLATE;
2488 #ifdef FEATURE_BROTLI
2489 csp->content_type &= ~CT_BROTLI;
2499 /*********************************************************************
2501 * Function : execute_content_filters
2503 * Description : Executes a given content filter.
2506 * 1 : csp = Current client state (buffers, headers, etc...)
2508 * Returns : Pointer to the modified buffer, or
2509 * NULL if filtering failed or wasn't necessary.
2511 *********************************************************************/
2512 char *execute_content_filters(struct client_state *csp)
2515 filter_function_ptr content_filter;
2517 assert(content_filters_enabled(csp->action));
2519 if (0 == csp->iob->eod - csp->iob->cur)
2522 * No content (probably status code 301, 302 ...),
2523 * no filtering necessary.
2528 if (JB_ERR_OK != prepare_for_filtering(csp))
2531 * We failed to de-chunk or decompress, don't accept
2532 * another request on the client connection.
2534 csp->flags &= ~CSP_FLAG_CLIENT_CONNECTION_KEEP_ALIVE;
2538 if (0 == csp->iob->eod - csp->iob->cur)
2541 * Clown alarm: chunked and/or compressed nothing delivered.
2546 content_filter = get_filter_function(csp);
2547 content = (content_filter != NULL) ? (*content_filter)(csp) : NULL;
2549 #ifdef FEATURE_EXTERNAL_FILTERS
2550 if ((csp->content_type & CT_TEXT) &&
2551 !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]))
2553 struct list_entry *filtername;
2554 size_t size = (size_t)csp->content_length;
2556 if (content == NULL)
2558 content = csp->iob->cur;
2559 size = (size_t)(csp->iob->eod - csp->iob->cur);
2562 for (filtername = csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]->first;
2563 filtername ; filtername = filtername->next)
2565 char *result = execute_external_filter(csp, filtername->str, content, &size);
2568 if (content != csp->iob->cur)
2575 csp->flags |= CSP_FLAG_MODIFIED;
2576 csp->content_length = size;
2578 #endif /* def FEATURE_EXTERNAL_FILTERS */
2585 /*********************************************************************
2587 * Function : execute_client_body_filters
2589 * Description : Executes client body filters for the request that is buffered
2590 * in the client_iob. The client_iob is updated with the filtered
2594 * 1 : csp = Current client state (buffers, headers, etc...)
2595 * 2 : content_length = content length. Upon successful filtering
2596 * the passed value is updated with the new content length.
2598 * Returns : 1 if the content has been filterd. 0 if it hasn't.
2600 *********************************************************************/
2601 int execute_client_body_filters(struct client_state *csp, size_t *content_length)
2603 char *filtered_content;
2605 assert(client_body_filters_enabled(csp->action));
2607 if (content_length == 0)
2610 * No content, no filtering necessary.
2615 filtered_content = pcrs_filter_request_body(csp, csp->client_iob->cur, content_length);
2616 if (filtered_content != NULL)
2618 freez(csp->client_iob->buf);
2619 csp->client_iob->buf = filtered_content;
2620 csp->client_iob->cur = csp->client_iob->buf;
2621 csp->client_iob->eod = csp->client_iob->cur + *content_length;
2622 csp->client_iob->size = *content_length;
2631 /*********************************************************************
2633 * Function : execute_client_body_taggers
2635 * Description : Executes client body taggers for the request that is
2636 * buffered in the client_iob.
2637 * XXX: Lots of code shared with header_tagger
2640 * 1 : csp = Current client state (buffers, headers, etc...)
2641 * 2 : content_length = content length.
2645 *********************************************************************/
2646 jb_err execute_client_body_taggers(struct client_state *csp, size_t content_length)
2648 enum filter_type wanted_filter_type = FT_CLIENT_BODY_TAGGER;
2649 int multi_action_index = ACTION_MULTI_CLIENT_BODY_TAGGER;
2652 struct re_filterfile_spec *b;
2653 struct list_entry *tag_name;
2655 assert(client_body_taggers_enabled(csp->action));
2657 if (content_length == 0)
2660 * No content, no tagging necessary.
2665 log_error(LOG_LEVEL_INFO, "Got to execute tagger on %N",
2666 content_length, csp->client_iob->cur);
2668 if (list_is_empty(csp->action->multi[multi_action_index])
2669 || filters_available(csp) == FALSE)
2671 /* Return early if no taggers apply or if none are available. */
2675 /* Execute all applying taggers */
2676 for (tag_name = csp->action->multi[multi_action_index]->first;
2677 NULL != tag_name; tag_name = tag_name->next)
2679 char *modified_tag = NULL;
2680 char *tag = csp->client_iob->cur;
2681 size_t size = content_length;
2684 b = get_filter(csp, tag_name->str, wanted_filter_type);
2690 joblist = b->joblist;
2692 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
2694 if (NULL == joblist)
2696 log_error(LOG_LEVEL_TAGGING,
2697 "Tagger %s has empty joblist. Nothing to do.", b->name);
2701 /* execute their pcrs_joblist on the body. */
2702 for (job = joblist; NULL != job; job = job->next)
2704 const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
2708 /* Success, continue with the modified version. */
2709 if (tag != csp->client_iob->cur)
2717 /* Tagger doesn't match */
2720 /* Regex failure, log it but continue anyway. */
2721 log_error(LOG_LEVEL_ERROR,
2722 "Problems with tagger \'%s\': %s",
2723 b->name, pcrs_strerror(hits));
2725 freez(modified_tag);
2729 if (b->dynamic) pcrs_free_joblist(joblist);
2731 /* If this tagger matched */
2732 if (tag != csp->client_iob->cur)
2737 * There is no technical limitation which makes
2738 * it impossible to use empty tags, but I assume
2739 * no one would do it intentionally.
2742 log_error(LOG_LEVEL_TAGGING,
2743 "Tagger \'%s\' created an empty tag. Ignored.", b->name);
2747 if (list_contains_item(csp->action->multi[ACTION_MULTI_SUPPRESS_TAG], tag))
2749 log_error(LOG_LEVEL_TAGGING,
2750 "Tagger \'%s\' didn't add tag \'%s\': suppressed",
2756 if (!list_contains_item(csp->tags, tag))
2758 if (JB_ERR_OK != enlist(csp->tags, tag))
2760 log_error(LOG_LEVEL_ERROR,
2761 "Insufficient memory to add tag \'%s\', "
2762 "based on tagger \'%s\'",
2767 char *action_message;
2769 * update the action bits right away, to make
2770 * tagging based on tags set by earlier taggers
2771 * of the same kind possible.
2773 if (update_action_bits_for_tag(csp, tag))
2775 action_message = "Action bits updated accordingly.";
2779 action_message = "No action bits update necessary.";
2782 log_error(LOG_LEVEL_TAGGING,
2783 "Tagger \'%s\' added tag \'%s\'. %s",
2784 b->name, tag, action_message);
2789 /* XXX: Is this log-worthy? */
2790 log_error(LOG_LEVEL_TAGGING,
2791 "Tagger \'%s\' didn't add tag \'%s\'. Tag already present",
2802 /*********************************************************************
2804 * Function : get_url_actions
2806 * Description : Gets the actions for this URL.
2809 * 1 : csp = Current client state (buffers, headers, etc...)
2810 * 2 : http = http_request request for blocked URLs
2814 *********************************************************************/
2815 void get_url_actions(struct client_state *csp, struct http_request *http)
2817 struct file_list *fl;
2818 struct url_actions *b;
2821 init_current_action(csp->action);
2823 for (i = 0; i < MAX_AF_FILES; i++)
2825 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2830 #ifdef FEATURE_CLIENT_TAGS
2831 apply_url_actions(csp->action, http, csp->client_tags, b);
2833 apply_url_actions(csp->action, http, b);
2840 /*********************************************************************
2842 * Function : apply_url_actions
2844 * Description : Applies a list of URL actions.
2847 * 1 : action = Destination.
2848 * 2 : http = Current URL
2849 * 3 : client_tags = list of client tags
2850 * 4 : b = list of URL actions to apply
2854 *********************************************************************/
2855 static void apply_url_actions(struct current_action_spec *action,
2856 struct http_request *http,
2857 #ifdef FEATURE_CLIENT_TAGS
2858 const struct list *client_tags,
2860 struct url_actions *b)
2864 /* Should never happen */
2868 for (b = b->next; NULL != b; b = b->next)
2870 if (url_match(b->url, http))
2872 merge_current_action(action, b->action);
2874 #ifdef FEATURE_CLIENT_TAGS
2875 if (client_tag_match(b->url, client_tags))
2877 merge_current_action(action, b->action);
2884 /*********************************************************************
2886 * Function : get_forward_override_settings
2888 * Description : Returns forward settings as specified with the
2889 * forward-override{} action. forward-override accepts
2890 * forward lines similar to the one used in the
2891 * configuration file, but without the URL pattern.
2897 * in the configuration file can be replaced with
2898 * the action section:
2900 * {+forward-override{forward .}}
2904 * 1 : csp = Current client state (buffers, headers, etc...)
2906 * Returns : Pointer to forwarding structure in case of success.
2907 * Invalid syntax is fatal.
2909 *********************************************************************/
2910 static const struct forward_spec *get_forward_override_settings(struct client_state *csp)
2912 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2913 char forward_settings[BUFFER_SIZE];
2914 char *http_parent = NULL;
2915 /* variable names were chosen for consistency reasons. */
2916 struct forward_spec *fwd = NULL;
2920 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2921 /* Should be enforced by load_one_actions_file() */
2922 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2924 /* Create a copy ssplit can modify */
2925 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2927 if (NULL != csp->fwd)
2930 * XXX: Currently necessary to prevent memory
2931 * leaks when the show-url-info cgi page is visited.
2933 unload_forward_spec(csp->fwd);
2937 * allocate a new forward node, valid only for
2938 * the lifetime of this request. Save its location
2939 * in csp as well, so sweep() can free it later on.
2941 fwd = csp->fwd = zalloc_or_die(sizeof(*fwd));
2943 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec));
2944 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2946 fwd->type = SOCKS_NONE;
2948 /* Parse the parent HTTP proxy host:port */
2949 http_parent = vec[1];
2952 else if ((vec_count == 2) && !strcasecmp(vec[0], "forward-webserver"))
2954 fwd->type = FORWARD_WEBSERVER;
2956 /* Parse the parent HTTP server host:port */
2957 http_parent = vec[1];
2960 else if (vec_count == 3)
2962 char *socks_proxy = NULL;
2964 if (!strcasecmp(vec[0], "forward-socks4"))
2966 fwd->type = SOCKS_4;
2967 socks_proxy = vec[1];
2969 else if (!strcasecmp(vec[0], "forward-socks4a"))
2971 fwd->type = SOCKS_4A;
2972 socks_proxy = vec[1];
2974 else if (!strcasecmp(vec[0], "forward-socks5"))
2976 fwd->type = SOCKS_5;
2977 socks_proxy = vec[1];
2979 else if (!strcasecmp(vec[0], "forward-socks5t"))
2981 fwd->type = SOCKS_5T;
2982 socks_proxy = vec[1];
2985 if (NULL != socks_proxy)
2987 /* Parse the SOCKS proxy [user:pass@]host[:port] */
2988 fwd->gateway_port = 1080;
2989 parse_forwarder_address(socks_proxy,
2990 &fwd->gateway_host, &fwd->gateway_port,
2991 &fwd->auth_username, &fwd->auth_password);
2993 http_parent = vec[2];
2997 if (NULL == http_parent)
2999 log_error(LOG_LEVEL_FATAL,
3000 "Invalid forward-override syntax in: %s", forward_override_line);
3001 /* Never get here - LOG_LEVEL_FATAL causes program exit */
3004 /* Parse http forwarding settings */
3005 if (strcmp(http_parent, ".") != 0)
3007 fwd->forward_port = 8000;
3008 parse_forwarder_address(http_parent,
3009 &fwd->forward_host, &fwd->forward_port,
3013 assert (NULL != fwd);
3015 log_error(LOG_LEVEL_CONNECT,
3016 "Overriding forwarding settings based on \'%s\'", forward_override_line);
3022 /*********************************************************************
3024 * Function : forward_url
3026 * Description : Should we forward this to another proxy?
3029 * 1 : csp = Current client state (buffers, headers, etc...)
3030 * 2 : http = http_request request for current URL
3032 * Returns : Pointer to forwarding information.
3034 *********************************************************************/
3035 const struct forward_spec *forward_url(struct client_state *csp,
3036 const struct http_request *http)
3038 static const struct forward_spec fwd_default[1]; /* Zero'ed due to being static. */
3039 struct forward_spec *fwd = csp->config->forward;
3041 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
3043 return get_forward_override_settings(csp);
3053 if (url_match(fwd->url, http))
3064 /*********************************************************************
3066 * Function : direct_response
3068 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
3069 * request and if so, return a HTTP 501 to the client.
3071 * FIXME: I have a stupid name and I should handle the
3072 * requests properly. Still, what we do here is rfc-
3073 * compliant, whereas ignoring or forwarding are not.
3076 * 1 : csp = Current client state (buffers, headers, etc...)
3078 * Returns : http_response if , NULL if nonmatch or handler fail
3080 *********************************************************************/
3081 struct http_response *direct_response(struct client_state *csp)
3083 struct http_response *rsp;
3084 struct list_entry *p;
3086 if ((0 == strcmpic(csp->http->gpc, "trace"))
3087 || (0 == strcmpic(csp->http->gpc, "options")))
3089 for (p = csp->headers->first; (p != NULL) ; p = p->next)
3091 if (!strncmpic(p->str, "Max-Forwards:", 13))
3093 unsigned int max_forwards;
3096 * If it's a Max-Forwards value of zero,
3097 * we have to intercept the request.
3099 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
3102 * FIXME: We could handle at least TRACE here,
3103 * but that would require a verbatim copy of
3104 * the request which we don't have anymore
3106 log_error(LOG_LEVEL_HEADER,
3107 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
3110 /* Get mem for response or fail*/
3111 if (NULL == (rsp = alloc_http_response()))
3113 return cgi_error_memory();
3116 rsp->status = strdup_or_die("501 Not Implemented");
3118 rsp->crunch_reason = UNSUPPORTED;
3120 return(finish_http_response(csp, rsp));
3129 /*********************************************************************
3131 * Function : content_requires_filtering
3133 * Description : Checks whether there are any content filters
3134 * enabled for the current request and if they
3135 * can actually be applied..
3138 * 1 : csp = Current client state (buffers, headers, etc...)
3140 * Returns : TRUE for yes, FALSE otherwise
3142 *********************************************************************/
3143 int content_requires_filtering(struct client_state *csp)
3145 if ((csp->content_type & CT_TABOO)
3146 && !(csp->action->flags & ACTION_FORCE_TEXT_MODE))
3152 * Are we enabling text mode by force?
3154 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
3157 * Do we really have to?
3159 if (csp->content_type & CT_TEXT)
3161 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
3165 csp->content_type |= CT_TEXT;
3166 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
3170 if (!(csp->content_type & CT_DECLARED))
3173 * The server didn't bother to declare a MIME-Type.
3174 * Assume it's text that can be filtered.
3176 * This also regularly happens with 304 responses,
3177 * therefore logging anything here would cause
3180 csp->content_type |= CT_TEXT;
3184 * Choose the applying filter function based on
3185 * the content type and action settings.
3187 if ((csp->content_type & CT_TEXT) &&
3188 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER]) ||
3189 !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER])))
3193 else if ((csp->content_type & CT_GIF) &&
3194 (csp->action->flags & ACTION_DEANIMATE))
3204 /*********************************************************************
3206 * Function : content_filters_enabled
3208 * Description : Checks whether there are any content filters
3209 * enabled for the current request.
3212 * 1 : action = Action spec to check.
3214 * Returns : TRUE for yes, FALSE otherwise
3216 *********************************************************************/
3217 int content_filters_enabled(const struct current_action_spec *action)
3219 return ((action->flags & ACTION_DEANIMATE) ||
3220 !list_is_empty(action->multi[ACTION_MULTI_FILTER]) ||
3221 !list_is_empty(action->multi[ACTION_MULTI_EXTERNAL_FILTER]));
3225 /*********************************************************************
3227 * Function : client_body_filters_enabled
3229 * Description : Checks whether there are any client body filters
3230 * enabled for the current request.
3233 * 1 : action = Action spec to check.
3235 * Returns : TRUE for yes, FALSE otherwise
3237 *********************************************************************/
3238 int client_body_filters_enabled(const struct current_action_spec *action)
3240 return !list_is_empty(action->multi[ACTION_MULTI_CLIENT_BODY_FILTER]);
3244 /*********************************************************************
3246 * Function : client_body_taggers_enabled
3248 * Description : Checks whether there are any client body taggers
3249 * enabled for the current request.
3252 * 1 : action = Action spec to check.
3254 * Returns : TRUE for yes, FALSE otherwise
3256 *********************************************************************/
3257 int client_body_taggers_enabled(const struct current_action_spec *action)
3259 return !list_is_empty(action->multi[ACTION_MULTI_CLIENT_BODY_TAGGER]);
3262 /*********************************************************************
3264 * Function : filters_available
3266 * Description : Checks whether there are any filters available.
3269 * 1 : csp = Current client state (buffers, headers, etc...)
3271 * Returns : TRUE for yes, FALSE otherwise.
3273 *********************************************************************/
3274 int filters_available(const struct client_state *csp)
3277 for (i = 0; i < MAX_AF_FILES; i++)
3279 const struct file_list *fl = csp->rlist[i];
3280 if ((NULL != fl) && (NULL != fl->f))
3288 #ifdef FEATURE_EXTENDED_STATISTICS
3290 struct filter_statistics_entry
3293 unsigned long long executions;
3294 unsigned long long response_bodies_modified;
3295 unsigned long long hits;
3297 struct filter_statistics_entry *next;
3300 static struct filter_statistics_entry *filter_statistics = NULL;
3303 /*********************************************************************
3305 * Function : register_filter_for_statistics
3307 * Description : Registers a filter so we can gather statistics for
3308 * it unless the filter has already been registered
3312 * 1 : filter = Name of the filter to register
3316 *********************************************************************/
3317 void register_filter_for_statistics(const char *filter)
3319 struct filter_statistics_entry *entry;
3321 privoxy_mutex_lock(&filter_statistics_mutex);
3323 if (filter_statistics == NULL)
3325 filter_statistics = zalloc_or_die(sizeof(struct filter_statistics_entry));
3326 entry = filter_statistics;
3327 entry->filter = strdup_or_die(filter);
3328 privoxy_mutex_unlock(&filter_statistics_mutex);
3331 entry = filter_statistics;
3332 while (entry != NULL)
3334 if (!strcmp(entry->filter, filter))
3336 /* Already registered, nothing to do. */
3339 if (entry->next == NULL)
3341 entry->next = zalloc_or_die(sizeof(struct filter_statistics_entry));
3342 entry->next->filter = strdup_or_die(filter);
3345 entry = entry->next;
3348 privoxy_mutex_unlock(&filter_statistics_mutex);
3353 /*********************************************************************
3355 * Function : update_filter_statistics
3357 * Description : Updates the statistics for a filter.
3360 * 1 : filter = Name of the filter to update
3361 * 2 : hits = Hit count.
3365 *********************************************************************/
3366 void update_filter_statistics(const char *filter, int hits)
3368 struct filter_statistics_entry *entry;
3370 privoxy_mutex_lock(&filter_statistics_mutex);
3372 entry = filter_statistics;
3373 while (entry != NULL)
3375 if (!strcmp(entry->filter, filter))
3377 entry->executions++;
3380 entry->response_bodies_modified++;
3381 entry->hits += (unsigned)hits;
3385 entry = entry->next;
3388 privoxy_mutex_unlock(&filter_statistics_mutex);
3393 /*********************************************************************
3395 * Function : get_filter_statistics
3397 * Description : Gets the statistics for a filter.
3400 * 1 : filter = Name of the filter to get statistics for.
3401 * 2 : executions = Storage for the execution count.
3402 * 3 : response_bodies_modified = Storage for the number
3403 * of modified response bodies.
3404 * 4 : hits = Storage for the number of hits.
3408 *********************************************************************/
3409 void get_filter_statistics(const char *filter, unsigned long long *executions,
3410 unsigned long long *response_bodies_modified,
3411 unsigned long long *hits)
3413 struct filter_statistics_entry *entry;
3415 privoxy_mutex_lock(&filter_statistics_mutex);
3417 entry = filter_statistics;
3418 while (entry != NULL)
3420 if (!strcmp(entry->filter, filter))
3422 *executions = entry->executions;
3423 *response_bodies_modified = entry->response_bodies_modified;
3424 *hits = entry->hits;
3427 entry = entry->next;
3430 privoxy_mutex_unlock(&filter_statistics_mutex);
3435 struct block_statistics_entry
3438 unsigned long long count;
3440 struct block_statistics_entry *next;
3443 static struct block_statistics_entry *block_statistics = NULL;
3445 /*********************************************************************
3447 * Function : register_block_reason_for_statistics
3449 * Description : Registers a block reason so we can gather statistics
3450 * for it unless the block reason has already been
3451 * registered before.
3454 * 1 : block_reason = Block reason to register
3458 *********************************************************************/
3459 void register_block_reason_for_statistics(const char *block_reason)
3461 struct block_statistics_entry *entry;
3463 privoxy_mutex_lock(&block_reason_statistics_mutex);
3465 if (block_statistics == NULL)
3467 block_statistics = zalloc_or_die(sizeof(struct block_statistics_entry));
3468 entry = block_statistics;
3469 entry->block_reason = strdup_or_die(block_reason);
3470 privoxy_mutex_unlock(&block_reason_statistics_mutex);
3473 entry = block_statistics;
3474 while (entry != NULL)
3476 if (!strcmp(entry->block_reason, block_reason))
3478 /* Already registered, nothing to do. */
3481 if (entry->next == NULL)
3483 entry->next = zalloc_or_die(sizeof(struct block_statistics_entry));
3484 entry->next->block_reason = strdup_or_die(block_reason);
3487 entry = entry->next;
3490 privoxy_mutex_unlock(&block_reason_statistics_mutex);
3495 /*********************************************************************
3497 * Function : increment_block_reason_counter
3499 * Description : Updates the counter for a block reason.
3502 * 1 : block_reason = Block reason to count
3506 *********************************************************************/
3507 static void increment_block_reason_counter(const char *block_reason)
3509 struct block_statistics_entry *entry;
3511 privoxy_mutex_lock(&block_reason_statistics_mutex);
3513 entry = block_statistics;
3514 while (entry != NULL)
3516 if (!strcmp(entry->block_reason, block_reason))
3521 entry = entry->next;
3524 privoxy_mutex_unlock(&block_reason_statistics_mutex);
3529 /*********************************************************************
3531 * Function : get_block_reason_count
3533 * Description : Gets number of times a block reason was used.
3536 * 1 : block_reason = Block reason to get statistics for.
3537 * 2 : count = Storage for the number of times the block
3542 *********************************************************************/
3543 void get_block_reason_count(const char *block_reason, unsigned long long *count)
3545 struct block_statistics_entry *entry;
3547 privoxy_mutex_lock(&block_reason_statistics_mutex);
3549 entry = block_statistics;
3550 while (entry != NULL)
3552 if (!strcmp(entry->block_reason, block_reason))
3554 *count = entry->count;
3557 entry = entry->next;
3560 privoxy_mutex_unlock(&block_reason_statistics_mutex);
3564 #endif /* def FEATURE_EXTENDED_STATISTICS */