1 const char filters_rcs[] = "$Id: filters.c,v 1.204 2017/03/08 13:13:18 fabiankeil Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/filters.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
8 * Copyright : Written by and Copyright (C) 2001-2016 the
9 * Privoxy team. http://www.privoxy.org/
11 * Based on the Internet Junkbuster originally written
12 * by and Copyright (C) 1997 Anonymous Coders and
13 * Junkbusters Corporation. http://www.junkbusters.com
15 * This program is free software; you can redistribute it
16 * and/or modify it under the terms of the GNU General
17 * Public License as published by the Free Software
18 * Foundation; either version 2 of the License, or (at
19 * your option) any later version.
21 * This program is distributed in the hope that it will
22 * be useful, but WITHOUT ANY WARRANTY; without even the
23 * implied warranty of MERCHANTABILITY or FITNESS FOR A
24 * PARTICULAR PURPOSE. See the GNU General Public
25 * License for more details.
27 * The GNU General Public License should be included with
28 * this file. If not, you can view it at
29 * http://www.gnu.org/copyleft/gpl.html
30 * or write to the Free Software Foundation, Inc., 59
31 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
33 *********************************************************************/
39 #include <sys/types.h>
48 #endif /* ndef __OS2__ */
49 #include <netinet/in.h>
52 #endif /* ndef _WIN32 */
56 #endif /* def __OS2__ */
64 #include "jbsockets.h"
70 #include "deanimate.h"
73 #ifdef FEATURE_CLIENT_TAGS
74 #include "client-tags.h"
81 const char filters_h_rcs[] = FILTERS_H_VERSION;
83 typedef char *(*filter_function_ptr)();
84 static filter_function_ptr get_filter_function(const struct client_state *csp);
85 static jb_err prepare_for_filtering(struct client_state *csp);
86 static void apply_url_actions(struct current_action_spec *action,
87 struct http_request *http,
88 #ifdef FEATURE_CLIENT_TAGS
89 const struct list *client_tags,
91 struct url_actions *b);
95 /*********************************************************************
97 * Function : sockaddr_storage_to_ip
99 * Description : Access internal structure of sockaddr_storage
102 * 1 : addr = socket address
103 * 2 : ip = IP address as array of octets in network order
104 * (it points into addr)
105 * 3 : len = length of IP address in octets
106 * 4 : port = port number in network order;
108 * Returns : 0 = no errror; -1 otherwise.
110 *********************************************************************/
111 static int sockaddr_storage_to_ip(const struct sockaddr_storage *addr,
112 uint8_t **ip, unsigned int *len,
115 assert(NULL != addr);
117 switch (addr->ss_family)
127 &(((struct sockaddr_in *)addr)->sin_addr.s_addr);
131 *port = &((struct sockaddr_in *)addr)->sin_port;
142 *ip = ((struct sockaddr_in6 *)addr)->sin6_addr.s6_addr;
146 *port = &((struct sockaddr_in6 *)addr)->sin6_port;
151 /* Unsupported address family */
159 /*********************************************************************
161 * Function : match_sockaddr
163 * Description : Check whether address matches network (IP address and port)
166 * 1 : network = socket address of subnework
167 * 2 : netmask = network mask as socket address
168 * 3 : address = checked socket address against given network
170 * Returns : 0 = doesn't match; 1 = does match
172 *********************************************************************/
173 static int match_sockaddr(const struct sockaddr_storage *network,
174 const struct sockaddr_storage *netmask,
175 const struct sockaddr_storage *address)
177 uint8_t *network_addr, *netmask_addr, *address_addr;
178 unsigned int addr_len;
179 in_port_t *network_port, *netmask_port, *address_port;
182 if (network->ss_family != netmask->ss_family)
184 /* This should never happen */
185 assert(network->ss_family == netmask->ss_family);
186 log_error(LOG_LEVEL_FATAL, "Network and netmask differ in family.");
189 sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
190 sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
191 sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
193 /* Check for family */
194 if ((network->ss_family == AF_INET) && (address->ss_family == AF_INET6)
195 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr))
197 /* Map AF_INET6 V4MAPPED address into AF_INET */
201 else if ((network->ss_family == AF_INET6) && (address->ss_family == AF_INET)
202 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr))
204 /* Map AF_INET6 V4MAPPED network into AF_INET */
210 /* XXX: Port check is signaled in netmask */
211 if (*netmask_port && *network_port != *address_port)
216 /* TODO: Optimize by checking by words insted of octets */
217 for (i = 0; (i < addr_len) && netmask_addr[i]; i++)
219 if ((network_addr[i] & netmask_addr[i]) !=
220 (address_addr[i] & netmask_addr[i]))
228 #endif /* def HAVE_RFC2553 */
231 /*********************************************************************
233 * Function : block_acl
235 * Description : Block this request?
236 * Decide yes or no based on ACL file.
239 * 1 : dst = The proxy or gateway address this is going to.
240 * Or NULL to check all possible targets.
241 * 2 : csp = Current client state (buffers, headers, etc...)
242 * Also includes the client IP address.
244 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
246 *********************************************************************/
247 int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
249 struct access_control_list *acl = csp->config->acl;
251 /* if not using an access control list, then permit the connection */
257 /* search the list */
262 match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
264 (csp->ip_addr_long & acl->src->mask) == acl->src->addr
270 /* Just want to check if they have any access */
271 if (acl->action == ACL_PERMIT)
283 * XXX: An undefined acl->dst is full of zeros and should be
284 * considered a wildcard address. sockaddr_storage_to_ip()
285 * fails on such destinations because of unknown sa_familly
286 * (glibc only?). However this test is not portable.
288 * So, we signal the acl->dst is wildcard in wildcard_dst.
291 match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
293 ((dst->addr & acl->dst->mask) == acl->dst->addr)
294 && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
298 if (acl->action == ACL_PERMIT)
316 /*********************************************************************
318 * Function : acl_addr
320 * Description : Called from `load_config' to parse an ACL address.
323 * 1 : aspec = String specifying ACL address.
324 * 2 : aca = struct access_control_addr to fill in.
326 * Returns : 0 => Ok, everything else is an error.
328 *********************************************************************/
329 int acl_addr(const char *aspec, struct access_control_addr *aca)
333 struct addrinfo hints, *result;
335 in_port_t *mask_port;
336 unsigned int addr_len;
339 #endif /* def HAVE_RFC2553 */
341 char *acl_spec = NULL;
344 /* XXX: Depend on ai_family */
352 * Use a temporary acl spec copy so we can log
353 * the unmodified original in case of parse errors.
355 acl_spec = strdup_or_die(aspec);
357 if ((p = strchr(acl_spec, '/')) != NULL)
360 if (privoxy_isdigit(*p) == 0)
365 masklength = atoi(p);
368 if ((masklength < 0) ||
380 if ((*acl_spec == '[') && (NULL != (p = strchr(acl_spec, ']'))))
383 memmove(acl_spec, acl_spec + 1, (size_t)(p - acl_spec));
392 p = strchr(acl_spec, ':');
402 memset(&hints, 0, sizeof(struct addrinfo));
403 hints.ai_family = AF_UNSPEC;
404 hints.ai_socktype = SOCK_STREAM;
406 i = getaddrinfo(acl_spec, p, &hints, &result);
410 log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s",
411 acl_spec, p, gai_strerror(i));
417 /* TODO: Allow multihomed hostnames */
418 memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
419 freeaddrinfo(result);
425 port = strtol(p, &endptr, 10);
427 if (port <= 0 || port > 65535 || *endptr != '\0')
434 aca->port = (unsigned long)port;
436 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
439 if (aca->addr == INADDR_NONE)
441 /* XXX: This will be logged as parse error. */
444 #endif /* def HAVE_RFC2553 */
446 /* build the netmask */
448 /* Clip masklength according to current family. */
449 if ((aca->addr.ss_family == AF_INET) && (masklength > 32))
454 aca->mask.ss_family = aca->addr.ss_family;
455 if (sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port))
462 /* ACL contains a port number, check ports in the future. */
467 * XXX: This could be optimized to operate on whole words instead
468 * of octets (128-bit CPU could do it in one iteration).
471 * Octets after prefix can be omitted because of
472 * previous initialization to zeros.
474 for (i = 0; (i < addr_len) && masklength; i++)
484 * XXX: This assumes MSB of octet is on the left side.
485 * This should be true for all architectures or solved
488 mask_data[i] = (uint8_t)~((1 << (8 - masklength)) - 1);
495 for (i=1; i <= masklength ; i++)
497 aca->mask |= (1U << (32 - i));
500 /* now mask off the host portion of the ip address
501 * (i.e. save on the network portion of the address).
503 aca->addr = aca->addr & aca->mask;
504 #endif /* def HAVE_RFC2553 */
509 #endif /* def FEATURE_ACL */
512 /*********************************************************************
514 * Function : connect_port_is_forbidden
516 * Description : Check to see if CONNECT requests to the destination
517 * port of this request are forbidden. The check is
518 * independend of the actual request method.
521 * 1 : csp = Current client state (buffers, headers, etc...)
523 * Returns : True if yes, false otherwise.
525 *********************************************************************/
526 int connect_port_is_forbidden(const struct client_state *csp)
528 return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
529 !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
534 /*********************************************************************
536 * Function : block_url
538 * Description : Called from `chat'. Check to see if we need to block this.
541 * 1 : csp = Current client state (buffers, headers, etc...)
543 * Returns : NULL => unblocked, else HTTP block response
545 *********************************************************************/
546 struct http_response *block_url(struct client_state *csp)
548 struct http_response *rsp;
549 const char *new_content_type = NULL;
552 * If it's not blocked, don't block it ;-)
554 if ((csp->action->flags & ACTION_BLOCK) == 0)
558 if (csp->action->flags & ACTION_REDIRECT)
560 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
563 * Else, prepare a response
565 if (NULL == (rsp = alloc_http_response()))
567 return cgi_error_memory();
571 * If it's an image-url, send back an image or redirect
572 * as specified by the relevant +image action
574 #ifdef FEATURE_IMAGE_BLOCKING
575 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
579 /* determine HOW images should be blocked */
580 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
582 if (csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
584 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
587 /* and handle accordingly: */
588 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
590 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
591 rsp->body = bindup(image_pattern_data, image_pattern_length);
592 if (rsp->body == NULL)
594 free_http_response(rsp);
595 return cgi_error_memory();
597 rsp->content_length = image_pattern_length;
599 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
601 free_http_response(rsp);
602 return cgi_error_memory();
605 else if (0 == strcmpic(p, "blank"))
607 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
608 rsp->body = bindup(image_blank_data, image_blank_length);
609 if (rsp->body == NULL)
611 free_http_response(rsp);
612 return cgi_error_memory();
614 rsp->content_length = image_blank_length;
616 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
618 free_http_response(rsp);
619 return cgi_error_memory();
624 rsp->status = strdup_or_die("302 Local Redirect from Privoxy");
626 if (enlist_unique_header(rsp->headers, "Location", p))
628 free_http_response(rsp);
629 return cgi_error_memory();
635 #endif /* def FEATURE_IMAGE_BLOCKING */
636 if (csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
639 * Send empty document.
641 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
644 rsp->body = strdup_or_die(" ");
645 rsp->content_length = 1;
647 if (csp->config->feature_flags & RUNTIME_FEATURE_EMPTY_DOC_RETURNS_OK)
650 * Workaround for firefox bug 492459
651 * https://bugzilla.mozilla.org/show_bug.cgi?id=492459
652 * Return a 200 OK status for pages blocked with +handle-as-empty-document
653 * if the "handle-as-empty-doc-returns-ok" runtime config option is set.
655 rsp->status = strdup_or_die("200 Request blocked by Privoxy");
659 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
662 if (new_content_type != 0)
664 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
665 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
667 free_http_response(rsp);
668 return cgi_error_memory();
675 * Else, generate an HTML "blocked" message:
679 struct map * exports;
681 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
683 exports = default_exports(csp, NULL);
686 free_http_response(rsp);
687 return cgi_error_memory();
690 #ifdef FEATURE_FORCE_LOAD
691 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
693 * Export the force conditional block killer if
695 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
696 * - Privoxy is configured to enforce blocks, or
697 * - it's a CONNECT request and enforcing wouldn't work anyway.
699 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
700 || (0 == strcmpic(csp->http->gpc, "connect")))
701 #endif /* ndef FEATURE_FORCE_LOAD */
703 err = map_block_killer(exports, "force-support");
706 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
707 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
708 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
709 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
712 const char *block_reason;
713 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
715 block_reason = csp->action->string[ACTION_STRING_BLOCK];
719 assert(connect_port_is_forbidden(csp));
720 block_reason = "Forbidden CONNECT port.";
722 err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
727 free_http_response(rsp);
728 return cgi_error_memory();
731 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
734 free_http_response(rsp);
735 return cgi_error_memory();
738 rsp->crunch_reason = BLOCKED;
740 return finish_http_response(csp, rsp);
746 /*********************************************************************
748 * Function : trust_url FIXME: I should be called distrust_url
750 * Description : Calls is_untrusted_url to determine if the URL is trusted
751 * and if not, returns a HTTP 403 response with a reject message.
754 * 1 : csp = Current client state (buffers, headers, etc...)
756 * Returns : NULL => trusted, else http_response.
758 *********************************************************************/
759 struct http_response *trust_url(struct client_state *csp)
761 struct http_response *rsp;
762 struct map * exports;
763 char buf[BUFFER_SIZE];
765 struct pattern_spec **tl;
766 struct pattern_spec *t;
770 * Don't bother to work on trusted URLs
772 if (!is_untrusted_url(csp))
778 * Else, prepare a response:
780 if (NULL == (rsp = alloc_http_response()))
782 return cgi_error_memory();
785 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
786 exports = default_exports(csp, NULL);
789 free_http_response(rsp);
790 return cgi_error_memory();
794 * Export the protocol, host, port, and referrer information
796 err = map(exports, "hostport", 1, csp->http->hostport, 1);
797 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
798 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
800 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
802 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
806 if (!err) err = map(exports, "referrer", 1, "none set", 1);
812 free_http_response(rsp);
813 return cgi_error_memory();
817 * Export the trust list
819 p = strdup_or_die("");
820 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
822 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
823 string_append(&p, buf);
825 err = map(exports, "trusted-referrers", 1, p, 0);
830 free_http_response(rsp);
831 return cgi_error_memory();
835 * Export the trust info, if available
837 if (csp->config->trust_info->first)
839 struct list_entry *l;
841 p = strdup_or_die("");
842 for (l = csp->config->trust_info->first; l ; l = l->next)
844 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
845 string_append(&p, buf);
847 err = map(exports, "trust-info", 1, p, 0);
851 err = map_block_killer(exports, "have-trust-info");
857 free_http_response(rsp);
858 return cgi_error_memory();
862 * Export the force conditional block killer if
864 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
865 * - Privoxy is configured to enforce blocks, or
866 * - it's a CONNECT request and enforcing wouldn't work anyway.
868 #ifdef FEATURE_FORCE_LOAD
869 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
870 || (0 == strcmpic(csp->http->gpc, "connect")))
872 err = map_block_killer(exports, "force-support");
876 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
878 #else /* ifndef FEATURE_FORCE_LOAD */
879 err = map_block_killer(exports, "force-support");
880 #endif /* ndef FEATURE_FORCE_LOAD */
885 free_http_response(rsp);
886 return cgi_error_memory();
892 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
895 free_http_response(rsp);
896 return cgi_error_memory();
898 rsp->crunch_reason = UNTRUSTED;
900 return finish_http_response(csp, rsp);
902 #endif /* def FEATURE_TRUST */
905 /*********************************************************************
907 * Function : compile_dynamic_pcrs_job_list
909 * Description : Compiles a dynamic pcrs job list (one with variables
910 * resolved at request time)
913 * 1 : csp = Current client state (buffers, headers, etc...)
914 * 2 : b = The filter list to compile
916 * Returns : NULL in case of errors, otherwise the
919 *********************************************************************/
920 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
922 struct list_entry *pattern;
923 pcrs_job *job_list = NULL;
924 pcrs_job *dummy = NULL;
925 pcrs_job *lastjob = NULL;
928 const struct pcrs_variable variables[] =
930 {"url", csp->http->url, 1},
931 {"path", csp->http->path, 1},
932 {"host", csp->http->host, 1},
933 {"origin", csp->ip_addr_str, 1},
934 {"listen-address", csp->listen_addr_str, 1},
938 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
940 assert(pattern->str != NULL);
942 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
945 log_error(LOG_LEVEL_ERROR,
946 "Compiling dynamic pcrs job '%s' for '%s' failed with error code %d: %s",
947 pattern->str, b->name, error, pcrs_strerror(error));
952 if (error == PCRS_WARN_TRUNCATION)
954 log_error(LOG_LEVEL_ERROR,
955 "At least one of the variables in \'%s\' had to "
956 "be truncated before compilation", pattern->str);
958 if (job_list == NULL)
964 lastjob->next = dummy;
974 /*********************************************************************
976 * Function : rewrite_url
978 * Description : Rewrites a URL with a single pcrs command
979 * and returns the result if it differs from the
980 * original and isn't obviously invalid.
983 * 1 : old_url = URL to rewrite.
984 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
987 * Returns : NULL if the pcrs_command didn't change the url, or
988 * the result of the modification.
990 *********************************************************************/
991 char *rewrite_url(char *old_url, const char *pcrs_command)
993 char *new_url = NULL;
997 assert(pcrs_command);
999 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1003 log_error(LOG_LEVEL_REDIRECTS,
1004 "pcrs command \"%s\" didn't change \"%s\".",
1005 pcrs_command, old_url);
1010 log_error(LOG_LEVEL_REDIRECTS,
1011 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1012 pcrs_command, old_url, pcrs_strerror(hits));
1015 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1017 log_error(LOG_LEVEL_ERROR,
1018 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1019 "but the result doesn't look like a valid URL and will be ignored.",
1020 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1025 log_error(LOG_LEVEL_REDIRECTS,
1026 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1027 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1035 #ifdef FEATURE_FAST_REDIRECTS
1036 /*********************************************************************
1038 * Function : get_last_url
1040 * Description : Search for the last URL inside a string.
1041 * If the string already is a URL, it will
1042 * be the first URL found.
1045 * 1 : subject = the string to check
1046 * 2 : redirect_mode = +fast-redirect{} mode
1048 * Returns : NULL if no URL was found, or
1049 * the last URL found.
1051 *********************************************************************/
1052 char *get_last_url(char *subject, const char *redirect_mode)
1054 char *new_url = NULL;
1058 assert(redirect_mode);
1060 subject = strdup(subject);
1061 if (subject == NULL)
1063 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1067 if (0 == strcmpic(redirect_mode, "check-decoded-url") && strchr(subject, '%'))
1069 char *url_segment = NULL;
1070 char **url_segments;
1071 size_t max_segments;
1074 log_error(LOG_LEVEL_REDIRECTS,
1075 "Checking \"%s\" for encoded redirects.", subject);
1078 * Check each parameter in the URL separately.
1079 * Sectionize the URL at "?" and "&",
1080 * go backwards through the segments, URL-decode them
1081 * and look for a URL in the decoded result.
1082 * Stop the search after the first match.
1084 * XXX: This estimate is guaranteed to be high enough as we
1085 * let ssplit() ignore empty fields, but also a bit wasteful.
1087 max_segments = strlen(subject) / 2;
1088 url_segments = malloc(max_segments * sizeof(char *));
1090 if (NULL == url_segments)
1092 log_error(LOG_LEVEL_ERROR,
1093 "Out of memory while decoding URL: %s", subject);
1098 segments = ssplit(subject, "?&", url_segments, max_segments);
1100 while (segments-- > 0)
1102 char *dtoken = url_decode(url_segments[segments]);
1105 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", url_segments[segments]);
1108 url_segment = strstr(dtoken, "http://");
1109 if (NULL == url_segment)
1111 url_segment = strstr(dtoken, "https://");
1113 if (NULL != url_segment)
1115 url_segment = strdup_or_die(url_segment);
1122 freez(url_segments);
1124 if (url_segment == NULL)
1128 subject = url_segment;
1132 /* Look for a URL inside this one, without decoding anything. */
1133 log_error(LOG_LEVEL_REDIRECTS,
1134 "Checking \"%s\" for unencoded redirects.", subject);
1138 * Find the last URL encoded in the request
1141 while ((tmp = strstr(tmp, "http://")) != NULL)
1145 tmp = (new_url != NULL) ? new_url : subject;
1146 while ((tmp = strstr(tmp, "https://")) != NULL)
1151 if ((new_url != NULL)
1152 && ( (new_url != subject)
1153 || (0 == strncmpic(subject, "http://", 7))
1154 || (0 == strncmpic(subject, "https://", 8))
1158 * Return new URL if we found a redirect
1159 * or if the subject already was a URL.
1161 * The second case makes sure that we can
1162 * chain get_last_url after another redirection check
1163 * (like rewrite_url) without losing earlier redirects.
1165 new_url = strdup(new_url);
1174 #endif /* def FEATURE_FAST_REDIRECTS */
1177 /*********************************************************************
1179 * Function : redirect_url
1181 * Description : Checks if Privoxy should answer the request with
1182 * a HTTP redirect and generates the redirect if
1186 * 1 : csp = Current client state (buffers, headers, etc...)
1188 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1190 *********************************************************************/
1191 struct http_response *redirect_url(struct client_state *csp)
1193 struct http_response *rsp;
1194 #ifdef FEATURE_FAST_REDIRECTS
1196 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1197 * as compile-time option? The user can easily disable
1198 * it in his action file.
1200 char * redirect_mode;
1201 #endif /* def FEATURE_FAST_REDIRECTS */
1202 char *old_url = NULL;
1203 char *new_url = NULL;
1204 char *redirection_string;
1206 if ((csp->action->flags & ACTION_REDIRECT))
1208 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1211 * If the redirection string begins with 's',
1212 * assume it's a pcrs command, otherwise treat it as
1213 * properly formatted URL and use it for the redirection
1216 * According to (the now obsolete) RFC 2616 section 14.30
1217 * the URL has to be absolute and if the user tries:
1218 * +redirect{sadly/this/will/be/parsed/as/pcrs_command.html}
1219 * she would get undefined results anyway.
1221 * RFC 7231 7.1.2 actually allows relative references,
1222 * but those start with a leading slash (RFC 3986 4.2) and
1223 * thus can't be mistaken for pcrs commands either.
1226 if (*redirection_string == 's')
1228 old_url = csp->http->url;
1229 new_url = rewrite_url(old_url, redirection_string);
1233 log_error(LOG_LEVEL_REDIRECTS,
1234 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1235 redirection_string);
1236 new_url = strdup(redirection_string);
1240 #ifdef FEATURE_FAST_REDIRECTS
1241 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1243 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1246 * If it exists, use the previously rewritten URL as input
1247 * otherwise just use the old path.
1249 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1250 new_url = get_last_url(old_url, redirect_mode);
1255 * Disable redirect checkers, so that they
1256 * will be only run more than once if the user
1257 * also enables them through tags.
1259 * From a performance point of view
1260 * it doesn't matter, but the duplicated
1261 * log messages are annoying.
1263 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1264 #endif /* def FEATURE_FAST_REDIRECTS */
1265 csp->action->flags &= ~ACTION_REDIRECT;
1267 /* Did any redirect action trigger? */
1270 if (url_requires_percent_encoding(new_url))
1273 log_error(LOG_LEVEL_REDIRECTS, "Percent-encoding redirect URL: %N",
1274 strlen(new_url), new_url);
1275 encoded_url = percent_encode_url(new_url);
1277 if (encoded_url == NULL)
1279 return cgi_error_memory();
1281 new_url = encoded_url;
1282 assert(FALSE == url_requires_percent_encoding(new_url));
1285 if (0 == strcmpic(new_url, csp->http->url))
1287 log_error(LOG_LEVEL_ERROR,
1288 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1289 csp->http->url, new_url);
1294 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1296 if (NULL == (rsp = alloc_http_response()))
1299 return cgi_error_memory();
1302 rsp->status = strdup_or_die("302 Local Redirect from Privoxy");
1303 if (enlist_unique_header(rsp->headers, "Location", new_url))
1306 free_http_response(rsp);
1307 return cgi_error_memory();
1309 rsp->crunch_reason = REDIRECTED;
1312 return finish_http_response(csp, rsp);
1316 /* Only reached if no redirect is required */
1322 #ifdef FEATURE_IMAGE_BLOCKING
1323 /*********************************************************************
1325 * Function : is_imageurl
1327 * Description : Given a URL, decide whether it should be treated
1328 * as image URL or not.
1331 * 1 : csp = Current client state (buffers, headers, etc...)
1333 * Returns : True (nonzero) if URL is an image URL, false (0)
1336 *********************************************************************/
1337 int is_imageurl(const struct client_state *csp)
1339 return ((csp->action->flags & ACTION_IMAGE) != 0);
1342 #endif /* def FEATURE_IMAGE_BLOCKING */
1345 #ifdef FEATURE_TRUST
1346 /*********************************************************************
1348 * Function : is_untrusted_url
1350 * Description : Should we "distrust" this URL (and block it)?
1352 * Yes if it matches a line in the trustfile, or if the
1353 * referrer matches a line starting with "+" in the
1358 * 1 : csp = Current client state (buffers, headers, etc...)
1360 * Returns : 0 => trusted, 1 => untrusted
1362 *********************************************************************/
1363 int is_untrusted_url(const struct client_state *csp)
1365 struct file_list *fl;
1366 struct block_spec *b;
1367 struct pattern_spec **trusted_url;
1368 struct http_request rhttp[1];
1369 const char * referer;
1373 * If we don't have a trustlist, we trust everybody
1375 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1380 memset(rhttp, '\0', sizeof(*rhttp));
1383 * Do we trust the request URL itself?
1385 for (b = b->next; b ; b = b->next)
1387 if (url_match(b->url, csp->http))
1393 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1395 /* no referrer was supplied */
1401 * If not, do we maybe trust its referrer?
1403 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
1409 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1411 if (url_match(*trusted_url, rhttp))
1413 /* if the URL's referrer is from a trusted referrer, then
1414 * add the target spec to the trustfile as an unblocked
1415 * domain and return 0 (which means it's OK).
1420 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1424 char * new_entry = strdup_or_die("~");
1426 string_append(&new_entry, csp->http->hostport);
1428 path = csp->http->path;
1429 if ( (path[0] == '/')
1431 && ((path_end = strchr(path + 2, '/')) != NULL))
1433 /* since this path points into a user's home space
1434 * be sure to include this spec in the trustfile.
1436 long path_len = path_end - path; /* save offset */
1437 path = strdup(path); /* Copy string */
1440 path_end = path + path_len; /* regenerate ptr to new buffer */
1441 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1443 string_join(&new_entry, path);
1447 * Give a reason for generating this entry.
1449 string_append(&new_entry, " # Trusted referrer was: ");
1450 string_append(&new_entry, referer);
1452 if (new_entry != NULL)
1454 if (-1 == fprintf(fp, "%s\n", new_entry))
1456 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1457 new_entry, csp->config->trustfile);
1463 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1464 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1471 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1472 csp->http->hostport, csp->config->trustfile);
1480 #endif /* def FEATURE_TRUST */
1483 /*********************************************************************
1485 * Function : get_filter
1487 * Description : Get a filter with a given name and type.
1488 * Note that taggers are filters, too.
1491 * 1 : csp = Current client state (buffers, headers, etc...)
1492 * 2 : requested_name = Name of the content filter to get
1493 * 3 : requested_type = Type of the filter to tagger to lookup
1495 * Returns : A pointer to the requested filter
1496 * or NULL if the filter wasn't found
1498 *********************************************************************/
1499 struct re_filterfile_spec *get_filter(const struct client_state *csp,
1500 const char *requested_name,
1501 enum filter_type requested_type)
1504 struct re_filterfile_spec *b;
1505 struct file_list *fl;
1507 for (i = 0; i < MAX_AF_FILES; i++)
1510 if ((NULL == fl) || (NULL == fl->f))
1513 * Either there are no filter files left or this
1514 * filter file just contains no valid filters.
1516 * Continue to be sure we don't miss valid filter
1517 * files that are chained after empty or invalid ones.
1522 for (b = fl->f; b != NULL; b = b->next)
1524 if (b->type != requested_type)
1526 /* The callers isn't interested in this filter type. */
1529 if (strcmp(b->name, requested_name) == 0)
1531 /* The requested filter has been found. Abort search. */
1537 /* No filter with the given name and type exists. */
1543 /*********************************************************************
1545 * Function : pcrs_filter_response
1547 * Description : Execute all text substitutions from all applying
1548 * +filter actions on the text buffer that's been
1549 * accumulated in csp->iob->buf.
1552 * 1 : csp = Current client state (buffers, headers, etc...)
1554 * Returns : a pointer to the (newly allocated) modified buffer.
1555 * or NULL if there were no hits or something went wrong
1557 *********************************************************************/
1558 static char *pcrs_filter_response(struct client_state *csp)
1561 size_t size, prev_size;
1567 struct re_filterfile_spec *b;
1568 struct list_entry *filtername;
1573 if (csp->iob->cur >= csp->iob->eod)
1578 if (filters_available(csp) == FALSE)
1580 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1581 "content filtering enabled, but no content filters available.");
1585 size = (size_t)(csp->iob->eod - csp->iob->cur);
1586 old = csp->iob->cur;
1589 * For all applying +filter actions, look if a filter by that
1590 * name exists and if yes, execute it's pcrs_joblist on the
1593 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1594 filtername != NULL; filtername = filtername->next)
1596 int current_hits = 0; /* Number of hits caused by this filter */
1597 int job_number = 0; /* Which job we're currently executing */
1598 int job_hits = 0; /* How many hits the current job caused */
1601 b = get_filter(csp, filtername->str, FT_CONTENT_FILTER);
1607 joblist = b->joblist;
1609 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1611 if (NULL == joblist)
1613 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1618 /* Apply all jobs from the joblist */
1619 for (job = joblist; NULL != job; job = job->next)
1622 job_hits = pcrs_execute(job, old, size, &new, &size);
1627 * That went well. Continue filtering
1628 * and use the result of this job as
1629 * input for the next one.
1631 current_hits += job_hits;
1632 if (old != csp->iob->cur)
1641 * This job caused an unexpected error. Inform the user
1642 * and skip the rest of the jobs in this filter. We could
1643 * continue with the next job, but usually the jobs
1644 * depend on each other or are similar enough to
1645 * fail for the same reason.
1647 * At the moment our pcrs expects the error codes of pcre 3.4,
1648 * but newer pcre versions can return additional error codes.
1649 * As a result pcrs_strerror()'s error message might be
1650 * "Unknown error ...", therefore we print the numerical value
1653 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1654 * should we use LOG_LEVEL_RE_FILTER instead?
1656 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1657 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1662 if (b->dynamic) pcrs_free_joblist(joblist);
1664 log_error(LOG_LEVEL_RE_FILTER,
1665 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
1666 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1668 hits += current_hits;
1672 * If there were no hits, destroy our copy and let
1673 * chat() use the original in csp->iob
1681 csp->flags |= CSP_FLAG_MODIFIED;
1682 csp->content_length = size;
1683 clear_iob(csp->iob);
1690 #ifdef FEATURE_EXTERNAL_FILTERS
1691 /*********************************************************************
1693 * Function : get_external_filter
1695 * Description : Lookup the code to execute for an external filter.
1696 * Masks the misuse of the re_filterfile_spec.
1699 * 1 : csp = Current client state (buffers, headers, etc...)
1700 * 2 : name = Name of the content filter to get
1702 * Returns : A pointer to the requested code
1703 * or NULL if the filter wasn't found
1705 *********************************************************************/
1706 static const char *get_external_filter(const struct client_state *csp,
1709 struct re_filterfile_spec *external_filter;
1711 external_filter = get_filter(csp, name, FT_EXTERNAL_CONTENT_FILTER);
1712 if (external_filter == NULL)
1714 log_error(LOG_LEVEL_FATAL,
1715 "Didn't find stuff to execute for external filter: %s",
1719 return external_filter->patterns->first->str;
1724 /*********************************************************************
1726 * Function : set_privoxy_variables
1728 * Description : Sets a couple of privoxy-specific environment variables
1731 * 1 : csp = Current client state (buffers, headers, etc...)
1735 *********************************************************************/
1736 static void set_privoxy_variables(const struct client_state *csp)
1743 { "PRIVOXY_URL", csp->http->url },
1744 { "PRIVOXY_PATH", csp->http->path },
1745 { "PRIVOXY_HOST", csp->http->host },
1746 { "PRIVOXY_ORIGIN", csp->ip_addr_str },
1747 { "PRIVOXY_LISTEN_ADDRESS", csp->listen_addr_str },
1750 for (i = 0; i < SZ(env); i++)
1752 if (setenv(env[i].name, env[i].value, 1))
1754 log_error(LOG_LEVEL_ERROR, "Failed to set %s=%s: %E",
1755 env[i].name, env[i].value);
1761 /*********************************************************************
1763 * Function : execute_external_filter
1765 * Description : Pipe content into external filter and return the output
1768 * 1 : csp = Current client state (buffers, headers, etc...)
1769 * 2 : name = Name of the external filter to execute
1770 * 3 : content = The original content to filter
1771 * 4 : size = The size of the content buffer
1773 * Returns : a pointer to the (newly allocated) modified buffer.
1774 * or NULL if there were no hits or something went wrong
1776 *********************************************************************/
1777 static char *execute_external_filter(const struct client_state *csp,
1778 const char *name, char *content, size_t *size)
1781 char file_name[FILENAME_MAX];
1783 char *filter_output;
1787 const char *external_filter;
1789 if (csp->config->temporary_directory == NULL)
1791 log_error(LOG_LEVEL_ERROR,
1792 "No temporary-directory configured. Can't execute filter: %s",
1797 external_filter = get_external_filter(csp, name);
1799 if (sizeof(file_name) < snprintf(file_name, sizeof(file_name),
1800 "%s/privoxy-XXXXXXXX", csp->config->temporary_directory))
1802 log_error(LOG_LEVEL_ERROR, "temporary-directory path too long");
1806 fd = mkstemp(file_name);
1809 log_error(LOG_LEVEL_ERROR, "mkstemp() failed to create %s: %E", file_name);
1813 fp = fdopen(fd, "w");
1816 log_error(LOG_LEVEL_ERROR, "fdopen() failed: %E");
1822 * The size may be zero if a previous filter discarded everything.
1824 * This isn't necessary unintentional, so we just don't try
1825 * to fwrite() nothing and let the user deal with the rest.
1827 if ((*size != 0) && fwrite(content, *size, 1, fp) != 1)
1829 log_error(LOG_LEVEL_ERROR, "fwrite(..., %d, 1, ..) failed: %E", *size);
1836 if (sizeof(cmd) < snprintf(cmd, sizeof(cmd), "%s < %s", external_filter, file_name))
1838 log_error(LOG_LEVEL_ERROR,
1839 "temporary-directory or external filter path too long");
1844 log_error(LOG_LEVEL_RE_FILTER, "Executing '%s': %s", name, cmd);
1847 * The locking is necessary to prevent other threads
1848 * from overwriting the environment variables before
1849 * the popen fork. Afterwards this no longer matters.
1851 privoxy_mutex_lock(&external_filter_mutex);
1852 set_privoxy_variables(csp);
1853 fp = popen(cmd, "r");
1854 privoxy_mutex_unlock(&external_filter_mutex);
1857 log_error(LOG_LEVEL_ERROR, "popen(\"%s\", \"r\") failed: %E", cmd);
1862 /* Allocate at least one byte */
1863 filter_output = malloc_or_die(*size + 1);
1866 while (!feof(fp) && !ferror(fp))
1869 /* Could be bigger ... */
1870 enum { READ_LENGTH = 2048 };
1872 if (new_size + READ_LENGTH >= *size)
1876 /* Could be considered wasteful if the content is 'large'. */
1877 *size += (*size >= READ_LENGTH) ? *size : READ_LENGTH;
1879 p = realloc(filter_output, *size);
1882 log_error(LOG_LEVEL_ERROR, "Out of memory while reading "
1883 "external filter output. Using what we got so far.");
1888 assert(new_size + READ_LENGTH < *size);
1889 len = fread(&filter_output[new_size], 1, READ_LENGTH, fp);
1899 log_error(LOG_LEVEL_ERROR, "Executing %s failed: %E", cmd);
1903 log_error(LOG_LEVEL_RE_FILTER,
1904 "Executing '%s' resulted in return value %d. "
1905 "Read %d of up to %d bytes.", name, (ret >> 8), new_size, *size);
1911 return filter_output;
1914 #endif /* def FEATURE_EXTERNAL_FILTERS */
1917 /*********************************************************************
1919 * Function : gif_deanimate_response
1921 * Description : Deanimate the GIF image that has been accumulated in
1922 * csp->iob->buf, set csp->content_length to the modified
1923 * size and raise the CSP_FLAG_MODIFIED flag.
1926 * 1 : csp = Current client state (buffers, headers, etc...)
1928 * Returns : a pointer to the (newly allocated) modified buffer.
1929 * or NULL in case something went wrong.
1931 *********************************************************************/
1933 char *gif_deanimate_response(struct client_state *csp)
1935 static char *gif_deanimate_response(struct client_state *csp)
1938 struct binbuffer *in, *out;
1942 size = (size_t)(csp->iob->eod - csp->iob->cur);
1944 in = zalloc_or_die(sizeof(*in));
1945 out = zalloc_or_die(sizeof(*out));
1947 in->buffer = csp->iob->cur;
1950 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1952 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1959 if ((int)size == out->offset)
1961 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1965 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1967 csp->content_length = out->offset;
1968 csp->flags |= CSP_FLAG_MODIFIED;
1978 /*********************************************************************
1980 * Function : get_filter_function
1982 * Description : Decides which content filter function has
1983 * to be applied (if any). Only considers functions
1984 * for internal filters which are mutually-exclusive.
1987 * 1 : csp = Current client state (buffers, headers, etc...)
1989 * Returns : The content filter function to run, or
1990 * NULL if no content filter is active
1992 *********************************************************************/
1993 static filter_function_ptr get_filter_function(const struct client_state *csp)
1995 filter_function_ptr filter_function = NULL;
1998 * Choose the applying filter function based on
1999 * the content type and action settings.
2001 if ((csp->content_type & CT_TEXT) &&
2002 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2004 filter_function = pcrs_filter_response;
2006 else if ((csp->content_type & CT_GIF) &&
2007 (csp->action->flags & ACTION_DEANIMATE))
2009 filter_function = gif_deanimate_response;
2012 return filter_function;
2016 /*********************************************************************
2018 * Function : remove_chunked_transfer_coding
2020 * Description : In-situ remove the "chunked" transfer coding as defined
2021 * in RFC 7230 4.1 from a buffer. XXX: The implementation
2022 * is neither complete nor compliant (TODO #129).
2025 * 1 : buffer = Pointer to the text buffer
2026 * 2 : size = In: Number of bytes to be processed,
2027 * Out: Number of bytes after de-chunking.
2028 * (undefined in case of errors)
2030 * Returns : JB_ERR_OK for success,
2031 * JB_ERR_PARSE otherwise
2033 *********************************************************************/
2035 extern jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2037 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2041 unsigned int chunksize = 0;
2042 char *from_p, *to_p;
2043 const char *end_of_buffer = buffer + *size;
2047 log_error(LOG_LEVEL_FATAL, "Invalid chunked input. Buffer is empty.");
2048 return JB_ERR_PARSE;
2052 from_p = to_p = buffer;
2054 if (sscanf(buffer, "%x", &chunksize) != 1)
2056 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2057 return JB_ERR_PARSE;
2060 while (chunksize > 0U)
2063 * If the chunk-size is valid, we should have at least
2064 * chunk-size bytes of chunk-data and five bytes of
2065 * meta data (chunk-size, CRLF, CRLF) left in the buffer.
2067 if (chunksize + 5 >= *size - newsize)
2069 log_error(LOG_LEVEL_ERROR,
2070 "Chunk size %u exceeds buffered data left. "
2071 "Already digested %u of %u buffered bytes.",
2072 chunksize, (unsigned int)newsize, (unsigned int)*size);
2073 return JB_ERR_PARSE;
2077 * Skip the chunk-size, the optional chunk-ext and the CRLF
2078 * that is supposed to be located directly before the start
2081 if (NULL == (from_p = strstr(from_p, "\r\n")))
2083 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2084 return JB_ERR_PARSE;
2089 * The previous strstr() does not enforce chunk-validity
2090 * and is sattisfied as long a CRLF is left in the buffer.
2092 * Make sure the bytes we consider chunk-data are within
2095 if (from_p + chunksize >= end_of_buffer)
2097 log_error(LOG_LEVEL_ERROR,
2098 "End of chunk is beyond the end of the buffer.");
2099 return JB_ERR_PARSE;
2102 memmove(to_p, from_p, (size_t) chunksize);
2103 newsize += chunksize;
2104 to_p = buffer + newsize;
2105 from_p += chunksize;
2108 * Not merging this check with the previous one allows us
2109 * to keep chunks without trailing CRLF. It's not clear
2110 * if we actually have to care about those, though.
2112 if (from_p + 2 >= end_of_buffer)
2114 log_error(LOG_LEVEL_ERROR, "Not enough room for trailing CRLF.");
2115 return JB_ERR_PARSE;
2118 if (sscanf(from_p, "%x", &chunksize) != 1)
2120 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
2125 /* XXX: Should get its own loglevel. */
2126 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
2135 /*********************************************************************
2137 * Function : prepare_for_filtering
2139 * Description : If necessary, de-chunks and decompresses
2140 * the content so it can get filterd.
2143 * 1 : csp = Current client state (buffers, headers, etc...)
2145 * Returns : JB_ERR_OK for success,
2146 * JB_ERR_PARSE otherwise
2148 *********************************************************************/
2149 static jb_err prepare_for_filtering(struct client_state *csp)
2151 jb_err err = JB_ERR_OK;
2154 * If the body has a "chunked" transfer-encoding,
2155 * get rid of it, adjusting size and iob->eod
2157 if (csp->flags & CSP_FLAG_CHUNKED)
2159 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2161 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2162 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2163 if (JB_ERR_OK == err)
2165 csp->iob->eod = csp->iob->cur + size;
2166 csp->flags |= CSP_FLAG_MODIFIED;
2170 return JB_ERR_PARSE;
2176 * If the body has a supported transfer-encoding,
2177 * decompress it, adjusting size and iob->eod.
2179 if (csp->content_type & (CT_GZIP|CT_DEFLATE))
2181 if (0 == csp->iob->eod - csp->iob->cur)
2183 /* Nothing left after de-chunking. */
2187 err = decompress_iob(csp);
2189 if (JB_ERR_OK == err)
2191 csp->flags |= CSP_FLAG_MODIFIED;
2192 csp->content_type &= ~CT_TABOO;
2197 * Unset CT_GZIP and CT_DEFLATE to remember not
2198 * to modify the Content-Encoding header later.
2200 csp->content_type &= ~CT_GZIP;
2201 csp->content_type &= ~CT_DEFLATE;
2210 /*********************************************************************
2212 * Function : execute_content_filters
2214 * Description : Executes a given content filter.
2217 * 1 : csp = Current client state (buffers, headers, etc...)
2219 * Returns : Pointer to the modified buffer, or
2220 * NULL if filtering failed or wasn't necessary.
2222 *********************************************************************/
2223 char *execute_content_filters(struct client_state *csp)
2226 filter_function_ptr content_filter;
2228 assert(content_filters_enabled(csp->action));
2230 if (0 == csp->iob->eod - csp->iob->cur)
2233 * No content (probably status code 301, 302 ...),
2234 * no filtering necessary.
2239 if (JB_ERR_OK != prepare_for_filtering(csp))
2242 * failed to de-chunk or decompress.
2247 if (0 == csp->iob->eod - csp->iob->cur)
2250 * Clown alarm: chunked and/or compressed nothing delivered.
2255 content_filter = get_filter_function(csp);
2256 content = (content_filter != NULL) ? (*content_filter)(csp) : NULL;
2258 #ifdef FEATURE_EXTERNAL_FILTERS
2259 if ((csp->content_type & CT_TEXT) &&
2260 !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]))
2262 struct list_entry *filtername;
2263 size_t size = (size_t)csp->content_length;
2265 if (content == NULL)
2267 content = csp->iob->cur;
2268 size = (size_t)(csp->iob->eod - csp->iob->cur);
2271 for (filtername = csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]->first;
2272 filtername ; filtername = filtername->next)
2274 char *result = execute_external_filter(csp, filtername->str, content, &size);
2277 if (content != csp->iob->cur)
2284 csp->flags |= CSP_FLAG_MODIFIED;
2285 csp->content_length = size;
2287 #endif /* def FEATURE_EXTERNAL_FILTERS */
2294 /*********************************************************************
2296 * Function : get_url_actions
2298 * Description : Gets the actions for this URL.
2301 * 1 : csp = Current client state (buffers, headers, etc...)
2302 * 2 : http = http_request request for blocked URLs
2306 *********************************************************************/
2307 void get_url_actions(struct client_state *csp, struct http_request *http)
2309 struct file_list *fl;
2310 struct url_actions *b;
2313 init_current_action(csp->action);
2315 for (i = 0; i < MAX_AF_FILES; i++)
2317 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2322 #ifdef FEATURE_CLIENT_TAGS
2323 apply_url_actions(csp->action, http, csp->client_tags, b);
2325 apply_url_actions(csp->action, http, b);
2332 /*********************************************************************
2334 * Function : apply_url_actions
2336 * Description : Applies a list of URL actions.
2339 * 1 : action = Destination.
2340 * 2 : http = Current URL
2341 * 3 : client_tags = list of client tags
2342 * 4 : b = list of URL actions to apply
2346 *********************************************************************/
2347 static void apply_url_actions(struct current_action_spec *action,
2348 struct http_request *http,
2349 #ifdef FEATURE_CLIENT_TAGS
2350 const struct list *client_tags,
2352 struct url_actions *b)
2356 /* Should never happen */
2360 for (b = b->next; NULL != b; b = b->next)
2362 if (url_match(b->url, http))
2364 merge_current_action(action, b->action);
2366 #ifdef FEATURE_CLIENT_TAGS
2367 if (client_tag_match(b->url, client_tags))
2369 merge_current_action(action, b->action);
2376 /*********************************************************************
2378 * Function : get_forward_override_settings
2380 * Description : Returns forward settings as specified with the
2381 * forward-override{} action. forward-override accepts
2382 * forward lines similar to the one used in the
2383 * configuration file, but without the URL pattern.
2389 * in the configuration file can be replaced with
2390 * the action section:
2392 * {+forward-override{forward .}}
2396 * 1 : csp = Current client state (buffers, headers, etc...)
2398 * Returns : Pointer to forwarding structure in case of success.
2399 * Invalid syntax is fatal.
2401 *********************************************************************/
2402 static const struct forward_spec *get_forward_override_settings(struct client_state *csp)
2404 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2405 char forward_settings[BUFFER_SIZE];
2406 char *http_parent = NULL;
2407 /* variable names were chosen for consistency reasons. */
2408 struct forward_spec *fwd = NULL;
2412 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2413 /* Should be enforced by load_one_actions_file() */
2414 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2416 /* Create a copy ssplit can modify */
2417 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2419 if (NULL != csp->fwd)
2422 * XXX: Currently necessary to prevent memory
2423 * leaks when the show-url-info cgi page is visited.
2425 unload_forward_spec(csp->fwd);
2429 * allocate a new forward node, valid only for
2430 * the lifetime of this request. Save its location
2431 * in csp as well, so sweep() can free it later on.
2433 fwd = csp->fwd = zalloc_or_die(sizeof(*fwd));
2435 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec));
2436 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2438 fwd->type = SOCKS_NONE;
2440 /* Parse the parent HTTP proxy host:port */
2441 http_parent = vec[1];
2444 else if ((vec_count == 2) && !strcasecmp(vec[0], "forward-webserver"))
2446 fwd->type = FORWARD_WEBSERVER;
2448 /* Parse the parent HTTP server host:port */
2449 http_parent = vec[1];
2452 else if (vec_count == 3)
2454 char *socks_proxy = NULL;
2456 if (!strcasecmp(vec[0], "forward-socks4"))
2458 fwd->type = SOCKS_4;
2459 socks_proxy = vec[1];
2461 else if (!strcasecmp(vec[0], "forward-socks4a"))
2463 fwd->type = SOCKS_4A;
2464 socks_proxy = vec[1];
2466 else if (!strcasecmp(vec[0], "forward-socks5"))
2468 fwd->type = SOCKS_5;
2469 socks_proxy = vec[1];
2471 else if (!strcasecmp(vec[0], "forward-socks5t"))
2473 fwd->type = SOCKS_5T;
2474 socks_proxy = vec[1];
2477 if (NULL != socks_proxy)
2479 /* Parse the SOCKS proxy host[:port] */
2480 fwd->gateway_port = 1080;
2481 parse_forwarder_address(socks_proxy,
2482 &fwd->gateway_host, &fwd->gateway_port);
2484 http_parent = vec[2];
2488 if (NULL == http_parent)
2490 log_error(LOG_LEVEL_FATAL,
2491 "Invalid forward-override syntax in: %s", forward_override_line);
2492 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2495 /* Parse http forwarding settings */
2496 if (strcmp(http_parent, ".") != 0)
2498 fwd->forward_port = 8000;
2499 parse_forwarder_address(http_parent,
2500 &fwd->forward_host, &fwd->forward_port);
2503 assert (NULL != fwd);
2505 log_error(LOG_LEVEL_CONNECT,
2506 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2512 /*********************************************************************
2514 * Function : forward_url
2516 * Description : Should we forward this to another proxy?
2519 * 1 : csp = Current client state (buffers, headers, etc...)
2520 * 2 : http = http_request request for current URL
2522 * Returns : Pointer to forwarding information.
2524 *********************************************************************/
2525 const struct forward_spec *forward_url(struct client_state *csp,
2526 const struct http_request *http)
2528 static const struct forward_spec fwd_default[1]; /* Zero'ed due to being static. */
2529 struct forward_spec *fwd = csp->config->forward;
2531 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2533 return get_forward_override_settings(csp);
2543 if (url_match(fwd->url, http))
2554 /*********************************************************************
2556 * Function : direct_response
2558 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2559 * request and if so, return a HTTP 501 to the client.
2561 * FIXME: I have a stupid name and I should handle the
2562 * requests properly. Still, what we do here is rfc-
2563 * compliant, whereas ignoring or forwarding are not.
2566 * 1 : csp = Current client state (buffers, headers, etc...)
2568 * Returns : http_response if , NULL if nonmatch or handler fail
2570 *********************************************************************/
2571 struct http_response *direct_response(struct client_state *csp)
2573 struct http_response *rsp;
2574 struct list_entry *p;
2576 if ((0 == strcmpic(csp->http->gpc, "trace"))
2577 || (0 == strcmpic(csp->http->gpc, "options")))
2579 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2581 if (!strncmpic(p->str, "Max-Forwards:", 13))
2583 unsigned int max_forwards;
2586 * If it's a Max-Forwards value of zero,
2587 * we have to intercept the request.
2589 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2592 * FIXME: We could handle at least TRACE here,
2593 * but that would require a verbatim copy of
2594 * the request which we don't have anymore
2596 log_error(LOG_LEVEL_HEADER,
2597 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2600 /* Get mem for response or fail*/
2601 if (NULL == (rsp = alloc_http_response()))
2603 return cgi_error_memory();
2606 rsp->status = strdup_or_die("501 Not Implemented");
2608 rsp->crunch_reason = UNSUPPORTED;
2610 return(finish_http_response(csp, rsp));
2619 /*********************************************************************
2621 * Function : content_requires_filtering
2623 * Description : Checks whether there are any content filters
2624 * enabled for the current request and if they
2625 * can actually be applied..
2628 * 1 : csp = Current client state (buffers, headers, etc...)
2630 * Returns : TRUE for yes, FALSE otherwise
2632 *********************************************************************/
2633 int content_requires_filtering(struct client_state *csp)
2635 if ((csp->content_type & CT_TABOO)
2636 && !(csp->action->flags & ACTION_FORCE_TEXT_MODE))
2642 * Are we enabling text mode by force?
2644 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2647 * Do we really have to?
2649 if (csp->content_type & CT_TEXT)
2651 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2655 csp->content_type |= CT_TEXT;
2656 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2660 if (!(csp->content_type & CT_DECLARED))
2663 * The server didn't bother to declare a MIME-Type.
2664 * Assume it's text that can be filtered.
2666 * This also regulary happens with 304 responses,
2667 * therefore logging anything here would cause
2670 csp->content_type |= CT_TEXT;
2674 * Choose the applying filter function based on
2675 * the content type and action settings.
2677 if ((csp->content_type & CT_TEXT) &&
2678 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER]) ||
2679 !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER])))
2683 else if ((csp->content_type & CT_GIF) &&
2684 (csp->action->flags & ACTION_DEANIMATE))
2694 /*********************************************************************
2696 * Function : content_filters_enabled
2698 * Description : Checks whether there are any content filters
2699 * enabled for the current request.
2702 * 1 : action = Action spec to check.
2704 * Returns : TRUE for yes, FALSE otherwise
2706 *********************************************************************/
2707 int content_filters_enabled(const struct current_action_spec *action)
2709 return ((action->flags & ACTION_DEANIMATE) ||
2710 !list_is_empty(action->multi[ACTION_MULTI_FILTER]) ||
2711 !list_is_empty(action->multi[ACTION_MULTI_EXTERNAL_FILTER]));
2715 /*********************************************************************
2717 * Function : filters_available
2719 * Description : Checks whether there are any filters available.
2722 * 1 : csp = Current client state (buffers, headers, etc...)
2724 * Returns : TRUE for yes, FALSE otherwise.
2726 *********************************************************************/
2727 int filters_available(const struct client_state *csp)
2730 for (i = 0; i < MAX_AF_FILES; i++)
2732 const struct file_list *fl = csp->rlist[i];
2733 if ((NULL != fl) && (NULL != fl->f))