1 const char filters_rcs[] = "$Id: filters.c,v 1.201 2016/03/17 10:40:53 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 remove_chunked_transfer_coding(char *buffer, size_t *size);
86 static jb_err prepare_for_filtering(struct client_state *csp);
87 static void apply_url_actions(struct current_action_spec *action,
88 struct http_request *http,
89 #ifdef FEATURE_CLIENT_TAGS
90 const struct list *client_tags,
92 struct url_actions *b);
96 /*********************************************************************
98 * Function : sockaddr_storage_to_ip
100 * Description : Access internal structure of sockaddr_storage
103 * 1 : addr = socket address
104 * 2 : ip = IP address as array of octets in network order
105 * (it points into addr)
106 * 3 : len = length of IP address in octets
107 * 4 : port = port number in network order;
109 * Returns : 0 = no errror; -1 otherwise.
111 *********************************************************************/
112 static int sockaddr_storage_to_ip(const struct sockaddr_storage *addr,
113 uint8_t **ip, unsigned int *len,
121 switch (addr->ss_family)
131 &(((struct sockaddr_in *)addr)->sin_addr.s_addr);
135 *port = &((struct sockaddr_in *)addr)->sin_port;
146 *ip = ((struct sockaddr_in6 *)addr)->sin6_addr.s6_addr;
150 *port = &((struct sockaddr_in6 *)addr)->sin6_port;
155 /* Unsupported address family */
163 /*********************************************************************
165 * Function : match_sockaddr
167 * Description : Check whether address matches network (IP address and port)
170 * 1 : network = socket address of subnework
171 * 2 : netmask = network mask as socket address
172 * 3 : address = checked socket address against given network
174 * Returns : 0 = doesn't match; 1 = does match
176 *********************************************************************/
177 static int match_sockaddr(const struct sockaddr_storage *network,
178 const struct sockaddr_storage *netmask,
179 const struct sockaddr_storage *address)
181 uint8_t *network_addr, *netmask_addr, *address_addr;
182 unsigned int addr_len;
183 in_port_t *network_port, *netmask_port, *address_port;
186 if (network->ss_family != netmask->ss_family)
188 /* This should never happen */
189 assert(network->ss_family == netmask->ss_family);
190 log_error(LOG_LEVEL_FATAL, "Network and netmask differ in family.");
193 sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
194 sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
195 sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
197 /* Check for family */
198 if ((network->ss_family == AF_INET) && (address->ss_family == AF_INET6)
199 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr))
201 /* Map AF_INET6 V4MAPPED address into AF_INET */
205 else if ((network->ss_family == AF_INET6) && (address->ss_family == AF_INET)
206 && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr))
208 /* Map AF_INET6 V4MAPPED network into AF_INET */
214 /* XXX: Port check is signaled in netmask */
215 if (*netmask_port && *network_port != *address_port)
220 /* TODO: Optimize by checking by words insted of octets */
221 for (i = 0; (i < addr_len) && netmask_addr[i]; i++)
223 if ((network_addr[i] & netmask_addr[i]) !=
224 (address_addr[i] & netmask_addr[i]))
232 #endif /* def HAVE_RFC2553 */
235 /*********************************************************************
237 * Function : block_acl
239 * Description : Block this request?
240 * Decide yes or no based on ACL file.
243 * 1 : dst = The proxy or gateway address this is going to.
244 * Or NULL to check all possible targets.
245 * 2 : csp = Current client state (buffers, headers, etc...)
246 * Also includes the client IP address.
248 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
250 *********************************************************************/
251 int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
253 struct access_control_list *acl = csp->config->acl;
255 /* if not using an access control list, then permit the connection */
261 /* search the list */
266 match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
268 (csp->ip_addr_long & acl->src->mask) == acl->src->addr
274 /* Just want to check if they have any access */
275 if (acl->action == ACL_PERMIT)
287 * XXX: An undefined acl->dst is full of zeros and should be
288 * considered a wildcard address. sockaddr_storage_to_ip()
289 * fails on such destinations because of unknown sa_familly
290 * (glibc only?). However this test is not portable.
292 * So, we signal the acl->dst is wildcard in wildcard_dst.
295 match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
297 ((dst->addr & acl->dst->mask) == acl->dst->addr)
298 && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
302 if (acl->action == ACL_PERMIT)
320 /*********************************************************************
322 * Function : acl_addr
324 * Description : Called from `load_config' to parse an ACL address.
327 * 1 : aspec = String specifying ACL address.
328 * 2 : aca = struct access_control_addr to fill in.
330 * Returns : 0 => Ok, everything else is an error.
332 *********************************************************************/
333 int acl_addr(const char *aspec, struct access_control_addr *aca)
337 struct addrinfo hints, *result;
339 in_port_t *mask_port;
340 unsigned int addr_len;
343 #endif /* def HAVE_RFC2553 */
345 char *acl_spec = NULL;
348 /* XXX: Depend on ai_family */
356 * Use a temporary acl spec copy so we can log
357 * the unmodified original in case of parse errors.
359 acl_spec = strdup_or_die(aspec);
361 if ((p = strchr(acl_spec, '/')) != NULL)
364 if (privoxy_isdigit(*p) == 0)
369 masklength = atoi(p);
372 if ((masklength < 0) ||
384 if ((*acl_spec == '[') && (NULL != (p = strchr(acl_spec, ']'))))
387 memmove(acl_spec, acl_spec + 1, (size_t)(p - acl_spec));
396 p = strchr(acl_spec, ':');
406 memset(&hints, 0, sizeof(struct addrinfo));
407 hints.ai_family = AF_UNSPEC;
408 hints.ai_socktype = SOCK_STREAM;
410 i = getaddrinfo(acl_spec, p, &hints, &result);
414 log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s",
415 acl_spec, p, gai_strerror(i));
421 /* TODO: Allow multihomed hostnames */
422 memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
423 freeaddrinfo(result);
429 port = strtol(p, &endptr, 10);
431 if (port <= 0 || port > 65535 || *endptr != '\0')
438 aca->port = (unsigned long)port;
440 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
443 if (aca->addr == INADDR_NONE)
445 /* XXX: This will be logged as parse error. */
448 #endif /* def HAVE_RFC2553 */
450 /* build the netmask */
452 /* Clip masklength according to current family. */
453 if ((aca->addr.ss_family == AF_INET) && (masklength > 32))
458 aca->mask.ss_family = aca->addr.ss_family;
459 if (sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port))
466 /* ACL contains a port number, check ports in the future. */
471 * XXX: This could be optimized to operate on whole words instead
472 * of octets (128-bit CPU could do it in one iteration).
475 * Octets after prefix can be omitted because of
476 * previous initialization to zeros.
478 for (i = 0; (i < addr_len) && masklength; i++)
488 * XXX: This assumes MSB of octet is on the left side.
489 * This should be true for all architectures or solved
492 mask_data[i] = (uint8_t)~((1 << (8 - masklength)) - 1);
499 for (i=1; i <= masklength ; i++)
501 aca->mask |= (1U << (32 - i));
504 /* now mask off the host portion of the ip address
505 * (i.e. save on the network portion of the address).
507 aca->addr = aca->addr & aca->mask;
508 #endif /* def HAVE_RFC2553 */
513 #endif /* def FEATURE_ACL */
516 /*********************************************************************
518 * Function : connect_port_is_forbidden
520 * Description : Check to see if CONNECT requests to the destination
521 * port of this request are forbidden. The check is
522 * independend of the actual request method.
525 * 1 : csp = Current client state (buffers, headers, etc...)
527 * Returns : True if yes, false otherwise.
529 *********************************************************************/
530 int connect_port_is_forbidden(const struct client_state *csp)
532 return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
533 !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
538 /*********************************************************************
540 * Function : block_url
542 * Description : Called from `chat'. Check to see if we need to block this.
545 * 1 : csp = Current client state (buffers, headers, etc...)
547 * Returns : NULL => unblocked, else HTTP block response
549 *********************************************************************/
550 struct http_response *block_url(struct client_state *csp)
552 struct http_response *rsp;
553 const char *new_content_type = NULL;
556 * If it's not blocked, don't block it ;-)
558 if ((csp->action->flags & ACTION_BLOCK) == 0)
562 if (csp->action->flags & ACTION_REDIRECT)
564 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
567 * Else, prepare a response
569 if (NULL == (rsp = alloc_http_response()))
571 return cgi_error_memory();
575 * If it's an image-url, send back an image or redirect
576 * as specified by the relevant +image action
578 #ifdef FEATURE_IMAGE_BLOCKING
579 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
583 /* determine HOW images should be blocked */
584 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
586 if (csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
588 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
591 /* and handle accordingly: */
592 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
594 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
595 rsp->body = bindup(image_pattern_data, image_pattern_length);
596 if (rsp->body == NULL)
598 free_http_response(rsp);
599 return cgi_error_memory();
601 rsp->content_length = image_pattern_length;
603 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
605 free_http_response(rsp);
606 return cgi_error_memory();
609 else if (0 == strcmpic(p, "blank"))
611 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
612 rsp->body = bindup(image_blank_data, image_blank_length);
613 if (rsp->body == NULL)
615 free_http_response(rsp);
616 return cgi_error_memory();
618 rsp->content_length = image_blank_length;
620 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
622 free_http_response(rsp);
623 return cgi_error_memory();
628 rsp->status = strdup_or_die("302 Local Redirect from Privoxy");
630 if (enlist_unique_header(rsp->headers, "Location", p))
632 free_http_response(rsp);
633 return cgi_error_memory();
639 #endif /* def FEATURE_IMAGE_BLOCKING */
640 if (csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
643 * Send empty document.
645 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
648 rsp->body = strdup_or_die(" ");
649 rsp->content_length = 1;
651 if (csp->config->feature_flags & RUNTIME_FEATURE_EMPTY_DOC_RETURNS_OK)
654 * Workaround for firefox bug 492459
655 * https://bugzilla.mozilla.org/show_bug.cgi?id=492459
656 * Return a 200 OK status for pages blocked with +handle-as-empty-document
657 * if the "handle-as-empty-doc-returns-ok" runtime config option is set.
659 rsp->status = strdup_or_die("200 Request blocked by Privoxy");
663 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
666 if (new_content_type != 0)
668 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
669 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
671 free_http_response(rsp);
672 return cgi_error_memory();
679 * Else, generate an HTML "blocked" message:
683 struct map * exports;
685 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
687 exports = default_exports(csp, NULL);
690 free_http_response(rsp);
691 return cgi_error_memory();
694 #ifdef FEATURE_FORCE_LOAD
695 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
697 * Export the force conditional block killer if
699 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
700 * - Privoxy is configured to enforce blocks, or
701 * - it's a CONNECT request and enforcing wouldn't work anyway.
703 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
704 || (0 == strcmpic(csp->http->gpc, "connect")))
705 #endif /* ndef FEATURE_FORCE_LOAD */
707 err = map_block_killer(exports, "force-support");
710 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
711 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
712 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
713 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
716 const char *block_reason;
717 if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
719 block_reason = csp->action->string[ACTION_STRING_BLOCK];
723 assert(connect_port_is_forbidden(csp));
724 block_reason = "Forbidden CONNECT port.";
726 err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
731 free_http_response(rsp);
732 return cgi_error_memory();
735 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
738 free_http_response(rsp);
739 return cgi_error_memory();
742 rsp->crunch_reason = BLOCKED;
744 return finish_http_response(csp, rsp);
750 /*********************************************************************
752 * Function : trust_url FIXME: I should be called distrust_url
754 * Description : Calls is_untrusted_url to determine if the URL is trusted
755 * and if not, returns a HTTP 403 response with a reject message.
758 * 1 : csp = Current client state (buffers, headers, etc...)
760 * Returns : NULL => trusted, else http_response.
762 *********************************************************************/
763 struct http_response *trust_url(struct client_state *csp)
765 struct http_response *rsp;
766 struct map * exports;
767 char buf[BUFFER_SIZE];
769 struct pattern_spec **tl;
770 struct pattern_spec *t;
774 * Don't bother to work on trusted URLs
776 if (!is_untrusted_url(csp))
782 * Else, prepare a response:
784 if (NULL == (rsp = alloc_http_response()))
786 return cgi_error_memory();
789 rsp->status = strdup_or_die("403 Request blocked by Privoxy");
790 exports = default_exports(csp, NULL);
793 free_http_response(rsp);
794 return cgi_error_memory();
798 * Export the protocol, host, port, and referrer information
800 err = map(exports, "hostport", 1, csp->http->hostport, 1);
801 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
802 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
804 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
806 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
810 if (!err) err = map(exports, "referrer", 1, "none set", 1);
816 free_http_response(rsp);
817 return cgi_error_memory();
821 * Export the trust list
823 p = strdup_or_die("");
824 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
826 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
827 string_append(&p, buf);
829 err = map(exports, "trusted-referrers", 1, p, 0);
834 free_http_response(rsp);
835 return cgi_error_memory();
839 * Export the trust info, if available
841 if (csp->config->trust_info->first)
843 struct list_entry *l;
845 p = strdup_or_die("");
846 for (l = csp->config->trust_info->first; l ; l = l->next)
848 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
849 string_append(&p, buf);
851 err = map(exports, "trust-info", 1, p, 0);
855 err = map_block_killer(exports, "have-trust-info");
861 free_http_response(rsp);
862 return cgi_error_memory();
866 * Export the force conditional block killer if
868 * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
869 * - Privoxy is configured to enforce blocks, or
870 * - it's a CONNECT request and enforcing wouldn't work anyway.
872 #ifdef FEATURE_FORCE_LOAD
873 if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
874 || (0 == strcmpic(csp->http->gpc, "connect")))
876 err = map_block_killer(exports, "force-support");
880 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
882 #else /* ifndef FEATURE_FORCE_LOAD */
883 err = map_block_killer(exports, "force-support");
884 #endif /* ndef FEATURE_FORCE_LOAD */
889 free_http_response(rsp);
890 return cgi_error_memory();
896 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
899 free_http_response(rsp);
900 return cgi_error_memory();
902 rsp->crunch_reason = UNTRUSTED;
904 return finish_http_response(csp, rsp);
906 #endif /* def FEATURE_TRUST */
909 /*********************************************************************
911 * Function : compile_dynamic_pcrs_job_list
913 * Description : Compiles a dynamic pcrs job list (one with variables
914 * resolved at request time)
917 * 1 : csp = Current client state (buffers, headers, etc...)
918 * 2 : b = The filter list to compile
920 * Returns : NULL in case of errors, otherwise the
923 *********************************************************************/
924 pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
926 struct list_entry *pattern;
927 pcrs_job *job_list = NULL;
928 pcrs_job *dummy = NULL;
929 pcrs_job *lastjob = NULL;
932 const struct pcrs_variable variables[] =
934 {"url", csp->http->url, 1},
935 {"path", csp->http->path, 1},
936 {"host", csp->http->host, 1},
937 {"origin", csp->ip_addr_str, 1},
938 {"listen-address", csp->listen_addr_str, 1},
942 for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
944 assert(pattern->str != NULL);
946 dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
949 log_error(LOG_LEVEL_ERROR,
950 "Compiling dynamic pcrs job '%s' for '%s' failed with error code %d: %s",
951 pattern->str, b->name, error, pcrs_strerror(error));
956 if (error == PCRS_WARN_TRUNCATION)
958 log_error(LOG_LEVEL_ERROR,
959 "At least one of the variables in \'%s\' had to "
960 "be truncated before compilation", pattern->str);
962 if (job_list == NULL)
968 lastjob->next = dummy;
978 /*********************************************************************
980 * Function : rewrite_url
982 * Description : Rewrites a URL with a single pcrs command
983 * and returns the result if it differs from the
984 * original and isn't obviously invalid.
987 * 1 : old_url = URL to rewrite.
988 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
991 * Returns : NULL if the pcrs_command didn't change the url, or
992 * the result of the modification.
994 *********************************************************************/
995 char *rewrite_url(char *old_url, const char *pcrs_command)
997 char *new_url = NULL;
1001 assert(pcrs_command);
1003 new_url = pcrs_execute_single_command(old_url, pcrs_command, &hits);
1007 log_error(LOG_LEVEL_REDIRECTS,
1008 "pcrs command \"%s\" didn't change \"%s\".",
1009 pcrs_command, old_url);
1014 log_error(LOG_LEVEL_REDIRECTS,
1015 "executing pcrs command \"%s\" to rewrite %s failed: %s",
1016 pcrs_command, old_url, pcrs_strerror(hits));
1019 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1021 log_error(LOG_LEVEL_ERROR,
1022 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1023 "but the result doesn't look like a valid URL and will be ignored.",
1024 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1029 log_error(LOG_LEVEL_REDIRECTS,
1030 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1031 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1039 #ifdef FEATURE_FAST_REDIRECTS
1040 /*********************************************************************
1042 * Function : get_last_url
1044 * Description : Search for the last URL inside a string.
1045 * If the string already is a URL, it will
1046 * be the first URL found.
1049 * 1 : subject = the string to check
1050 * 2 : redirect_mode = +fast-redirect{} mode
1052 * Returns : NULL if no URL was found, or
1053 * the last URL found.
1055 *********************************************************************/
1056 char *get_last_url(char *subject, const char *redirect_mode)
1058 char *new_url = NULL;
1062 assert(redirect_mode);
1064 subject = strdup(subject);
1065 if (subject == NULL)
1067 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1071 if (0 == strcmpic(redirect_mode, "check-decoded-url") && strchr(subject, '%'))
1073 char *url_segment = NULL;
1074 char **url_segments;
1075 size_t max_segments;
1078 log_error(LOG_LEVEL_REDIRECTS,
1079 "Checking \"%s\" for encoded redirects.", subject);
1082 * Check each parameter in the URL separately.
1083 * Sectionize the URL at "?" and "&",
1084 * go backwards through the segments, URL-decode them
1085 * and look for a URL in the decoded result.
1086 * Stop the search after the first match.
1088 * XXX: This estimate is guaranteed to be high enough as we
1089 * let ssplit() ignore empty fields, but also a bit wasteful.
1091 max_segments = strlen(subject) / 2;
1092 url_segments = malloc(max_segments * sizeof(char *));
1094 if (NULL == url_segments)
1096 log_error(LOG_LEVEL_ERROR,
1097 "Out of memory while decoding URL: %s", subject);
1102 segments = ssplit(subject, "?&", url_segments, max_segments);
1104 while (segments-- > 0)
1106 char *dtoken = url_decode(url_segments[segments]);
1109 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", url_segments[segments]);
1112 url_segment = strstr(dtoken, "http://");
1113 if (NULL == url_segment)
1115 url_segment = strstr(dtoken, "https://");
1117 if (NULL != url_segment)
1119 url_segment = strdup_or_die(url_segment);
1126 freez(url_segments);
1128 if (url_segment == NULL)
1132 subject = url_segment;
1136 /* Look for a URL inside this one, without decoding anything. */
1137 log_error(LOG_LEVEL_REDIRECTS,
1138 "Checking \"%s\" for unencoded redirects.", subject);
1142 * Find the last URL encoded in the request
1145 while ((tmp = strstr(tmp, "http://")) != NULL)
1149 tmp = (new_url != NULL) ? new_url : subject;
1150 while ((tmp = strstr(tmp, "https://")) != NULL)
1155 if ((new_url != NULL)
1156 && ( (new_url != subject)
1157 || (0 == strncmpic(subject, "http://", 7))
1158 || (0 == strncmpic(subject, "https://", 8))
1162 * Return new URL if we found a redirect
1163 * or if the subject already was a URL.
1165 * The second case makes sure that we can
1166 * chain get_last_url after another redirection check
1167 * (like rewrite_url) without losing earlier redirects.
1169 new_url = strdup(new_url);
1178 #endif /* def FEATURE_FAST_REDIRECTS */
1181 /*********************************************************************
1183 * Function : redirect_url
1185 * Description : Checks if Privoxy should answer the request with
1186 * a HTTP redirect and generates the redirect if
1190 * 1 : csp = Current client state (buffers, headers, etc...)
1192 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1194 *********************************************************************/
1195 struct http_response *redirect_url(struct client_state *csp)
1197 struct http_response *rsp;
1198 #ifdef FEATURE_FAST_REDIRECTS
1200 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1201 * as compile-time option? The user can easily disable
1202 * it in his action file.
1204 char * redirect_mode;
1205 #endif /* def FEATURE_FAST_REDIRECTS */
1206 char *old_url = NULL;
1207 char *new_url = NULL;
1208 char *redirection_string;
1210 if ((csp->action->flags & ACTION_REDIRECT))
1212 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1215 * If the redirection string begins with 's',
1216 * assume it's a pcrs command, otherwise treat it as
1217 * properly formatted URL and use it for the redirection
1220 * According to (the now obsolete) RFC 2616 section 14.30
1221 * the URL has to be absolute and if the user tries:
1222 * +redirect{sadly/this/will/be/parsed/as/pcrs_command.html}
1223 * she would get undefined results anyway.
1225 * RFC 7231 7.1.2 actually allows relative references,
1226 * but those start with a leading slash (RFC 3986 4.2) and
1227 * thus can't be mistaken for pcrs commands either.
1230 if (*redirection_string == 's')
1232 old_url = csp->http->url;
1233 new_url = rewrite_url(old_url, redirection_string);
1237 log_error(LOG_LEVEL_REDIRECTS,
1238 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1239 redirection_string);
1240 new_url = strdup(redirection_string);
1244 #ifdef FEATURE_FAST_REDIRECTS
1245 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1247 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1250 * If it exists, use the previously rewritten URL as input
1251 * otherwise just use the old path.
1253 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1254 new_url = get_last_url(old_url, redirect_mode);
1259 * Disable redirect checkers, so that they
1260 * will be only run more than once if the user
1261 * also enables them through tags.
1263 * From a performance point of view
1264 * it doesn't matter, but the duplicated
1265 * log messages are annoying.
1267 csp->action->flags &= ~ACTION_FAST_REDIRECTS;
1268 #endif /* def FEATURE_FAST_REDIRECTS */
1269 csp->action->flags &= ~ACTION_REDIRECT;
1271 /* Did any redirect action trigger? */
1274 if (url_requires_percent_encoding(new_url))
1277 log_error(LOG_LEVEL_REDIRECTS, "Percent-encoding redirect URL: %N",
1278 strlen(new_url), new_url);
1279 encoded_url = percent_encode_url(new_url);
1281 if (encoded_url == NULL)
1283 return cgi_error_memory();
1285 new_url = encoded_url;
1286 assert(FALSE == url_requires_percent_encoding(new_url));
1289 if (0 == strcmpic(new_url, csp->http->url))
1291 log_error(LOG_LEVEL_ERROR,
1292 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1293 csp->http->url, new_url);
1298 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1300 if (NULL == (rsp = alloc_http_response()))
1303 return cgi_error_memory();
1306 rsp->status = strdup_or_die("302 Local Redirect from Privoxy");
1307 if (enlist_unique_header(rsp->headers, "Location", new_url))
1310 free_http_response(rsp);
1311 return cgi_error_memory();
1313 rsp->crunch_reason = REDIRECTED;
1316 return finish_http_response(csp, rsp);
1320 /* Only reached if no redirect is required */
1326 #ifdef FEATURE_IMAGE_BLOCKING
1327 /*********************************************************************
1329 * Function : is_imageurl
1331 * Description : Given a URL, decide whether it is an image or not,
1332 * using either the info from a previous +image action
1333 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1334 * is MSIE and not on a Mac, tell from the browser's accept
1338 * 1 : csp = Current client state (buffers, headers, etc...)
1340 * Returns : True (nonzero) if URL is an image, false (0)
1343 *********************************************************************/
1344 int is_imageurl(const struct client_state *csp)
1346 #ifdef FEATURE_IMAGE_DETECT_MSIE
1349 tmp = get_header_value(csp->headers, "User-Agent:");
1350 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1352 tmp = get_header_value(csp->headers, "Accept:");
1353 if (tmp && strstr(tmp, "image/gif"))
1355 /* Client will accept HTML. If this seems counterintuitive,
1365 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1367 return ((csp->action->flags & ACTION_IMAGE) != 0);
1370 #endif /* def FEATURE_IMAGE_BLOCKING */
1373 #ifdef FEATURE_TRUST
1374 /*********************************************************************
1376 * Function : is_untrusted_url
1378 * Description : Should we "distrust" this URL (and block it)?
1380 * Yes if it matches a line in the trustfile, or if the
1381 * referrer matches a line starting with "+" in the
1386 * 1 : csp = Current client state (buffers, headers, etc...)
1388 * Returns : 0 => trusted, 1 => untrusted
1390 *********************************************************************/
1391 int is_untrusted_url(const struct client_state *csp)
1393 struct file_list *fl;
1394 struct block_spec *b;
1395 struct pattern_spec **trusted_url;
1396 struct http_request rhttp[1];
1397 const char * referer;
1401 * If we don't have a trustlist, we trust everybody
1403 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1408 memset(rhttp, '\0', sizeof(*rhttp));
1411 * Do we trust the request URL itself?
1413 for (b = b->next; b ; b = b->next)
1415 if (url_match(b->url, csp->http))
1421 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1423 /* no referrer was supplied */
1429 * If not, do we maybe trust its referrer?
1431 err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
1437 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1439 if (url_match(*trusted_url, rhttp))
1441 /* if the URL's referrer is from a trusted referrer, then
1442 * add the target spec to the trustfile as an unblocked
1443 * domain and return 0 (which means it's OK).
1448 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1452 char * new_entry = strdup_or_die("~");
1454 string_append(&new_entry, csp->http->hostport);
1456 path = csp->http->path;
1457 if ( (path[0] == '/')
1459 && ((path_end = strchr(path + 2, '/')) != NULL))
1461 /* since this path points into a user's home space
1462 * be sure to include this spec in the trustfile.
1464 long path_len = path_end - path; /* save offset */
1465 path = strdup(path); /* Copy string */
1468 path_end = path + path_len; /* regenerate ptr to new buffer */
1469 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1471 string_join(&new_entry, path);
1475 * Give a reason for generating this entry.
1477 string_append(&new_entry, " # Trusted referrer was: ");
1478 string_append(&new_entry, referer);
1480 if (new_entry != NULL)
1482 if (-1 == fprintf(fp, "%s\n", new_entry))
1484 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1485 new_entry, csp->config->trustfile);
1491 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1492 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1499 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1500 csp->http->hostport, csp->config->trustfile);
1508 #endif /* def FEATURE_TRUST */
1511 /*********************************************************************
1513 * Function : get_filter
1515 * Description : Get a filter with a given name and type.
1516 * Note that taggers are filters, too.
1519 * 1 : csp = Current client state (buffers, headers, etc...)
1520 * 2 : requested_name = Name of the content filter to get
1521 * 3 : requested_type = Type of the filter to tagger to lookup
1523 * Returns : A pointer to the requested filter
1524 * or NULL if the filter wasn't found
1526 *********************************************************************/
1527 struct re_filterfile_spec *get_filter(const struct client_state *csp,
1528 const char *requested_name,
1529 enum filter_type requested_type)
1532 struct re_filterfile_spec *b;
1533 struct file_list *fl;
1535 for (i = 0; i < MAX_AF_FILES; i++)
1538 if ((NULL == fl) || (NULL == fl->f))
1541 * Either there are no filter files left or this
1542 * filter file just contains no valid filters.
1544 * Continue to be sure we don't miss valid filter
1545 * files that are chained after empty or invalid ones.
1550 for (b = fl->f; b != NULL; b = b->next)
1552 if (b->type != requested_type)
1554 /* The callers isn't interested in this filter type. */
1557 if (strcmp(b->name, requested_name) == 0)
1559 /* The requested filter has been found. Abort search. */
1565 /* No filter with the given name and type exists. */
1571 /*********************************************************************
1573 * Function : pcrs_filter_response
1575 * Description : Execute all text substitutions from all applying
1576 * +filter actions on the text buffer that's been
1577 * accumulated in csp->iob->buf.
1580 * 1 : csp = Current client state (buffers, headers, etc...)
1582 * Returns : a pointer to the (newly allocated) modified buffer.
1583 * or NULL if there were no hits or something went wrong
1585 *********************************************************************/
1586 static char *pcrs_filter_response(struct client_state *csp)
1589 size_t size, prev_size;
1595 struct re_filterfile_spec *b;
1596 struct list_entry *filtername;
1601 if (csp->iob->cur >= csp->iob->eod)
1606 if (filters_available(csp) == FALSE)
1608 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1609 "content filtering enabled, but no content filters available.");
1613 size = (size_t)(csp->iob->eod - csp->iob->cur);
1614 old = csp->iob->cur;
1617 * For all applying +filter actions, look if a filter by that
1618 * name exists and if yes, execute it's pcrs_joblist on the
1621 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1622 filtername != NULL; filtername = filtername->next)
1624 int current_hits = 0; /* Number of hits caused by this filter */
1625 int job_number = 0; /* Which job we're currently executing */
1626 int job_hits = 0; /* How many hits the current job caused */
1629 b = get_filter(csp, filtername->str, FT_CONTENT_FILTER);
1635 joblist = b->joblist;
1637 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1639 if (NULL == joblist)
1641 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1646 /* Apply all jobs from the joblist */
1647 for (job = joblist; NULL != job; job = job->next)
1650 job_hits = pcrs_execute(job, old, size, &new, &size);
1655 * That went well. Continue filtering
1656 * and use the result of this job as
1657 * input for the next one.
1659 current_hits += job_hits;
1660 if (old != csp->iob->cur)
1669 * This job caused an unexpected error. Inform the user
1670 * and skip the rest of the jobs in this filter. We could
1671 * continue with the next job, but usually the jobs
1672 * depend on each other or are similar enough to
1673 * fail for the same reason.
1675 * At the moment our pcrs expects the error codes of pcre 3.4,
1676 * but newer pcre versions can return additional error codes.
1677 * As a result pcrs_strerror()'s error message might be
1678 * "Unknown error ...", therefore we print the numerical value
1681 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1682 * should we use LOG_LEVEL_RE_FILTER instead?
1684 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1685 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1690 if (b->dynamic) pcrs_free_joblist(joblist);
1692 log_error(LOG_LEVEL_RE_FILTER,
1693 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
1694 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1696 hits += current_hits;
1700 * If there were no hits, destroy our copy and let
1701 * chat() use the original in csp->iob
1709 csp->flags |= CSP_FLAG_MODIFIED;
1710 csp->content_length = size;
1711 clear_iob(csp->iob);
1718 #ifdef FEATURE_EXTERNAL_FILTERS
1719 /*********************************************************************
1721 * Function : get_external_filter
1723 * Description : Lookup the code to execute for an external filter.
1724 * Masks the misuse of the re_filterfile_spec.
1727 * 1 : csp = Current client state (buffers, headers, etc...)
1728 * 2 : name = Name of the content filter to get
1730 * Returns : A pointer to the requested code
1731 * or NULL if the filter wasn't found
1733 *********************************************************************/
1734 static const char *get_external_filter(const struct client_state *csp,
1737 struct re_filterfile_spec *external_filter;
1739 external_filter = get_filter(csp, name, FT_EXTERNAL_CONTENT_FILTER);
1740 if (external_filter == NULL)
1742 log_error(LOG_LEVEL_FATAL,
1743 "Didn't find stuff to execute for external filter: %s",
1747 return external_filter->patterns->first->str;
1752 /*********************************************************************
1754 * Function : set_privoxy_variables
1756 * Description : Sets a couple of privoxy-specific environment variables
1759 * 1 : csp = Current client state (buffers, headers, etc...)
1763 *********************************************************************/
1764 static void set_privoxy_variables(const struct client_state *csp)
1771 { "PRIVOXY_URL", csp->http->url },
1772 { "PRIVOXY_PATH", csp->http->path },
1773 { "PRIVOXY_HOST", csp->http->host },
1774 { "PRIVOXY_ORIGIN", csp->ip_addr_str },
1775 { "PRIVOXY_LISTEN_ADDRESS", csp->listen_addr_str },
1778 for (i = 0; i < SZ(env); i++)
1780 if (setenv(env[i].name, env[i].value, 1))
1782 log_error(LOG_LEVEL_ERROR, "Failed to set %s=%s: %E",
1783 env[i].name, env[i].value);
1789 /*********************************************************************
1791 * Function : execute_external_filter
1793 * Description : Pipe content into external filter and return the output
1796 * 1 : csp = Current client state (buffers, headers, etc...)
1797 * 2 : name = Name of the external filter to execute
1798 * 3 : content = The original content to filter
1799 * 4 : size = The size of the content buffer
1801 * Returns : a pointer to the (newly allocated) modified buffer.
1802 * or NULL if there were no hits or something went wrong
1804 *********************************************************************/
1805 static char *execute_external_filter(const struct client_state *csp,
1806 const char *name, char *content, size_t *size)
1809 char file_name[FILENAME_MAX];
1811 char *filter_output;
1815 const char *external_filter;
1817 if (csp->config->temporary_directory == NULL)
1819 log_error(LOG_LEVEL_ERROR,
1820 "No temporary-directory configured. Can't execute filter: %s",
1825 external_filter = get_external_filter(csp, name);
1827 if (sizeof(file_name) < snprintf(file_name, sizeof(file_name),
1828 "%s/privoxy-XXXXXXXX", csp->config->temporary_directory))
1830 log_error(LOG_LEVEL_ERROR, "temporary-directory path too long");
1834 fd = mkstemp(file_name);
1837 log_error(LOG_LEVEL_ERROR, "mkstemp() failed to create %s: %E", file_name);
1841 fp = fdopen(fd, "w");
1844 log_error(LOG_LEVEL_ERROR, "fdopen() failed: %E");
1850 * The size may be zero if a previous filter discarded everything.
1852 * This isn't necessary unintentional, so we just don't try
1853 * to fwrite() nothing and let the user deal with the rest.
1855 if ((*size != 0) && fwrite(content, *size, 1, fp) != 1)
1857 log_error(LOG_LEVEL_ERROR, "fwrite(..., %d, 1, ..) failed: %E", *size);
1864 if (sizeof(cmd) < snprintf(cmd, sizeof(cmd), "%s < %s", external_filter, file_name))
1866 log_error(LOG_LEVEL_ERROR,
1867 "temporary-directory or external filter path too long");
1872 log_error(LOG_LEVEL_RE_FILTER, "Executing '%s': %s", name, cmd);
1875 * The locking is necessary to prevent other threads
1876 * from overwriting the environment variables before
1877 * the popen fork. Afterwards this no longer matters.
1879 privoxy_mutex_lock(&external_filter_mutex);
1880 set_privoxy_variables(csp);
1881 fp = popen(cmd, "r");
1882 privoxy_mutex_unlock(&external_filter_mutex);
1885 log_error(LOG_LEVEL_ERROR, "popen(\"%s\", \"r\") failed: %E", cmd);
1890 /* Allocate at least one byte */
1891 filter_output = malloc_or_die(*size + 1);
1894 while (!feof(fp) && !ferror(fp))
1897 /* Could be bigger ... */
1898 enum { READ_LENGTH = 2048 };
1900 if (new_size + READ_LENGTH >= *size)
1904 /* Could be considered wasteful if the content is 'large'. */
1905 *size += (*size >= READ_LENGTH) ? *size : READ_LENGTH;
1907 p = realloc(filter_output, *size);
1910 log_error(LOG_LEVEL_ERROR, "Out of memory while reading "
1911 "external filter output. Using what we got so far.");
1916 assert(new_size + READ_LENGTH < *size);
1917 len = fread(&filter_output[new_size], 1, READ_LENGTH, fp);
1927 log_error(LOG_LEVEL_ERROR, "Executing %s failed: %E", cmd);
1931 log_error(LOG_LEVEL_RE_FILTER,
1932 "Executing '%s' resulted in return value %d. "
1933 "Read %d of up to %d bytes.", name, (ret >> 8), new_size, *size);
1939 return filter_output;
1942 #endif /* def FEATURE_EXTERNAL_FILTERS */
1945 /*********************************************************************
1947 * Function : gif_deanimate_response
1949 * Description : Deanimate the GIF image that has been accumulated in
1950 * csp->iob->buf, set csp->content_length to the modified
1951 * size and raise the CSP_FLAG_MODIFIED flag.
1954 * 1 : csp = Current client state (buffers, headers, etc...)
1956 * Returns : a pointer to the (newly allocated) modified buffer.
1957 * or NULL in case something went wrong.
1959 *********************************************************************/
1960 static char *gif_deanimate_response(struct client_state *csp)
1962 struct binbuffer *in, *out;
1966 size = (size_t)(csp->iob->eod - csp->iob->cur);
1968 in = zalloc_or_die(sizeof(*in));
1969 out = zalloc_or_die(sizeof(*out));
1971 in->buffer = csp->iob->cur;
1974 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1976 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1983 if ((int)size == out->offset)
1985 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1989 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1991 csp->content_length = out->offset;
1992 csp->flags |= CSP_FLAG_MODIFIED;
2002 /*********************************************************************
2004 * Function : get_filter_function
2006 * Description : Decides which content filter function has
2007 * to be applied (if any). Only considers functions
2008 * for internal filters which are mutually-exclusive.
2011 * 1 : csp = Current client state (buffers, headers, etc...)
2013 * Returns : The content filter function to run, or
2014 * NULL if no content filter is active
2016 *********************************************************************/
2017 static filter_function_ptr get_filter_function(const struct client_state *csp)
2019 filter_function_ptr filter_function = NULL;
2022 * Choose the applying filter function based on
2023 * the content type and action settings.
2025 if ((csp->content_type & CT_TEXT) &&
2026 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2028 filter_function = pcrs_filter_response;
2030 else if ((csp->content_type & CT_GIF) &&
2031 (csp->action->flags & ACTION_DEANIMATE))
2033 filter_function = gif_deanimate_response;
2036 return filter_function;
2040 /*********************************************************************
2042 * Function : remove_chunked_transfer_coding
2044 * Description : In-situ remove the "chunked" transfer coding as defined
2045 * in RFC 7230 4.1 from a buffer. XXX: The implementation
2046 * is neither complete nor compliant (TODO #129).
2049 * 1 : buffer = Pointer to the text buffer
2050 * 2 : size = In: Number of bytes to be processed,
2051 * Out: Number of bytes after de-chunking.
2052 * (undefined in case of errors)
2054 * Returns : JB_ERR_OK for success,
2055 * JB_ERR_PARSE otherwise
2057 *********************************************************************/
2058 static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
2061 unsigned int chunksize = 0;
2062 char *from_p, *to_p;
2063 const char *end_of_buffer = buffer + *size;
2066 from_p = to_p = buffer;
2068 if (sscanf(buffer, "%x", &chunksize) != 1)
2070 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2071 return JB_ERR_PARSE;
2074 while (chunksize > 0U)
2077 * If the chunk-size is valid, we should have at least
2078 * chunk-size bytes of chunk-data and five bytes of
2079 * meta data (chunk-size, CRLF, CRLF) left in the buffer.
2081 if (chunksize + 5 >= *size - newsize)
2083 log_error(LOG_LEVEL_ERROR,
2084 "Chunk size %u exceeds buffered data left. "
2085 "Already digested %u of %u buffered bytes.",
2086 chunksize, (unsigned int)newsize, (unsigned int)*size);
2087 return JB_ERR_PARSE;
2091 * Skip the chunk-size, the optional chunk-ext and the CRLF
2092 * that is supposed to be located directly before the start
2095 if (NULL == (from_p = strstr(from_p, "\r\n")))
2097 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2098 return JB_ERR_PARSE;
2103 * The previous strstr() does not enforce chunk-validity
2104 * and is sattisfied as long a CRLF is left in the buffer.
2106 * Make sure the bytes we consider chunk-data are within
2109 if (from_p + chunksize >= end_of_buffer)
2111 log_error(LOG_LEVEL_ERROR,
2112 "End of chunk is beyond the end of the buffer.");
2113 return JB_ERR_PARSE;
2116 memmove(to_p, from_p, (size_t) chunksize);
2117 newsize += chunksize;
2118 to_p = buffer + newsize;
2119 from_p += chunksize;
2122 * Not merging this check with the previous one allows us
2123 * to keep chunks without trailing CRLF. It's not clear
2124 * if we actually have to care about those, though.
2126 if (from_p + 2 >= end_of_buffer)
2128 log_error(LOG_LEVEL_ERROR, "Not enough room for trailing CRLF.");
2129 return JB_ERR_PARSE;
2132 if (sscanf(from_p, "%x", &chunksize) != 1)
2134 log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
2139 /* XXX: Should get its own loglevel. */
2140 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
2149 /*********************************************************************
2151 * Function : prepare_for_filtering
2153 * Description : If necessary, de-chunks and decompresses
2154 * the content so it can get filterd.
2157 * 1 : csp = Current client state (buffers, headers, etc...)
2159 * Returns : JB_ERR_OK for success,
2160 * JB_ERR_PARSE otherwise
2162 *********************************************************************/
2163 static jb_err prepare_for_filtering(struct client_state *csp)
2165 jb_err err = JB_ERR_OK;
2168 * If the body has a "chunked" transfer-encoding,
2169 * get rid of it, adjusting size and iob->eod
2171 if (csp->flags & CSP_FLAG_CHUNKED)
2173 size_t size = (size_t)(csp->iob->eod - csp->iob->cur);
2175 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
2176 err = remove_chunked_transfer_coding(csp->iob->cur, &size);
2177 if (JB_ERR_OK == err)
2179 csp->iob->eod = csp->iob->cur + size;
2180 csp->flags |= CSP_FLAG_MODIFIED;
2184 return JB_ERR_PARSE;
2190 * If the body has a supported transfer-encoding,
2191 * decompress it, adjusting size and iob->eod.
2193 if (csp->content_type & (CT_GZIP|CT_DEFLATE))
2195 if (0 == csp->iob->eod - csp->iob->cur)
2197 /* Nothing left after de-chunking. */
2201 err = decompress_iob(csp);
2203 if (JB_ERR_OK == err)
2205 csp->flags |= CSP_FLAG_MODIFIED;
2206 csp->content_type &= ~CT_TABOO;
2211 * Unset CT_GZIP and CT_DEFLATE to remember not
2212 * to modify the Content-Encoding header later.
2214 csp->content_type &= ~CT_GZIP;
2215 csp->content_type &= ~CT_DEFLATE;
2224 /*********************************************************************
2226 * Function : execute_content_filters
2228 * Description : Executes a given content filter.
2231 * 1 : csp = Current client state (buffers, headers, etc...)
2233 * Returns : Pointer to the modified buffer, or
2234 * NULL if filtering failed or wasn't necessary.
2236 *********************************************************************/
2237 char *execute_content_filters(struct client_state *csp)
2240 filter_function_ptr content_filter;
2242 assert(content_filters_enabled(csp->action));
2244 if (0 == csp->iob->eod - csp->iob->cur)
2247 * No content (probably status code 301, 302 ...),
2248 * no filtering necessary.
2253 if (JB_ERR_OK != prepare_for_filtering(csp))
2256 * failed to de-chunk or decompress.
2261 if (0 == csp->iob->eod - csp->iob->cur)
2264 * Clown alarm: chunked and/or compressed nothing delivered.
2269 content_filter = get_filter_function(csp);
2270 content = (content_filter != NULL) ? (*content_filter)(csp) : NULL;
2272 #ifdef FEATURE_EXTERNAL_FILTERS
2273 if ((csp->content_type & CT_TEXT) &&
2274 !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]))
2276 struct list_entry *filtername;
2277 size_t size = (size_t)csp->content_length;
2279 if (content == NULL)
2281 content = csp->iob->cur;
2282 size = (size_t)(csp->iob->eod - csp->iob->cur);
2285 for (filtername = csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]->first;
2286 filtername ; filtername = filtername->next)
2288 char *result = execute_external_filter(csp, filtername->str, content, &size);
2291 if (content != csp->iob->cur)
2298 csp->flags |= CSP_FLAG_MODIFIED;
2299 csp->content_length = size;
2301 #endif /* def FEATURE_EXTERNAL_FILTERS */
2308 /*********************************************************************
2310 * Function : get_url_actions
2312 * Description : Gets the actions for this URL.
2315 * 1 : csp = Current client state (buffers, headers, etc...)
2316 * 2 : http = http_request request for blocked URLs
2320 *********************************************************************/
2321 void get_url_actions(struct client_state *csp, struct http_request *http)
2323 struct file_list *fl;
2324 struct url_actions *b;
2327 init_current_action(csp->action);
2329 for (i = 0; i < MAX_AF_FILES; i++)
2331 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2336 #ifdef FEATURE_CLIENT_TAGS
2337 apply_url_actions(csp->action, http, csp->client_tags, b);
2339 apply_url_actions(csp->action, http, b);
2346 /*********************************************************************
2348 * Function : apply_url_actions
2350 * Description : Applies a list of URL actions.
2353 * 1 : action = Destination.
2354 * 2 : http = Current URL
2355 * 3 : client_tags = list of client tags
2356 * 4 : b = list of URL actions to apply
2360 *********************************************************************/
2361 static void apply_url_actions(struct current_action_spec *action,
2362 struct http_request *http,
2363 #ifdef FEATURE_CLIENT_TAGS
2364 const struct list *client_tags,
2366 struct url_actions *b)
2370 /* Should never happen */
2374 for (b = b->next; NULL != b; b = b->next)
2376 if (url_match(b->url, http))
2378 merge_current_action(action, b->action);
2380 #ifdef FEATURE_CLIENT_TAGS
2381 if (client_tag_match(b->url, client_tags))
2383 merge_current_action(action, b->action);
2390 /*********************************************************************
2392 * Function : get_forward_override_settings
2394 * Description : Returns forward settings as specified with the
2395 * forward-override{} action. forward-override accepts
2396 * forward lines similar to the one used in the
2397 * configuration file, but without the URL pattern.
2403 * in the configuration file can be replaced with
2404 * the action section:
2406 * {+forward-override{forward .}}
2410 * 1 : csp = Current client state (buffers, headers, etc...)
2412 * Returns : Pointer to forwarding structure in case of success.
2413 * Invalid syntax is fatal.
2415 *********************************************************************/
2416 static const struct forward_spec *get_forward_override_settings(struct client_state *csp)
2418 const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
2419 char forward_settings[BUFFER_SIZE];
2420 char *http_parent = NULL;
2421 /* variable names were chosen for consistency reasons. */
2422 struct forward_spec *fwd = NULL;
2426 assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
2427 /* Should be enforced by load_one_actions_file() */
2428 assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
2430 /* Create a copy ssplit can modify */
2431 strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
2433 if (NULL != csp->fwd)
2436 * XXX: Currently necessary to prevent memory
2437 * leaks when the show-url-info cgi page is visited.
2439 unload_forward_spec(csp->fwd);
2443 * allocate a new forward node, valid only for
2444 * the lifetime of this request. Save its location
2445 * in csp as well, so sweep() can free it later on.
2447 fwd = csp->fwd = zalloc_or_die(sizeof(*fwd));
2449 vec_count = ssplit(forward_settings, " \t", vec, SZ(vec));
2450 if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
2452 fwd->type = SOCKS_NONE;
2454 /* Parse the parent HTTP proxy host:port */
2455 http_parent = vec[1];
2458 else if ((vec_count == 2) && !strcasecmp(vec[0], "forward-webserver"))
2460 fwd->type = FORWARD_WEBSERVER;
2462 /* Parse the parent HTTP server host:port */
2463 http_parent = vec[1];
2466 else if (vec_count == 3)
2468 char *socks_proxy = NULL;
2470 if (!strcasecmp(vec[0], "forward-socks4"))
2472 fwd->type = SOCKS_4;
2473 socks_proxy = vec[1];
2475 else if (!strcasecmp(vec[0], "forward-socks4a"))
2477 fwd->type = SOCKS_4A;
2478 socks_proxy = vec[1];
2480 else if (!strcasecmp(vec[0], "forward-socks5"))
2482 fwd->type = SOCKS_5;
2483 socks_proxy = vec[1];
2485 else if (!strcasecmp(vec[0], "forward-socks5t"))
2487 fwd->type = SOCKS_5T;
2488 socks_proxy = vec[1];
2491 if (NULL != socks_proxy)
2493 /* Parse the SOCKS proxy host[:port] */
2494 fwd->gateway_port = 1080;
2495 parse_forwarder_address(socks_proxy,
2496 &fwd->gateway_host, &fwd->gateway_port);
2498 http_parent = vec[2];
2502 if (NULL == http_parent)
2504 log_error(LOG_LEVEL_FATAL,
2505 "Invalid forward-override syntax in: %s", forward_override_line);
2506 /* Never get here - LOG_LEVEL_FATAL causes program exit */
2509 /* Parse http forwarding settings */
2510 if (strcmp(http_parent, ".") != 0)
2512 fwd->forward_port = 8000;
2513 parse_forwarder_address(http_parent,
2514 &fwd->forward_host, &fwd->forward_port);
2517 assert (NULL != fwd);
2519 log_error(LOG_LEVEL_CONNECT,
2520 "Overriding forwarding settings based on \'%s\'", forward_override_line);
2526 /*********************************************************************
2528 * Function : forward_url
2530 * Description : Should we forward this to another proxy?
2533 * 1 : csp = Current client state (buffers, headers, etc...)
2534 * 2 : http = http_request request for current URL
2536 * Returns : Pointer to forwarding information.
2538 *********************************************************************/
2539 const struct forward_spec *forward_url(struct client_state *csp,
2540 const struct http_request *http)
2542 static const struct forward_spec fwd_default[1]; /* Zero'ed due to being static. */
2543 struct forward_spec *fwd = csp->config->forward;
2545 if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
2547 return get_forward_override_settings(csp);
2557 if (url_match(fwd->url, http))
2568 /*********************************************************************
2570 * Function : direct_response
2572 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2573 * request and if so, return a HTTP 501 to the client.
2575 * FIXME: I have a stupid name and I should handle the
2576 * requests properly. Still, what we do here is rfc-
2577 * compliant, whereas ignoring or forwarding are not.
2580 * 1 : csp = Current client state (buffers, headers, etc...)
2582 * Returns : http_response if , NULL if nonmatch or handler fail
2584 *********************************************************************/
2585 struct http_response *direct_response(struct client_state *csp)
2587 struct http_response *rsp;
2588 struct list_entry *p;
2590 if ((0 == strcmpic(csp->http->gpc, "trace"))
2591 || (0 == strcmpic(csp->http->gpc, "options")))
2593 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2595 if (!strncmpic(p->str, "Max-Forwards:", 13))
2597 unsigned int max_forwards;
2600 * If it's a Max-Forwards value of zero,
2601 * we have to intercept the request.
2603 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2606 * FIXME: We could handle at least TRACE here,
2607 * but that would require a verbatim copy of
2608 * the request which we don't have anymore
2610 log_error(LOG_LEVEL_HEADER,
2611 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2614 /* Get mem for response or fail*/
2615 if (NULL == (rsp = alloc_http_response()))
2617 return cgi_error_memory();
2620 rsp->status = strdup_or_die("501 Not Implemented");
2622 rsp->crunch_reason = UNSUPPORTED;
2624 return(finish_http_response(csp, rsp));
2633 /*********************************************************************
2635 * Function : content_requires_filtering
2637 * Description : Checks whether there are any content filters
2638 * enabled for the current request and if they
2639 * can actually be applied..
2642 * 1 : csp = Current client state (buffers, headers, etc...)
2644 * Returns : TRUE for yes, FALSE otherwise
2646 *********************************************************************/
2647 int content_requires_filtering(struct client_state *csp)
2649 if ((csp->content_type & CT_TABOO)
2650 && !(csp->action->flags & ACTION_FORCE_TEXT_MODE))
2656 * Are we enabling text mode by force?
2658 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
2661 * Do we really have to?
2663 if (csp->content_type & CT_TEXT)
2665 log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
2669 csp->content_type |= CT_TEXT;
2670 log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
2674 if (!(csp->content_type & CT_DECLARED))
2677 * The server didn't bother to declare a MIME-Type.
2678 * Assume it's text that can be filtered.
2680 * This also regulary happens with 304 responses,
2681 * therefore logging anything here would cause
2684 csp->content_type |= CT_TEXT;
2688 * Choose the applying filter function based on
2689 * the content type and action settings.
2691 if ((csp->content_type & CT_TEXT) &&
2692 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER]) ||
2693 !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER])))
2697 else if ((csp->content_type & CT_GIF) &&
2698 (csp->action->flags & ACTION_DEANIMATE))
2708 /*********************************************************************
2710 * Function : content_filters_enabled
2712 * Description : Checks whether there are any content filters
2713 * enabled for the current request.
2716 * 1 : action = Action spec to check.
2718 * Returns : TRUE for yes, FALSE otherwise
2720 *********************************************************************/
2721 int content_filters_enabled(const struct current_action_spec *action)
2723 return ((action->flags & ACTION_DEANIMATE) ||
2724 !list_is_empty(action->multi[ACTION_MULTI_FILTER]) ||
2725 !list_is_empty(action->multi[ACTION_MULTI_EXTERNAL_FILTER]));
2729 /*********************************************************************
2731 * Function : filters_available
2733 * Description : Checks whether there are any filters available.
2736 * 1 : csp = Current client state (buffers, headers, etc...)
2738 * Returns : TRUE for yes, FALSE otherwise.
2740 *********************************************************************/
2741 int filters_available(const struct client_state *csp)
2744 for (i = 0; i < MAX_AF_FILES; i++)
2746 const struct file_list *fl = csp->rlist[i];
2747 if ((NULL != fl) && (NULL != fl->f))