1 const char filters_rcs[] = "$Id: filters.c,v 1.47 2002/03/13 00:30:52 jongfoster Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/filters.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
7 * Functions declared include:
8 * `acl_addr', `add_stats', `block_acl', `block_imageurl',
9 * `block_url', `url_actions', `domain_split',
10 * `filter_popups', `forward_url', 'redirect_url',
11 * `ij_untrusted_url', `intercept_url', `pcrs_filter_respose',
12 * 'ijb_send_banner', and `trust_url'
14 * Copyright : Written by and Copyright (C) 2001 the SourceForge
15 * IJBSWA team. http://ijbswa.sourceforge.net
17 * Based on the Internet Junkbuster originally written
18 * by and Copyright (C) 1997 Anonymous Coders and
19 * Junkbusters Corporation. http://www.junkbusters.com
21 * This program is free software; you can redistribute it
22 * and/or modify it under the terms of the GNU General
23 * Public License as published by the Free Software
24 * Foundation; either version 2 of the License, or (at
25 * your option) any later version.
27 * This program is distributed in the hope that it will
28 * be useful, but WITHOUT ANY WARRANTY; without even the
29 * implied warranty of MERCHANTABILITY or FITNESS FOR A
30 * PARTICULAR PURPOSE. See the GNU General Public
31 * License for more details.
33 * The GNU General Public License should be included with
34 * this file. If not, you can view it at
35 * http://www.gnu.org/copyleft/gpl.html
36 * or write to the Free Software Foundation, Inc., 59
37 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
41 * Revision 1.47 2002/03/13 00:30:52 jongfoster
43 * Added option of always sending redirect for imageblock,
44 * currently disabled with #if 0.
46 * Revision 1.46 2002/03/12 01:42:49 oes
47 * Introduced modular filters
49 * Revision 1.45 2002/03/08 16:47:50 oes
50 * Added choice beween GIF and PNG built-in images
52 * Revision 1.44 2002/03/07 03:49:31 oes
53 * - Fixed compiler warnings etc
54 * - Changed built-in images from GIF to PNG
55 * (with regard to Unisys patent issue)
56 * - Added a 4x4 pattern PNG which is less intrusive
57 * than the logo but also clearly marks the deleted banners
59 * Revision 1.43 2002/01/22 23:51:59 jongfoster
60 * Replacing strsav() with the safer string_append().
62 * Adding missing html_encode() to error message generators. Where encoded
63 * and unencoded versions of a string were provided, removing the unencoded
66 * Revision 1.42 2002/01/17 21:00:32 jongfoster
67 * Moving all our URL and URL pattern parsing code to urlmatch.c.
69 * Using a single, simple url_match(pattern,url) function - rather than
70 * the 3-line match routine which was repeated all over the place.
72 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
74 * Using parse_http_url() to parse URLs without faking a HTTP
75 * request line for parse_http_request().
77 * Revision 1.41 2001/11/13 00:14:07 jongfoster
78 * Fixing stupid bug now I've figured out what || means.
79 * (It always returns 0 or 1, not one of it's paramaters.)
81 * Revision 1.40 2001/10/26 17:37:55 oes
82 * - Re-enabled Netscape 200/404 bug workaround in block_url():
83 * - Removed OS/2 special case
84 * - Made block_url() independant from sed() having been run
85 * - Made trust_url independant from sed() having been run
86 * - Made is_imageurl independant from sed() having been run.
87 * It now checks User-Agent: and Accept: by itself.
90 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
91 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
92 * threads to call select() simultaneously. So, it's time to do a real, live,
93 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
94 * (native). Both versions will work, but using __OS2__ offers multi-threading.
96 * Revision 1.38 2001/10/23 21:32:33 jongfoster
97 * Adding error-checking to selected functions
99 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
100 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
101 * filters.c. Added a FIXME in front of the offending code. I'll gladly
102 * put in a better/more robust fix for all parties if one is presented...
103 * It seems that just returning 200 instead of 404 would pretty much fix
104 * it for everyone, but I don't know all the history of the problem.
106 * Revision 1.36 2001/10/10 16:44:16 oes
107 * Added match_portlist function
109 * Revision 1.35 2001/10/07 15:41:23 oes
110 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
112 * New function remove_chunked_transfer_coding that strips chunked
113 * transfer coding to plain and is called by pcrs_filter_response
114 * and gif_deanimate_response if neccessary
116 * Improved handling of zero-change re_filter runs
118 * pcrs_filter_response and gif_deanimate_response now remove
119 * chunked transfer codeing before processing the body.
121 * Revision 1.34 2001/09/20 15:49:36 steudten
123 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
124 * See cgi.c fill_template().
126 * Revision 1.33 2001/09/16 17:05:14 jongfoster
127 * Removing unused #include showarg.h
129 * Revision 1.32 2001/09/16 13:21:27 jongfoster
130 * Changes to use new list functions.
132 * Revision 1.31 2001/09/16 11:38:02 jongfoster
133 * Splitting fill_template() into 2 functions:
134 * template_load() loads the file
135 * template_fill() performs the PCRS regexps.
136 * This is because the CGI edit interface has a "table row"
137 * template which is used many times in the page - this
138 * change means it's only loaded from disk once.
140 * Revision 1.30 2001/09/16 11:00:10 jongfoster
141 * New function alloc_http_response, for symmetry with free_http_response
143 * Revision 1.29 2001/09/13 23:32:40 jongfoster
144 * Moving image data to cgi.c rather than cgi.h
145 * Fixing a GPF under Win32 (and any other OS that protects global
146 * constants from being written to).
148 * Revision 1.28 2001/09/10 10:18:51 oes
149 * Silenced compiler warnings
151 * Revision 1.27 2001/08/05 16:06:20 jongfoster
152 * Modifiying "struct map" so that there are now separate header and
153 * "map_entry" structures. This means that functions which modify a
154 * map no longer need to return a pointer to the modified map.
155 * Also, it no longer reverses the order of the entries (which may be
156 * important with some advanced template substitutions).
158 * Revision 1.26 2001/07/30 22:08:36 jongfoster
159 * Tidying up #defines:
160 * - All feature #defines are now of the form FEATURE_xxx
161 * - Permanently turned off WIN_GUI_EDIT
162 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
164 * Revision 1.25 2001/07/26 10:09:46 oes
165 * Made browser detection a little less naive
167 * Revision 1.24 2001/07/25 17:22:51 oes
168 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
170 * Revision 1.23 2001/07/23 13:40:12 oes
171 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
173 * Revision 1.22 2001/07/18 12:29:34 oes
174 * - Made gif_deanimate_response respect
175 * csp->action->string[ACTION_STRING_DEANIMATE]
176 * - Logging cosmetics
178 * Revision 1.21 2001/07/13 13:59:53 oes
179 * - Introduced gif_deanimate_response which shares the
180 * generic content modification interface of pcrs_filter_response
181 * and acts as a wrapper to deanimate.c:gif_deanimate()
182 * - Renamed re_process_buffer to pcrs_filter_response
183 * - pcrs_filter_response now returns NULL on failiure
184 * - Removed all #ifdef PCRS
186 * Revision 1.20 2001/07/01 17:01:04 oes
187 * Added comments and missing return statement in is_untrusted_url()
189 * Revision 1.19 2001/06/29 21:45:41 oes
190 * Indentation, CRLF->LF, Tab-> Space
192 * Revision 1.18 2001/06/29 13:27:38 oes
193 * - Cleaned up, renamed and reorderd functions
194 * and improved comments
197 * - Ported to CGI platform. Now delivers
198 * http_response or NULL
199 * - Unified HTML and GIF generation (moved image detection
200 * and GIF generation here from jcc.c:chat())
201 * - Fixed HTTP status to:
202 * - 403 (Forbidden) for the "blocked" HTML message
203 * - 200 (OK) for GIF answers
204 * - 302 (Redirect) for redirect to GIF
207 * - Ported to CGI platform. Now delivers
208 * http_response or NULL
209 * - Separated detection of untrusted URL into
210 * (bool)is_untrusted_url
211 * - Added enforcement of untrusted requests
213 * - Moved redirect_url() from cgi.c to here
214 * and ported it to the CGI platform
216 * - Removed logentry from cancelled commit
218 * Revision 1.17 2001/06/09 10:55:28 jongfoster
219 * Changing BUFSIZ ==> BUFFER_SIZE
221 * Revision 1.16 2001/06/07 23:10:26 jongfoster
222 * Allowing unanchored domain patterns to back off and retry
223 * if they partially match. Optimized right-anchored patterns.
224 * Moving ACL and forward files into config file.
225 * Replacing struct gateway with struct forward_spec
227 * Revision 1.15 2001/06/03 19:12:00 oes
228 * extracted-CGI relevant stuff
230 * Revision 1.14 2001/06/01 10:30:55 oes
231 * Added optional left-anchoring to domaincmp
233 * Revision 1.13 2001/05/31 21:21:30 jongfoster
234 * Permissionsfile / actions file changes:
235 * - Changed "permission" to "action" throughout
236 * - changes to file format to allow string parameters
237 * - Moved helper functions to actions.c
239 * Revision 1.12 2001/05/31 17:35:20 oes
241 * - Enhanced domain part globbing with infix and prefix asterisk
242 * matching and optional unanchored operation
244 * Revision 1.11 2001/05/29 11:53:23 oes
245 * "See why" link added to "blocked" page
247 * Revision 1.10 2001/05/29 09:50:24 jongfoster
248 * Unified blocklist/imagelist/permissionslist.
249 * File format is still under discussion, but the internal changes
252 * Also modified interceptor behaviour:
253 * - We now intercept all URLs beginning with one of the following
254 * prefixes (and *only* these prefixes):
256 * * http://ijbswa.sf.net/config/
257 * * http://ijbswa.sourceforge.net/config/
258 * - New interceptors "home page" - go to http://i.j.b/ to see it.
259 * - Internal changes so that intercepted and fast redirect pages
260 * are not replaced with an image.
261 * - Interceptors now have the option to send a binary page direct
262 * to the client. (i.e. ijb-send-banner uses this)
263 * - Implemented show-url-info interceptor. (Which is why I needed
264 * the above interceptors changes - a typical URL is
265 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
266 * The previous mechanism would not have intercepted that, and
267 * if it had been intercepted then it then it would have replaced
270 * Revision 1.9 2001/05/27 22:17:04 oes
272 * - re_process_buffer no longer writes the modified buffer
273 * to the client, which was very ugly. It now returns the
274 * buffer, which it is then written by chat.
276 * - content_length now adjusts the Content-Length: header
277 * for modified documents rather than crunch()ing it.
278 * (Length info in csp->content_length, which is 0 for
279 * unmodified documents)
281 * - For this to work, sed() is called twice when filtering.
283 * Revision 1.8 2001/05/26 17:13:28 jongfoster
284 * Filled in a function comment.
286 * Revision 1.7 2001/05/26 15:26:15 jongfoster
287 * ACL feature now provides more security by immediately dropping
288 * connections from untrusted hosts.
290 * Revision 1.6 2001/05/26 00:28:36 jongfoster
291 * Automatic reloading of config file.
292 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
293 * Most of the global variables have been moved to a new
294 * struct configuration_spec, accessed through csp->config->globalname
295 * Most of the globals remaining are used by the Win32 GUI.
297 * Revision 1.5 2001/05/25 22:34:30 jongfoster
300 * Revision 1.4 2001/05/22 18:46:04 oes
302 * - Enabled filtering banners by size rather than URL
303 * by adding patterns that replace all standard banner
304 * sizes with the "Junkbuster" gif to the re_filterfile
306 * - Enabled filtering WebBugs by providing a pattern
307 * which kills all 1x1 images
309 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
310 * which is selected by the (nonstandard and therefore
311 * capital) letter 'U' in the option string.
312 * It causes the quantifiers to be ungreedy by default.
313 * Appending a ? turns back to greedy (!).
315 * - Added a new interceptor ijb-send-banner, which
316 * sends back the "Junkbuster" gif. Without imagelist or
317 * MSIE detection support, or if tinygif = 1, or the
318 * URL isn't recognized as an imageurl, a lame HTML
319 * explanation is sent instead.
321 * - Added new feature, which permits blocking remote
322 * script redirects and firing back a local redirect
324 * The feature is conditionally compiled, i.e. it
325 * can be disabled with --disable-fast-redirects,
326 * plus it must be activated by a "fast-redirects"
327 * line in the config file, has its own log level
328 * and of course wants to be displayed by show-proxy-args
329 * Note: Boy, all the #ifdefs in 1001 locations and
330 * all the fumbling with configure.in and acconfig.h
331 * were *way* more work than the feature itself :-(
333 * - Because a generic redirect template was needed for
334 * this, tinygif = 3 now uses the same.
336 * - Moved GIFs, and other static HTTP response templates
341 * - Removed some >400 CRs again (Jon, you really worked
344 * Revision 1.3 2001/05/20 16:44:47 jongfoster
345 * Removing last hardcoded JunkBusters.com URLs.
347 * Revision 1.2 2001/05/20 01:21:20 jongfoster
348 * Version 2.9.4 checkin.
349 * - Merged popupfile and cookiefile, and added control over PCRS
350 * filtering, in new "permissionsfile".
351 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
352 * file error you now get a message box (in the Win32 GUI) rather
353 * than the program exiting with no explanation.
354 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
356 * - Removed tabs from "config"
357 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
358 * - Bumped up version number.
360 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
361 * Initial import of version 2.9.3 source tree
364 *********************************************************************/
370 #include <sys/types.h>
379 #endif /* ndef __OS2__ */
380 #include <netinet/in.h>
382 #include <winsock2.h>
383 #endif /* ndef _WIN32 */
387 #endif /* def __OS2__ */
395 #include "jbsockets.h"
396 #include "miscutil.h"
400 #include "deanimate.h"
401 #include "urlmatch.h"
407 const char filters_h_rcs[] = FILTERS_H_VERSION;
409 /* Fix a problem with Solaris. There should be no effect on other
411 * Solaris's isspace() is a macro which uses it's argument directly
412 * as an array index. Therefore we need to make sure that high-bit
413 * characters generate +ve values, and ideally we also want to make
414 * the argument match the declared parameter type of "int".
416 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
420 /*********************************************************************
422 * Function : block_acl
424 * Description : Block this request?
425 * Decide yes or no based on ACL file.
428 * 1 : dst = The proxy or gateway address this is going to.
429 * Or NULL to check all possible targets.
430 * 2 : csp = Current client state (buffers, headers, etc...)
431 * Also includes the client IP address.
433 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
435 *********************************************************************/
436 int block_acl(struct access_control_addr *dst, struct client_state *csp)
438 struct access_control_list *acl = csp->config->acl;
440 /* if not using an access control list, then permit the connection */
446 /* search the list */
449 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
453 /* Just want to check if they have any access */
454 if (acl->action == ACL_PERMIT)
459 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
460 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
462 if (acl->action == ACL_PERMIT)
480 /*********************************************************************
482 * Function : acl_addr
484 * Description : Called from `load_config' to parse an ACL address.
487 * 1 : aspec = String specifying ACL address.
488 * 2 : aca = struct access_control_addr to fill in.
490 * Returns : 0 => Ok, everything else is an error.
492 *********************************************************************/
493 int acl_addr(char *aspec, struct access_control_addr *aca)
495 int i, masklength, port;
501 if ((p = strchr(aspec, '/')) != NULL)
505 if (ijb_isdigit(*p) == 0)
509 masklength = atoi(p);
512 if ((masklength < 0) || (masklength > 32))
517 if ((p = strchr(aspec, ':')) != NULL)
521 if (ijb_isdigit(*p) == 0)
530 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
532 if (aca->addr == INADDR_NONE)
537 /* build the netmask */
539 for (i=1; i <= masklength ; i++)
541 aca->mask |= (1 << (32 - i));
544 /* now mask off the host portion of the ip address
545 * (i.e. save on the network portion of the address).
547 aca->addr = aca->addr & aca->mask;
552 #endif /* def FEATURE_ACL */
555 /*********************************************************************
557 * Function : match_portlist
559 * Description : Check if a given number is covered by a comma
560 * separated list of numbers and ranges (a,b-c,d,..)
563 * 1 : portlist = String with list
564 * 2 : port = port to check
566 * Returns : 0 => no match
569 *********************************************************************/
570 int match_portlist(const char *portlist, int port)
572 char *min, *max, *next, *portlist_copy;
574 min = next = portlist_copy = strdup(portlist);
577 * Zero-terminate first item and remember offset for next
579 if (NULL != (next = strchr(portlist_copy, (int) ',')))
585 * Loop through all items, checking for match
589 if (NULL == (max = strchr(min, (int) '-')))
592 * No dash, check for equality
594 if (port == atoi(min))
603 * This is a range, so check if between min and max,
604 * or, if max was omitted, between min and 65K
607 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
621 * Zero-terminate next item and remember offset for n+1
623 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
635 /*********************************************************************
637 * Function : block_url
639 * Description : Called from `chat'. Check to see if we need to block this.
642 * 1 : csp = Current client state (buffers, headers, etc...)
644 * Returns : NULL => unblocked, else HTTP block response
646 *********************************************************************/
647 struct http_response *block_url(struct client_state *csp)
649 #ifdef FEATURE_IMAGE_BLOCKING
651 #endif /* def FEATURE_IMAGE_BLOCKING */
652 struct http_response *rsp;
655 * If it's not blocked, don't block it ;-)
657 if ((csp->action->flags & ACTION_BLOCK) == 0)
663 * Else, prepare a response
665 if (NULL == (rsp = alloc_http_response()))
667 return cgi_error_memory();
671 * If it's an image-url, send back an image or redirect
672 * as specified by the relevant +image action
674 #ifdef FEATURE_IMAGE_BLOCKING
675 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
678 /* determine HOW images should be blocked */
679 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
681 #if 1 /* Two alternative strategies, use this one for now: */
683 /* and handle accordingly: */
684 if ((p == NULL) || (0 == strcmpic(p, "logo")))
686 rsp->body = bindup(image_logo_data, image_logo_length);
687 if (rsp->body == NULL)
689 free_http_response(rsp);
690 return cgi_error_memory();
692 rsp->content_length = image_logo_length;
694 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
696 free_http_response(rsp);
697 return cgi_error_memory();
701 else if (0 == strcmpic(p, "blank"))
703 rsp->body = bindup(image_blank_data, image_blank_length);
704 if (rsp->body == NULL)
706 free_http_response(rsp);
707 return cgi_error_memory();
709 rsp->content_length = image_blank_length;
711 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
713 free_http_response(rsp);
714 return cgi_error_memory();
718 else if (0 == strcmpic(p, "pattern"))
720 rsp->body = bindup(image_pattern_data, image_pattern_length);
721 if (rsp->body == NULL)
723 free_http_response(rsp);
724 return cgi_error_memory();
726 rsp->content_length = image_pattern_length;
728 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
730 free_http_response(rsp);
731 return cgi_error_memory();
737 rsp->status = strdup("302 Local Redirect from Junkbuster");
738 if (rsp->status == NULL)
740 free_http_response(rsp);
741 return cgi_error_memory();
744 if (enlist_unique_header(rsp->headers, "Location", p))
746 free_http_response(rsp);
747 return cgi_error_memory();
751 #else /* Following code is disabled for now */
753 /* and handle accordingly: */
754 if ((p == NULL) || (0 == strcmpic(p, "logo")))
756 p = CGI_PREFIX "send-banner?type=logo";
758 else if (0 == strcmpic(p, "blank"))
760 p = CGI_PREFIX "send-banner?type=blank";
762 else if (0 == strcmpic(p, "pattern"))
764 p = CGI_PREFIX "send-banner?type=pattern";
766 rsp->status = strdup("302 Local Redirect from Junkbuster");
767 if (rsp->status == NULL)
769 free_http_response(rsp);
770 return cgi_error_memory();
773 if (enlist_unique_header(rsp->headers, "Location", p))
775 free_http_response(rsp);
776 return cgi_error_memory();
778 #endif /* Preceeding code is disabled for now */
781 #endif /* def FEATURE_IMAGE_BLOCKING */
784 * Else, generate an HTML "blocked" message:
788 struct map * exports;
791 * Workaround for stupid Netscape bug which prevents
792 * pages from being displayed if loading a referenced
793 * JavaScript or style sheet fails. So make it appear
794 * as if it succeeded.
796 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
797 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
798 && !strstr(p, "Gecko") /* save Mozilla, */
799 && !strstr(p, "compatible") /* MSIE */
800 && !strstr(p, "Opera")) /* and Opera. */
802 rsp->status = strdup("200 Request for blocked URL");
806 rsp->status = strdup("404 Request for blocked URL");
809 if (rsp->status == NULL)
811 free_http_response(rsp);
812 return cgi_error_memory();
815 exports = default_exports(csp, NULL);
818 free_http_response(rsp);
819 return cgi_error_memory();
822 #ifdef FEATURE_FORCE_LOAD
823 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
824 #else /* ifndef FEATURE_FORCE_LOAD */
825 err = map_block_killer(exports, "force-support");
826 #endif /* ndef FEATURE_FORCE_LOAD */
828 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
829 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
834 free_http_response(rsp);
835 return cgi_error_memory();
838 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
841 free_http_response(rsp);
842 return cgi_error_memory();
846 return finish_http_response(rsp);
852 /*********************************************************************
854 * Function : trust_url FIXME: I should be called distrust_url
856 * Description : Calls is_untrusted_url to determine if the URL is trusted
857 * and if not, returns a HTTP 304 response with a reject message.
860 * 1 : csp = Current client state (buffers, headers, etc...)
862 * Returns : NULL => trusted, else http_response.
864 *********************************************************************/
865 struct http_response *trust_url(struct client_state *csp)
867 struct http_response *rsp;
868 struct map * exports;
869 char buf[BUFFER_SIZE];
871 struct url_spec **tl;
876 * Don't bother to work on trusted URLs
878 if (!is_untrusted_url(csp))
884 * Else, prepare a response:
886 if (NULL == (rsp = alloc_http_response()))
888 return cgi_error_memory();
891 exports = default_exports(csp, NULL);
894 free_http_response(rsp);
895 return cgi_error_memory();
899 * Export the host, port, and referrer information
901 err = map(exports, "hostport", 1, csp->http->hostport, 1);
902 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
904 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
906 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
910 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
916 free_http_response(rsp);
917 return cgi_error_memory();
921 * Export the trust list
924 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
926 sprintf(buf, "<li>%s</li>\n", t->spec);
927 string_append(&p, buf);
929 err = map(exports, "trusted-referrers", 1, p, 0);
934 free_http_response(rsp);
935 return cgi_error_memory();
939 * Export the trust info, if available
941 if (csp->config->trust_info->first)
943 struct list_entry *l;
946 for (l = csp->config->trust_info->first; l ; l = l->next)
948 sprintf(buf, "<li> <a href=%s>%s</a><br>\n",l->str, l->str);
949 string_append(&p, buf);
951 err = map(exports, "trust-info", 1, p, 0);
955 err = map_block_killer(exports, "have-trust-info");
961 free_http_response(rsp);
962 return cgi_error_memory();
966 * Export the force prefix or the force conditional block killer
968 #ifdef FEATURE_FORCE_LOAD
969 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
970 #else /* ifndef FEATURE_FORCE_LOAD */
971 err = map_block_killer(exports, "force-support");
972 #endif /* ndef FEATURE_FORCE_LOAD */
977 free_http_response(rsp);
978 return cgi_error_memory();
984 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
987 free_http_response(rsp);
988 return cgi_error_memory();
991 return finish_http_response(rsp);
993 #endif /* def FEATURE_TRUST */
996 #ifdef FEATURE_FAST_REDIRECTS
997 /*********************************************************************
999 * Function : redirect_url
1001 * Description : Checks for redirection URLs and returns a HTTP redirect
1002 * to the destination URL, if necessary
1005 * 1 : csp = Current client state (buffers, headers, etc...)
1007 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1009 *********************************************************************/
1010 struct http_response *redirect_url(struct client_state *csp)
1013 struct http_response *rsp;
1015 p = q = csp->http->path;
1016 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1019 * find the last URL encoded in the request
1021 while ((p = strstr(p, "http://")) != NULL)
1027 * if there was any, generate and return a HTTP redirect
1029 if (q != csp->http->path)
1031 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1033 if (NULL == (rsp = alloc_http_response()))
1035 return cgi_error_memory();
1038 if ( enlist_unique_header(rsp->headers, "Location", q)
1039 || (NULL == (rsp->status = strdup("302 Local Redirect from Junkbuster"))) )
1041 free_http_response(rsp);
1042 return cgi_error_memory();
1045 return finish_http_response(rsp);
1053 #endif /* def FEATURE_FAST_REDIRECTS */
1056 #ifdef FEATURE_IMAGE_BLOCKING
1057 /*********************************************************************
1059 * Function : is_imageurl
1061 * Description : Given a URL, decide whether it is an image or not,
1062 * using either the info from a previous +image action
1063 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, the info from
1064 * the browser's accept header.
1067 * 1 : csp = Current client state (buffers, headers, etc...)
1069 * Returns : True (nonzero) if URL is an image, false (0)
1072 *********************************************************************/
1073 int is_imageurl(struct client_state *csp)
1075 #ifdef FEATURE_IMAGE_DETECT_MSIE
1078 tmp = get_header_value(csp->headers, "User-Agent:");
1079 if (tmp && strstr(tmp, "MSIE"))
1081 tmp = get_header_value(csp->headers, "Accept:");
1082 if (tmp && strstr(tmp, "image/gif"))
1084 /* Client will accept HTML. If this seems counterintuitive,
1094 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1096 return ((csp->action->flags & ACTION_IMAGE) != 0);
1099 #endif /* def FEATURE_IMAGE_BLOCKING */
1102 #ifdef FEATURE_COOKIE_JAR
1103 /*********************************************************************
1105 * Function : is_untrusted_url
1107 * Description : Should we "distrust" this URL (and block it)?
1109 * Yes if it matches a line in the trustfile, or if the
1110 * referrer matches a line starting with "+" in the
1115 * 1 : csp = Current client state (buffers, headers, etc...)
1117 * Returns : 0 => trusted, 1 => untrusted
1119 *********************************************************************/
1120 int is_untrusted_url(struct client_state *csp)
1122 struct file_list *fl;
1123 struct block_spec *b;
1124 struct url_spec **trusted_url;
1125 struct http_request rhttp[1];
1126 const char * referer;
1130 * If we don't have a trustlist, we trust everybody
1132 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1137 memset(rhttp, '\0', sizeof(*rhttp));
1140 * Do we trust the request URL itself?
1142 for (b = b->next; b ; b = b->next)
1144 if (url_match(b->url, csp->http))
1150 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1152 /* no referrer was supplied */
1157 * If not, do we maybe trust its referrer?
1162 * Parse the URL from the referrer
1165 err = parse_http_url(referer, rhttp, csp);
1171 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1173 if (url_match(*trusted_url, rhttp))
1175 /* if the URL's referrer is from a trusted referrer, then
1176 * add the target spec to the trustfile as an unblocked
1177 * domain and return NULL (which means it's OK).
1182 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1186 char * new_entry = strdup("~");
1188 string_append(&new_entry, csp->http->hostport);
1190 path = csp->http->path;
1191 if ( (path[0] == '/')
1193 && ((path_end = strchr(path + 2, '/')) != NULL))
1195 /* since this path points into a user's home space
1196 * be sure to include this spec in the trustfile.
1198 int path_len = path_end - path; /* save offset */
1199 path = strdup(path); /* Copy string */
1202 path_end = path + path_len; /* regenerate ptr to new buffer */
1203 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1205 string_join(&new_entry, path);
1208 if (new_entry != NULL)
1210 fprintf(fp, "%s\n", new_entry);
1215 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1216 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1226 #endif /* def FEATURE_COOKIE_JAR */
1229 /*********************************************************************
1231 * Function : pcrs_filter_response
1233 * Description : Ecexute all text substitutions from all applying
1234 * +filter actions on the text buffer that's been accumulated
1235 * in csp->iob->buf. If this changes the contents, set
1236 * csp->content_length to the modified size and raise the
1237 * CSP_FLAG_MODIFIED flag.
1240 * 1 : csp = Current client state (buffers, headers, etc...)
1242 * Returns : a pointer to the (newly allocated) modified buffer.
1243 * or NULL if there were no hits or something went wrong
1245 *********************************************************************/
1246 char *pcrs_filter_response(struct client_state *csp)
1251 char *old = csp->iob->cur, *new = NULL;
1254 struct file_list *fl;
1255 struct re_filterfile_spec *b;
1256 struct list_entry *filtername;
1261 if (csp->iob->cur >= csp->iob->eod)
1265 size = csp->iob->eod - csp->iob->cur;
1267 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1269 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1274 * If the body has a "chunked" transfer-encoding,
1275 * get rid of it first, adjusting size and iob->eod
1277 if (csp->flags & CSP_FLAG_CHUNKED)
1279 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1280 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1284 csp->iob->eod = csp->iob->cur + size;
1285 csp->flags |= CSP_FLAG_MODIFIED;
1289 * For all applying +filter actions, look if a filter by that
1290 * name exists and if yes, execute it's pcrs_joblist on the
1293 for (b = fl->f; b; b = b->next)
1295 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1296 filtername ; filtername = filtername->next)
1298 if (strcmp(b->filtername, filtername->str) == 0)
1300 int current_hits = 0;
1302 if ( NULL == b->joblist )
1304 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->filtername);
1308 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1309 csp->http->hostport, csp->http->path, size, b->filtername);
1311 /* Apply all jobs from the joblist */
1312 for (job = b->joblist; NULL != job; job = job->next)
1314 current_hits += pcrs_execute(job, old, size, &new, &size);
1315 if (old != csp->iob->cur) free(old);
1319 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1320 hits += current_hits;
1326 * If there were no hits, destroy our copy and let
1327 * chat() use the original in csp->iob
1335 csp->flags |= CSP_FLAG_MODIFIED;
1336 csp->content_length = size;
1344 /*********************************************************************
1346 * Function : gif_deanimate_response
1348 * Description : Deanimate the GIF image that has been accumulated in
1349 * csp->iob->buf, set csp->content_length to the modified
1350 * size and raise the CSP_FLAG_MODIFIED flag.
1353 * 1 : csp = Current client state (buffers, headers, etc...)
1355 * Returns : a pointer to the (newly allocated) modified buffer.
1356 * or NULL in case something went wrong.
1358 *********************************************************************/
1359 char *gif_deanimate_response(struct client_state *csp)
1361 struct binbuffer *in, *out;
1363 size_t size = csp->iob->eod - csp->iob->cur;
1366 * If the body has a "chunked" transfer-encoding,
1367 * get rid of it first, adjusting size and iob->eod
1369 if (csp->flags & CSP_FLAG_CHUNKED)
1371 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1372 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1376 csp->iob->eod = csp->iob->cur + size;
1377 csp->flags |= CSP_FLAG_MODIFIED;
1380 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1381 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1383 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1387 in->buffer = csp->iob->cur;
1390 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1392 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1399 if ((int)size == out->offset)
1401 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1405 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1407 csp->content_length = out->offset;
1408 csp->flags |= CSP_FLAG_MODIFIED;
1418 /*********************************************************************
1420 * Function : remove_chunked_transfer_coding
1422 * Description : In-situ remove the "chunked" transfer coding as defined
1423 * in rfc2616 from a buffer.
1426 * 1 : buffer = Pointer to the text buffer
1427 * 2 : size = Number of bytes to be processed
1429 * Returns : The new size, i.e. the number of bytes from buffer which
1430 * are occupied by the stripped body, or 0 in case something
1433 *********************************************************************/
1434 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1437 unsigned int chunksize = 0;
1438 char *from_p, *to_p;
1441 from_p = to_p = buffer;
1443 if (sscanf(buffer, "%x", &chunksize) != 1)
1445 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1449 while (chunksize > 0)
1451 if (NULL == (from_p = strstr(from_p, "\r\n")))
1453 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1456 newsize += chunksize;
1459 memmove(to_p, from_p, (size_t) chunksize);
1460 to_p = buffer + newsize;
1461 from_p += chunksize + 2;
1463 if (sscanf(from_p, "%x", &chunksize) != 1)
1465 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1470 /* FIXME: Should this get its own loglevel? */
1471 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1477 /*********************************************************************
1479 * Function : url_actions
1481 * Description : Gets the actions for this URL.
1484 * 1 : http = http_request request for blocked URLs
1485 * 2 : csp = Current client state (buffers, headers, etc...)
1489 *********************************************************************/
1490 void url_actions(struct http_request *http,
1491 struct client_state *csp)
1493 struct file_list *fl;
1494 struct url_actions *b;
1496 init_current_action(csp->action);
1498 if (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL))
1503 apply_url_actions(csp->action, http, b);
1508 /*********************************************************************
1510 * Function : apply_url_actions
1512 * Description : Applies a list of URL actions.
1515 * 1 : action = Destination.
1516 * 2 : http = Current URL
1517 * 3 : b = list of URL actions to apply
1521 *********************************************************************/
1522 void apply_url_actions(struct current_action_spec *action,
1523 struct http_request *http,
1524 struct url_actions *b)
1528 /* Should never happen */
1532 for (b = b->next; NULL != b; b = b->next)
1534 if (url_match(b->url, http))
1536 merge_current_action(action, b->action);
1542 /*********************************************************************
1544 * Function : forward_url
1546 * Description : Should we forward this to another proxy?
1549 * 1 : http = http_request request for current URL
1550 * 2 : csp = Current client state (buffers, headers, etc...)
1552 * Returns : Pointer to forwarding information.
1554 *********************************************************************/
1555 const struct forward_spec * forward_url(struct http_request *http,
1556 struct client_state *csp)
1558 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1559 struct forward_spec *fwd = csp->config->forward;
1568 if (url_match(fwd->url, http))