1 const char filters_rcs[] = "$Id: filters.c,v 1.49 2002/03/16 20:29:14 oes 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 * Privoxy 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 1.49 2002/03/16 20:29:14 oes
44 * Revision 1.48 2002/03/13 20:25:34 oes
45 * Better logging for content filters
47 * Revision 1.47 2002/03/13 00:30:52 jongfoster
49 * Added option of always sending redirect for imageblock,
50 * currently disabled with #if 0.
52 * Revision 1.46 2002/03/12 01:42:49 oes
53 * Introduced modular filters
55 * Revision 1.45 2002/03/08 16:47:50 oes
56 * Added choice beween GIF and PNG built-in images
58 * Revision 1.44 2002/03/07 03:49:31 oes
59 * - Fixed compiler warnings etc
60 * - Changed built-in images from GIF to PNG
61 * (with regard to Unisys patent issue)
62 * - Added a 4x4 pattern PNG which is less intrusive
63 * than the logo but also clearly marks the deleted banners
65 * Revision 1.43 2002/01/22 23:51:59 jongfoster
66 * Replacing strsav() with the safer string_append().
68 * Adding missing html_encode() to error message generators. Where encoded
69 * and unencoded versions of a string were provided, removing the unencoded
72 * Revision 1.42 2002/01/17 21:00:32 jongfoster
73 * Moving all our URL and URL pattern parsing code to urlmatch.c.
75 * Using a single, simple url_match(pattern,url) function - rather than
76 * the 3-line match routine which was repeated all over the place.
78 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
80 * Using parse_http_url() to parse URLs without faking a HTTP
81 * request line for parse_http_request().
83 * Revision 1.41 2001/11/13 00:14:07 jongfoster
84 * Fixing stupid bug now I've figured out what || means.
85 * (It always returns 0 or 1, not one of it's paramaters.)
87 * Revision 1.40 2001/10/26 17:37:55 oes
88 * - Re-enabled Netscape 200/404 bug workaround in block_url():
89 * - Removed OS/2 special case
90 * - Made block_url() independant from sed() having been run
91 * - Made trust_url independant from sed() having been run
92 * - Made is_imageurl independant from sed() having been run.
93 * It now checks User-Agent: and Accept: by itself.
96 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
97 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
98 * threads to call select() simultaneously. So, it's time to do a real, live,
99 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
100 * (native). Both versions will work, but using __OS2__ offers multi-threading.
102 * Revision 1.38 2001/10/23 21:32:33 jongfoster
103 * Adding error-checking to selected functions
105 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
106 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
107 * filters.c. Added a FIXME in front of the offending code. I'll gladly
108 * put in a better/more robust fix for all parties if one is presented...
109 * It seems that just returning 200 instead of 404 would pretty much fix
110 * it for everyone, but I don't know all the history of the problem.
112 * Revision 1.36 2001/10/10 16:44:16 oes
113 * Added match_portlist function
115 * Revision 1.35 2001/10/07 15:41:23 oes
116 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
118 * New function remove_chunked_transfer_coding that strips chunked
119 * transfer coding to plain and is called by pcrs_filter_response
120 * and gif_deanimate_response if neccessary
122 * Improved handling of zero-change re_filter runs
124 * pcrs_filter_response and gif_deanimate_response now remove
125 * chunked transfer codeing before processing the body.
127 * Revision 1.34 2001/09/20 15:49:36 steudten
129 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
130 * See cgi.c fill_template().
132 * Revision 1.33 2001/09/16 17:05:14 jongfoster
133 * Removing unused #include showarg.h
135 * Revision 1.32 2001/09/16 13:21:27 jongfoster
136 * Changes to use new list functions.
138 * Revision 1.31 2001/09/16 11:38:02 jongfoster
139 * Splitting fill_template() into 2 functions:
140 * template_load() loads the file
141 * template_fill() performs the PCRS regexps.
142 * This is because the CGI edit interface has a "table row"
143 * template which is used many times in the page - this
144 * change means it's only loaded from disk once.
146 * Revision 1.30 2001/09/16 11:00:10 jongfoster
147 * New function alloc_http_response, for symmetry with free_http_response
149 * Revision 1.29 2001/09/13 23:32:40 jongfoster
150 * Moving image data to cgi.c rather than cgi.h
151 * Fixing a GPF under Win32 (and any other OS that protects global
152 * constants from being written to).
154 * Revision 1.28 2001/09/10 10:18:51 oes
155 * Silenced compiler warnings
157 * Revision 1.27 2001/08/05 16:06:20 jongfoster
158 * Modifiying "struct map" so that there are now separate header and
159 * "map_entry" structures. This means that functions which modify a
160 * map no longer need to return a pointer to the modified map.
161 * Also, it no longer reverses the order of the entries (which may be
162 * important with some advanced template substitutions).
164 * Revision 1.26 2001/07/30 22:08:36 jongfoster
165 * Tidying up #defines:
166 * - All feature #defines are now of the form FEATURE_xxx
167 * - Permanently turned off WIN_GUI_EDIT
168 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
170 * Revision 1.25 2001/07/26 10:09:46 oes
171 * Made browser detection a little less naive
173 * Revision 1.24 2001/07/25 17:22:51 oes
174 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
176 * Revision 1.23 2001/07/23 13:40:12 oes
177 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
179 * Revision 1.22 2001/07/18 12:29:34 oes
180 * - Made gif_deanimate_response respect
181 * csp->action->string[ACTION_STRING_DEANIMATE]
182 * - Logging cosmetics
184 * Revision 1.21 2001/07/13 13:59:53 oes
185 * - Introduced gif_deanimate_response which shares the
186 * generic content modification interface of pcrs_filter_response
187 * and acts as a wrapper to deanimate.c:gif_deanimate()
188 * - Renamed re_process_buffer to pcrs_filter_response
189 * - pcrs_filter_response now returns NULL on failiure
190 * - Removed all #ifdef PCRS
192 * Revision 1.20 2001/07/01 17:01:04 oes
193 * Added comments and missing return statement in is_untrusted_url()
195 * Revision 1.19 2001/06/29 21:45:41 oes
196 * Indentation, CRLF->LF, Tab-> Space
198 * Revision 1.18 2001/06/29 13:27:38 oes
199 * - Cleaned up, renamed and reorderd functions
200 * and improved comments
203 * - Ported to CGI platform. Now delivers
204 * http_response or NULL
205 * - Unified HTML and GIF generation (moved image detection
206 * and GIF generation here from jcc.c:chat())
207 * - Fixed HTTP status to:
208 * - 403 (Forbidden) for the "blocked" HTML message
209 * - 200 (OK) for GIF answers
210 * - 302 (Redirect) for redirect to GIF
213 * - Ported to CGI platform. Now delivers
214 * http_response or NULL
215 * - Separated detection of untrusted URL into
216 * (bool)is_untrusted_url
217 * - Added enforcement of untrusted requests
219 * - Moved redirect_url() from cgi.c to here
220 * and ported it to the CGI platform
222 * - Removed logentry from cancelled commit
224 * Revision 1.17 2001/06/09 10:55:28 jongfoster
225 * Changing BUFSIZ ==> BUFFER_SIZE
227 * Revision 1.16 2001/06/07 23:10:26 jongfoster
228 * Allowing unanchored domain patterns to back off and retry
229 * if they partially match. Optimized right-anchored patterns.
230 * Moving ACL and forward files into config file.
231 * Replacing struct gateway with struct forward_spec
233 * Revision 1.15 2001/06/03 19:12:00 oes
234 * extracted-CGI relevant stuff
236 * Revision 1.14 2001/06/01 10:30:55 oes
237 * Added optional left-anchoring to domaincmp
239 * Revision 1.13 2001/05/31 21:21:30 jongfoster
240 * Permissionsfile / actions file changes:
241 * - Changed "permission" to "action" throughout
242 * - changes to file format to allow string parameters
243 * - Moved helper functions to actions.c
245 * Revision 1.12 2001/05/31 17:35:20 oes
247 * - Enhanced domain part globbing with infix and prefix asterisk
248 * matching and optional unanchored operation
250 * Revision 1.11 2001/05/29 11:53:23 oes
251 * "See why" link added to "blocked" page
253 * Revision 1.10 2001/05/29 09:50:24 jongfoster
254 * Unified blocklist/imagelist/permissionslist.
255 * File format is still under discussion, but the internal changes
258 * Also modified interceptor behaviour:
259 * - We now intercept all URLs beginning with one of the following
260 * prefixes (and *only* these prefixes):
262 * * http://ijbswa.sf.net/config/
263 * * http://ijbswa.sourceforge.net/config/
264 * - New interceptors "home page" - go to http://i.j.b/ to see it.
265 * - Internal changes so that intercepted and fast redirect pages
266 * are not replaced with an image.
267 * - Interceptors now have the option to send a binary page direct
268 * to the client. (i.e. ijb-send-banner uses this)
269 * - Implemented show-url-info interceptor. (Which is why I needed
270 * the above interceptors changes - a typical URL is
271 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
272 * The previous mechanism would not have intercepted that, and
273 * if it had been intercepted then it then it would have replaced
276 * Revision 1.9 2001/05/27 22:17:04 oes
278 * - re_process_buffer no longer writes the modified buffer
279 * to the client, which was very ugly. It now returns the
280 * buffer, which it is then written by chat.
282 * - content_length now adjusts the Content-Length: header
283 * for modified documents rather than crunch()ing it.
284 * (Length info in csp->content_length, which is 0 for
285 * unmodified documents)
287 * - For this to work, sed() is called twice when filtering.
289 * Revision 1.8 2001/05/26 17:13:28 jongfoster
290 * Filled in a function comment.
292 * Revision 1.7 2001/05/26 15:26:15 jongfoster
293 * ACL feature now provides more security by immediately dropping
294 * connections from untrusted hosts.
296 * Revision 1.6 2001/05/26 00:28:36 jongfoster
297 * Automatic reloading of config file.
298 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
299 * Most of the global variables have been moved to a new
300 * struct configuration_spec, accessed through csp->config->globalname
301 * Most of the globals remaining are used by the Win32 GUI.
303 * Revision 1.5 2001/05/25 22:34:30 jongfoster
306 * Revision 1.4 2001/05/22 18:46:04 oes
308 * - Enabled filtering banners by size rather than URL
309 * by adding patterns that replace all standard banner
310 * sizes with the "Junkbuster" gif to the re_filterfile
312 * - Enabled filtering WebBugs by providing a pattern
313 * which kills all 1x1 images
315 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
316 * which is selected by the (nonstandard and therefore
317 * capital) letter 'U' in the option string.
318 * It causes the quantifiers to be ungreedy by default.
319 * Appending a ? turns back to greedy (!).
321 * - Added a new interceptor ijb-send-banner, which
322 * sends back the "Junkbuster" gif. Without imagelist or
323 * MSIE detection support, or if tinygif = 1, or the
324 * URL isn't recognized as an imageurl, a lame HTML
325 * explanation is sent instead.
327 * - Added new feature, which permits blocking remote
328 * script redirects and firing back a local redirect
330 * The feature is conditionally compiled, i.e. it
331 * can be disabled with --disable-fast-redirects,
332 * plus it must be activated by a "fast-redirects"
333 * line in the config file, has its own log level
334 * and of course wants to be displayed by show-proxy-args
335 * Note: Boy, all the #ifdefs in 1001 locations and
336 * all the fumbling with configure.in and acconfig.h
337 * were *way* more work than the feature itself :-(
339 * - Because a generic redirect template was needed for
340 * this, tinygif = 3 now uses the same.
342 * - Moved GIFs, and other static HTTP response templates
347 * - Removed some >400 CRs again (Jon, you really worked
350 * Revision 1.3 2001/05/20 16:44:47 jongfoster
351 * Removing last hardcoded JunkBusters.com URLs.
353 * Revision 1.2 2001/05/20 01:21:20 jongfoster
354 * Version 2.9.4 checkin.
355 * - Merged popupfile and cookiefile, and added control over PCRS
356 * filtering, in new "permissionsfile".
357 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
358 * file error you now get a message box (in the Win32 GUI) rather
359 * than the program exiting with no explanation.
360 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
362 * - Removed tabs from "config"
363 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
364 * - Bumped up version number.
366 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
367 * Initial import of version 2.9.3 source tree
370 *********************************************************************/
376 #include <sys/types.h>
385 #endif /* ndef __OS2__ */
386 #include <netinet/in.h>
388 #include <winsock2.h>
389 #endif /* ndef _WIN32 */
393 #endif /* def __OS2__ */
401 #include "jbsockets.h"
402 #include "miscutil.h"
406 #include "deanimate.h"
407 #include "urlmatch.h"
413 const char filters_h_rcs[] = FILTERS_H_VERSION;
415 /* Fix a problem with Solaris. There should be no effect on other
417 * Solaris's isspace() is a macro which uses it's argument directly
418 * as an array index. Therefore we need to make sure that high-bit
419 * characters generate +ve values, and ideally we also want to make
420 * the argument match the declared parameter type of "int".
422 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
426 /*********************************************************************
428 * Function : block_acl
430 * Description : Block this request?
431 * Decide yes or no based on ACL file.
434 * 1 : dst = The proxy or gateway address this is going to.
435 * Or NULL to check all possible targets.
436 * 2 : csp = Current client state (buffers, headers, etc...)
437 * Also includes the client IP address.
439 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
441 *********************************************************************/
442 int block_acl(struct access_control_addr *dst, struct client_state *csp)
444 struct access_control_list *acl = csp->config->acl;
446 /* if not using an access control list, then permit the connection */
452 /* search the list */
455 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
459 /* Just want to check if they have any access */
460 if (acl->action == ACL_PERMIT)
465 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
466 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
468 if (acl->action == ACL_PERMIT)
486 /*********************************************************************
488 * Function : acl_addr
490 * Description : Called from `load_config' to parse an ACL address.
493 * 1 : aspec = String specifying ACL address.
494 * 2 : aca = struct access_control_addr to fill in.
496 * Returns : 0 => Ok, everything else is an error.
498 *********************************************************************/
499 int acl_addr(char *aspec, struct access_control_addr *aca)
501 int i, masklength, port;
507 if ((p = strchr(aspec, '/')) != NULL)
511 if (ijb_isdigit(*p) == 0)
515 masklength = atoi(p);
518 if ((masklength < 0) || (masklength > 32))
523 if ((p = strchr(aspec, ':')) != NULL)
527 if (ijb_isdigit(*p) == 0)
536 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
538 if (aca->addr == INADDR_NONE)
543 /* build the netmask */
545 for (i=1; i <= masklength ; i++)
547 aca->mask |= (1 << (32 - i));
550 /* now mask off the host portion of the ip address
551 * (i.e. save on the network portion of the address).
553 aca->addr = aca->addr & aca->mask;
558 #endif /* def FEATURE_ACL */
561 /*********************************************************************
563 * Function : match_portlist
565 * Description : Check if a given number is covered by a comma
566 * separated list of numbers and ranges (a,b-c,d,..)
569 * 1 : portlist = String with list
570 * 2 : port = port to check
572 * Returns : 0 => no match
575 *********************************************************************/
576 int match_portlist(const char *portlist, int port)
578 char *min, *max, *next, *portlist_copy;
580 min = next = portlist_copy = strdup(portlist);
583 * Zero-terminate first item and remember offset for next
585 if (NULL != (next = strchr(portlist_copy, (int) ',')))
591 * Loop through all items, checking for match
595 if (NULL == (max = strchr(min, (int) '-')))
598 * No dash, check for equality
600 if (port == atoi(min))
609 * This is a range, so check if between min and max,
610 * or, if max was omitted, between min and 65K
613 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
627 * Zero-terminate next item and remember offset for n+1
629 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
641 /*********************************************************************
643 * Function : block_url
645 * Description : Called from `chat'. Check to see if we need to block this.
648 * 1 : csp = Current client state (buffers, headers, etc...)
650 * Returns : NULL => unblocked, else HTTP block response
652 *********************************************************************/
653 struct http_response *block_url(struct client_state *csp)
655 #ifdef FEATURE_IMAGE_BLOCKING
657 #endif /* def FEATURE_IMAGE_BLOCKING */
658 struct http_response *rsp;
661 * If it's not blocked, don't block it ;-)
663 if ((csp->action->flags & ACTION_BLOCK) == 0)
669 * Else, prepare a response
671 if (NULL == (rsp = alloc_http_response()))
673 return cgi_error_memory();
677 * If it's an image-url, send back an image or redirect
678 * as specified by the relevant +image action
680 #ifdef FEATURE_IMAGE_BLOCKING
681 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
684 /* determine HOW images should be blocked */
685 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
687 #if 1 /* Two alternative strategies, use this one for now: */
689 /* and handle accordingly: */
690 if ((p == NULL) || (0 == strcmpic(p, "logo")))
692 rsp->body = bindup(image_logo_data, image_logo_length);
693 if (rsp->body == NULL)
695 free_http_response(rsp);
696 return cgi_error_memory();
698 rsp->content_length = image_logo_length;
700 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
702 free_http_response(rsp);
703 return cgi_error_memory();
707 else if (0 == strcmpic(p, "blank"))
709 rsp->body = bindup(image_blank_data, image_blank_length);
710 if (rsp->body == NULL)
712 free_http_response(rsp);
713 return cgi_error_memory();
715 rsp->content_length = image_blank_length;
717 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
719 free_http_response(rsp);
720 return cgi_error_memory();
724 else if (0 == strcmpic(p, "pattern"))
726 rsp->body = bindup(image_pattern_data, image_pattern_length);
727 if (rsp->body == NULL)
729 free_http_response(rsp);
730 return cgi_error_memory();
732 rsp->content_length = image_pattern_length;
734 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
736 free_http_response(rsp);
737 return cgi_error_memory();
743 rsp->status = strdup("302 Local Redirect from Junkbuster");
744 if (rsp->status == NULL)
746 free_http_response(rsp);
747 return cgi_error_memory();
750 if (enlist_unique_header(rsp->headers, "Location", p))
752 free_http_response(rsp);
753 return cgi_error_memory();
757 #else /* Following code is disabled for now */
759 /* and handle accordingly: */
760 if ((p == NULL) || (0 == strcmpic(p, "logo")))
762 p = CGI_PREFIX "send-banner?type=logo";
764 else if (0 == strcmpic(p, "blank"))
766 p = CGI_PREFIX "send-banner?type=blank";
768 else if (0 == strcmpic(p, "pattern"))
770 p = CGI_PREFIX "send-banner?type=pattern";
772 rsp->status = strdup("302 Local Redirect from Junkbuster");
773 if (rsp->status == NULL)
775 free_http_response(rsp);
776 return cgi_error_memory();
779 if (enlist_unique_header(rsp->headers, "Location", p))
781 free_http_response(rsp);
782 return cgi_error_memory();
784 #endif /* Preceeding code is disabled for now */
787 #endif /* def FEATURE_IMAGE_BLOCKING */
790 * Else, generate an HTML "blocked" message:
794 struct map * exports;
797 * Workaround for stupid Netscape bug which prevents
798 * pages from being displayed if loading a referenced
799 * JavaScript or style sheet fails. So make it appear
800 * as if it succeeded.
802 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
803 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
804 && !strstr(p, "Gecko") /* save Mozilla, */
805 && !strstr(p, "compatible") /* MSIE */
806 && !strstr(p, "Opera")) /* and Opera. */
808 rsp->status = strdup("200 Request for blocked URL");
812 rsp->status = strdup("404 Request for blocked URL");
815 if (rsp->status == NULL)
817 free_http_response(rsp);
818 return cgi_error_memory();
821 exports = default_exports(csp, NULL);
824 free_http_response(rsp);
825 return cgi_error_memory();
828 #ifdef FEATURE_FORCE_LOAD
829 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
830 #else /* ifndef FEATURE_FORCE_LOAD */
831 err = map_block_killer(exports, "force-support");
832 #endif /* ndef FEATURE_FORCE_LOAD */
834 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
835 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
840 free_http_response(rsp);
841 return cgi_error_memory();
844 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
847 free_http_response(rsp);
848 return cgi_error_memory();
852 return finish_http_response(rsp);
858 /*********************************************************************
860 * Function : trust_url FIXME: I should be called distrust_url
862 * Description : Calls is_untrusted_url to determine if the URL is trusted
863 * and if not, returns a HTTP 304 response with a reject message.
866 * 1 : csp = Current client state (buffers, headers, etc...)
868 * Returns : NULL => trusted, else http_response.
870 *********************************************************************/
871 struct http_response *trust_url(struct client_state *csp)
873 struct http_response *rsp;
874 struct map * exports;
875 char buf[BUFFER_SIZE];
877 struct url_spec **tl;
882 * Don't bother to work on trusted URLs
884 if (!is_untrusted_url(csp))
890 * Else, prepare a response:
892 if (NULL == (rsp = alloc_http_response()))
894 return cgi_error_memory();
897 exports = default_exports(csp, NULL);
900 free_http_response(rsp);
901 return cgi_error_memory();
905 * Export the host, port, and referrer information
907 err = map(exports, "hostport", 1, csp->http->hostport, 1);
908 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
910 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
912 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
916 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
922 free_http_response(rsp);
923 return cgi_error_memory();
927 * Export the trust list
930 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
932 sprintf(buf, "<li>%s</li>\n", t->spec);
933 string_append(&p, buf);
935 err = map(exports, "trusted-referrers", 1, p, 0);
940 free_http_response(rsp);
941 return cgi_error_memory();
945 * Export the trust info, if available
947 if (csp->config->trust_info->first)
949 struct list_entry *l;
952 for (l = csp->config->trust_info->first; l ; l = l->next)
954 sprintf(buf, "<li> <a href=%s>%s</a><br>\n",l->str, l->str);
955 string_append(&p, buf);
957 err = map(exports, "trust-info", 1, p, 0);
961 err = map_block_killer(exports, "have-trust-info");
967 free_http_response(rsp);
968 return cgi_error_memory();
972 * Export the force prefix or the force conditional block killer
974 #ifdef FEATURE_FORCE_LOAD
975 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
976 #else /* ifndef FEATURE_FORCE_LOAD */
977 err = map_block_killer(exports, "force-support");
978 #endif /* ndef FEATURE_FORCE_LOAD */
983 free_http_response(rsp);
984 return cgi_error_memory();
990 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
993 free_http_response(rsp);
994 return cgi_error_memory();
997 return finish_http_response(rsp);
999 #endif /* def FEATURE_TRUST */
1002 #ifdef FEATURE_FAST_REDIRECTS
1003 /*********************************************************************
1005 * Function : redirect_url
1007 * Description : Checks for redirection URLs and returns a HTTP redirect
1008 * to the destination URL, if necessary
1011 * 1 : csp = Current client state (buffers, headers, etc...)
1013 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1015 *********************************************************************/
1016 struct http_response *redirect_url(struct client_state *csp)
1019 struct http_response *rsp;
1021 p = q = csp->http->path;
1022 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1025 * find the last URL encoded in the request
1027 while ((p = strstr(p, "http://")) != NULL)
1033 * if there was any, generate and return a HTTP redirect
1035 if (q != csp->http->path)
1037 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1039 if (NULL == (rsp = alloc_http_response()))
1041 return cgi_error_memory();
1044 if ( enlist_unique_header(rsp->headers, "Location", q)
1045 || (NULL == (rsp->status = strdup("302 Local Redirect from Junkbuster"))) )
1047 free_http_response(rsp);
1048 return cgi_error_memory();
1051 return finish_http_response(rsp);
1059 #endif /* def FEATURE_FAST_REDIRECTS */
1062 #ifdef FEATURE_IMAGE_BLOCKING
1063 /*********************************************************************
1065 * Function : is_imageurl
1067 * Description : Given a URL, decide whether it is an image or not,
1068 * using either the info from a previous +image action
1069 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, the info from
1070 * the browser's accept header.
1073 * 1 : csp = Current client state (buffers, headers, etc...)
1075 * Returns : True (nonzero) if URL is an image, false (0)
1078 *********************************************************************/
1079 int is_imageurl(struct client_state *csp)
1081 #ifdef FEATURE_IMAGE_DETECT_MSIE
1084 tmp = get_header_value(csp->headers, "User-Agent:");
1085 if (tmp && strstr(tmp, "MSIE"))
1087 tmp = get_header_value(csp->headers, "Accept:");
1088 if (tmp && strstr(tmp, "image/gif"))
1090 /* Client will accept HTML. If this seems counterintuitive,
1100 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1102 return ((csp->action->flags & ACTION_IMAGE) != 0);
1105 #endif /* def FEATURE_IMAGE_BLOCKING */
1108 #ifdef FEATURE_COOKIE_JAR
1109 /*********************************************************************
1111 * Function : is_untrusted_url
1113 * Description : Should we "distrust" this URL (and block it)?
1115 * Yes if it matches a line in the trustfile, or if the
1116 * referrer matches a line starting with "+" in the
1121 * 1 : csp = Current client state (buffers, headers, etc...)
1123 * Returns : 0 => trusted, 1 => untrusted
1125 *********************************************************************/
1126 int is_untrusted_url(struct client_state *csp)
1128 struct file_list *fl;
1129 struct block_spec *b;
1130 struct url_spec **trusted_url;
1131 struct http_request rhttp[1];
1132 const char * referer;
1136 * If we don't have a trustlist, we trust everybody
1138 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1143 memset(rhttp, '\0', sizeof(*rhttp));
1146 * Do we trust the request URL itself?
1148 for (b = b->next; b ; b = b->next)
1150 if (url_match(b->url, csp->http))
1156 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1158 /* no referrer was supplied */
1163 * If not, do we maybe trust its referrer?
1168 * Parse the URL from the referrer
1171 err = parse_http_url(referer, rhttp, csp);
1177 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1179 if (url_match(*trusted_url, rhttp))
1181 /* if the URL's referrer is from a trusted referrer, then
1182 * add the target spec to the trustfile as an unblocked
1183 * domain and return NULL (which means it's OK).
1188 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1192 char * new_entry = strdup("~");
1194 string_append(&new_entry, csp->http->hostport);
1196 path = csp->http->path;
1197 if ( (path[0] == '/')
1199 && ((path_end = strchr(path + 2, '/')) != NULL))
1201 /* since this path points into a user's home space
1202 * be sure to include this spec in the trustfile.
1204 int path_len = path_end - path; /* save offset */
1205 path = strdup(path); /* Copy string */
1208 path_end = path + path_len; /* regenerate ptr to new buffer */
1209 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1211 string_join(&new_entry, path);
1214 if (new_entry != NULL)
1216 fprintf(fp, "%s\n", new_entry);
1221 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1222 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1232 #endif /* def FEATURE_COOKIE_JAR */
1235 /*********************************************************************
1237 * Function : pcrs_filter_response
1239 * Description : Ecexute all text substitutions from all applying
1240 * +filter actions on the text buffer that's been accumulated
1241 * in csp->iob->buf. If this changes the contents, set
1242 * csp->content_length to the modified size and raise the
1243 * CSP_FLAG_MODIFIED flag.
1246 * 1 : csp = Current client state (buffers, headers, etc...)
1248 * Returns : a pointer to the (newly allocated) modified buffer.
1249 * or NULL if there were no hits or something went wrong
1251 *********************************************************************/
1252 char *pcrs_filter_response(struct client_state *csp)
1257 char *old = csp->iob->cur, *new = NULL;
1260 struct file_list *fl;
1261 struct re_filterfile_spec *b;
1262 struct list_entry *filtername;
1267 if (csp->iob->cur >= csp->iob->eod)
1271 size = csp->iob->eod - csp->iob->cur;
1273 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1275 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1280 * If the body has a "chunked" transfer-encoding,
1281 * get rid of it first, adjusting size and iob->eod
1283 if (csp->flags & CSP_FLAG_CHUNKED)
1285 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1286 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1290 csp->iob->eod = csp->iob->cur + size;
1291 csp->flags |= CSP_FLAG_MODIFIED;
1295 * For all applying +filter actions, look if a filter by that
1296 * name exists and if yes, execute it's pcrs_joblist on the
1299 for (b = fl->f; b; b = b->next)
1301 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1302 filtername ; filtername = filtername->next)
1304 if (strcmp(b->name, filtername->str) == 0)
1306 int current_hits = 0;
1308 if ( NULL == b->joblist )
1310 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1314 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1315 csp->http->hostport, csp->http->path, size, b->name);
1317 /* Apply all jobs from the joblist */
1318 for (job = b->joblist; NULL != job; job = job->next)
1320 current_hits += pcrs_execute(job, old, size, &new, &size);
1321 if (old != csp->iob->cur) free(old);
1325 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1326 hits += current_hits;
1332 * If there were no hits, destroy our copy and let
1333 * chat() use the original in csp->iob
1341 csp->flags |= CSP_FLAG_MODIFIED;
1342 csp->content_length = size;
1350 /*********************************************************************
1352 * Function : gif_deanimate_response
1354 * Description : Deanimate the GIF image that has been accumulated in
1355 * csp->iob->buf, set csp->content_length to the modified
1356 * size and raise the CSP_FLAG_MODIFIED flag.
1359 * 1 : csp = Current client state (buffers, headers, etc...)
1361 * Returns : a pointer to the (newly allocated) modified buffer.
1362 * or NULL in case something went wrong.
1364 *********************************************************************/
1365 char *gif_deanimate_response(struct client_state *csp)
1367 struct binbuffer *in, *out;
1369 size_t size = csp->iob->eod - csp->iob->cur;
1372 * If the body has a "chunked" transfer-encoding,
1373 * get rid of it first, adjusting size and iob->eod
1375 if (csp->flags & CSP_FLAG_CHUNKED)
1377 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1378 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1382 csp->iob->eod = csp->iob->cur + size;
1383 csp->flags |= CSP_FLAG_MODIFIED;
1386 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1387 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1389 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1393 in->buffer = csp->iob->cur;
1396 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1398 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1405 if ((int)size == out->offset)
1407 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1411 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1413 csp->content_length = out->offset;
1414 csp->flags |= CSP_FLAG_MODIFIED;
1424 /*********************************************************************
1426 * Function : remove_chunked_transfer_coding
1428 * Description : In-situ remove the "chunked" transfer coding as defined
1429 * in rfc2616 from a buffer.
1432 * 1 : buffer = Pointer to the text buffer
1433 * 2 : size = Number of bytes to be processed
1435 * Returns : The new size, i.e. the number of bytes from buffer which
1436 * are occupied by the stripped body, or 0 in case something
1439 *********************************************************************/
1440 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1443 unsigned int chunksize = 0;
1444 char *from_p, *to_p;
1447 from_p = to_p = buffer;
1449 if (sscanf(buffer, "%x", &chunksize) != 1)
1451 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1455 while (chunksize > 0)
1457 if (NULL == (from_p = strstr(from_p, "\r\n")))
1459 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1462 newsize += chunksize;
1465 memmove(to_p, from_p, (size_t) chunksize);
1466 to_p = buffer + newsize;
1467 from_p += chunksize + 2;
1469 if (sscanf(from_p, "%x", &chunksize) != 1)
1471 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1476 /* FIXME: Should this get its own loglevel? */
1477 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1483 /*********************************************************************
1485 * Function : url_actions
1487 * Description : Gets the actions for this URL.
1490 * 1 : http = http_request request for blocked URLs
1491 * 2 : csp = Current client state (buffers, headers, etc...)
1495 *********************************************************************/
1496 void url_actions(struct http_request *http,
1497 struct client_state *csp)
1499 struct file_list *fl;
1500 struct url_actions *b;
1502 init_current_action(csp->action);
1504 if (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL))
1509 apply_url_actions(csp->action, http, b);
1514 /*********************************************************************
1516 * Function : apply_url_actions
1518 * Description : Applies a list of URL actions.
1521 * 1 : action = Destination.
1522 * 2 : http = Current URL
1523 * 3 : b = list of URL actions to apply
1527 *********************************************************************/
1528 void apply_url_actions(struct current_action_spec *action,
1529 struct http_request *http,
1530 struct url_actions *b)
1534 /* Should never happen */
1538 for (b = b->next; NULL != b; b = b->next)
1540 if (url_match(b->url, http))
1542 merge_current_action(action, b->action);
1548 /*********************************************************************
1550 * Function : forward_url
1552 * Description : Should we forward this to another proxy?
1555 * 1 : http = http_request request for current URL
1556 * 2 : csp = Current client state (buffers, headers, etc...)
1558 * Returns : Pointer to forwarding information.
1560 *********************************************************************/
1561 const struct forward_spec * forward_url(struct http_request *http,
1562 struct client_state *csp)
1564 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1565 struct forward_spec *fwd = csp->config->forward;
1574 if (url_match(fwd->url, http))