1 const char filters_rcs[] = "$Id: filters.c,v 2.2 2002/09/04 15:38:24 oes Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/src/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://www.privoxy.org/
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 2.2 2002/09/04 15:38:24 oes
42 * Synced with the stable branch:
43 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
44 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
46 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
47 * - Bugfix: Executing a filters without jobs no longer results in
48 * turing off *all* filters.
49 * - Security fix: Malicious web servers can't cause a seg fault
50 * through bogus chunk sizes anymore
52 * Revision 2.1 2002/08/26 11:08:18 sarantis
55 * Revision 2.0 2002/06/04 14:34:21 jongfoster
56 * Moving source files to src/
58 * Revision 1.58 2002/04/24 02:11:17 oes
59 * Jon's multiple AF patch: url_actions now evaluates rules
62 * Revision 1.57 2002/04/08 20:38:34 swa
65 * Revision 1.56 2002/04/05 15:51:24 oes
66 * - bugfix: error-pages now get correct request protocol
67 * - fix for invalid HTML in trust info
69 * Revision 1.55 2002/04/02 16:13:51 oes
70 * Fix: No "Go there anyway" for SSL
72 * Revision 1.54 2002/04/02 14:55:56 oes
73 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
75 * Revision 1.53 2002/03/26 22:29:54 swa
76 * we have a new homepage!
78 * Revision 1.52 2002/03/24 16:35:57 jongfoster
81 * Revision 1.51 2002/03/24 15:23:33 jongfoster
84 * Revision 1.50 2002/03/24 13:25:43 swa
85 * name change related issues
87 * Revision 1.49 2002/03/16 20:29:14 oes
90 * Revision 1.48 2002/03/13 20:25:34 oes
91 * Better logging for content filters
93 * Revision 1.47 2002/03/13 00:30:52 jongfoster
95 * Added option of always sending redirect for imageblock,
96 * currently disabled with #if 0.
98 * Revision 1.46 2002/03/12 01:42:49 oes
99 * Introduced modular filters
101 * Revision 1.45 2002/03/08 16:47:50 oes
102 * Added choice beween GIF and PNG built-in images
104 * Revision 1.44 2002/03/07 03:49:31 oes
105 * - Fixed compiler warnings etc
106 * - Changed built-in images from GIF to PNG
107 * (with regard to Unisys patent issue)
108 * - Added a 4x4 pattern PNG which is less intrusive
109 * than the logo but also clearly marks the deleted banners
111 * Revision 1.43 2002/01/22 23:51:59 jongfoster
112 * Replacing strsav() with the safer string_append().
114 * Adding missing html_encode() to error message generators. Where encoded
115 * and unencoded versions of a string were provided, removing the unencoded
118 * Revision 1.42 2002/01/17 21:00:32 jongfoster
119 * Moving all our URL and URL pattern parsing code to urlmatch.c.
121 * Using a single, simple url_match(pattern,url) function - rather than
122 * the 3-line match routine which was repeated all over the place.
124 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
126 * Using parse_http_url() to parse URLs without faking a HTTP
127 * request line for parse_http_request().
129 * Revision 1.41 2001/11/13 00:14:07 jongfoster
130 * Fixing stupid bug now I've figured out what || means.
131 * (It always returns 0 or 1, not one of it's paramaters.)
133 * Revision 1.40 2001/10/26 17:37:55 oes
134 * - Re-enabled Netscape 200/404 bug workaround in block_url():
135 * - Removed OS/2 special case
136 * - Made block_url() independant from sed() having been run
137 * - Made trust_url independant from sed() having been run
138 * - Made is_imageurl independant from sed() having been run.
139 * It now checks User-Agent: and Accept: by itself.
142 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
143 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
144 * threads to call select() simultaneously. So, it's time to do a real, live,
145 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
146 * (native). Both versions will work, but using __OS2__ offers multi-threading.
148 * Revision 1.38 2001/10/23 21:32:33 jongfoster
149 * Adding error-checking to selected functions
151 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
152 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
153 * filters.c. Added a FIXME in front of the offending code. I'll gladly
154 * put in a better/more robust fix for all parties if one is presented...
155 * It seems that just returning 200 instead of 404 would pretty much fix
156 * it for everyone, but I don't know all the history of the problem.
158 * Revision 1.36 2001/10/10 16:44:16 oes
159 * Added match_portlist function
161 * Revision 1.35 2001/10/07 15:41:23 oes
162 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
164 * New function remove_chunked_transfer_coding that strips chunked
165 * transfer coding to plain and is called by pcrs_filter_response
166 * and gif_deanimate_response if neccessary
168 * Improved handling of zero-change re_filter runs
170 * pcrs_filter_response and gif_deanimate_response now remove
171 * chunked transfer codeing before processing the body.
173 * Revision 1.34 2001/09/20 15:49:36 steudten
175 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
176 * See cgi.c fill_template().
178 * Revision 1.33 2001/09/16 17:05:14 jongfoster
179 * Removing unused #include showarg.h
181 * Revision 1.32 2001/09/16 13:21:27 jongfoster
182 * Changes to use new list functions.
184 * Revision 1.31 2001/09/16 11:38:02 jongfoster
185 * Splitting fill_template() into 2 functions:
186 * template_load() loads the file
187 * template_fill() performs the PCRS regexps.
188 * This is because the CGI edit interface has a "table row"
189 * template which is used many times in the page - this
190 * change means it's only loaded from disk once.
192 * Revision 1.30 2001/09/16 11:00:10 jongfoster
193 * New function alloc_http_response, for symmetry with free_http_response
195 * Revision 1.29 2001/09/13 23:32:40 jongfoster
196 * Moving image data to cgi.c rather than cgi.h
197 * Fixing a GPF under Win32 (and any other OS that protects global
198 * constants from being written to).
200 * Revision 1.28 2001/09/10 10:18:51 oes
201 * Silenced compiler warnings
203 * Revision 1.27 2001/08/05 16:06:20 jongfoster
204 * Modifiying "struct map" so that there are now separate header and
205 * "map_entry" structures. This means that functions which modify a
206 * map no longer need to return a pointer to the modified map.
207 * Also, it no longer reverses the order of the entries (which may be
208 * important with some advanced template substitutions).
210 * Revision 1.26 2001/07/30 22:08:36 jongfoster
211 * Tidying up #defines:
212 * - All feature #defines are now of the form FEATURE_xxx
213 * - Permanently turned off WIN_GUI_EDIT
214 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
216 * Revision 1.25 2001/07/26 10:09:46 oes
217 * Made browser detection a little less naive
219 * Revision 1.24 2001/07/25 17:22:51 oes
220 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
222 * Revision 1.23 2001/07/23 13:40:12 oes
223 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
225 * Revision 1.22 2001/07/18 12:29:34 oes
226 * - Made gif_deanimate_response respect
227 * csp->action->string[ACTION_STRING_DEANIMATE]
228 * - Logging cosmetics
230 * Revision 1.21 2001/07/13 13:59:53 oes
231 * - Introduced gif_deanimate_response which shares the
232 * generic content modification interface of pcrs_filter_response
233 * and acts as a wrapper to deanimate.c:gif_deanimate()
234 * - Renamed re_process_buffer to pcrs_filter_response
235 * - pcrs_filter_response now returns NULL on failure
236 * - Removed all #ifdef PCRS
238 * Revision 1.20 2001/07/01 17:01:04 oes
239 * Added comments and missing return statement in is_untrusted_url()
241 * Revision 1.19 2001/06/29 21:45:41 oes
242 * Indentation, CRLF->LF, Tab-> Space
244 * Revision 1.18 2001/06/29 13:27:38 oes
245 * - Cleaned up, renamed and reorderd functions
246 * and improved comments
249 * - Ported to CGI platform. Now delivers
250 * http_response or NULL
251 * - Unified HTML and GIF generation (moved image detection
252 * and GIF generation here from jcc.c:chat())
253 * - Fixed HTTP status to:
254 * - 403 (Forbidden) for the "blocked" HTML message
255 * - 200 (OK) for GIF answers
256 * - 302 (Redirect) for redirect to GIF
259 * - Ported to CGI platform. Now delivers
260 * http_response or NULL
261 * - Separated detection of untrusted URL into
262 * (bool)is_untrusted_url
263 * - Added enforcement of untrusted requests
265 * - Moved redirect_url() from cgi.c to here
266 * and ported it to the CGI platform
268 * - Removed logentry from cancelled commit
270 * Revision 1.17 2001/06/09 10:55:28 jongfoster
271 * Changing BUFSIZ ==> BUFFER_SIZE
273 * Revision 1.16 2001/06/07 23:10:26 jongfoster
274 * Allowing unanchored domain patterns to back off and retry
275 * if they partially match. Optimized right-anchored patterns.
276 * Moving ACL and forward files into config file.
277 * Replacing struct gateway with struct forward_spec
279 * Revision 1.15 2001/06/03 19:12:00 oes
280 * extracted-CGI relevant stuff
282 * Revision 1.14 2001/06/01 10:30:55 oes
283 * Added optional left-anchoring to domaincmp
285 * Revision 1.13 2001/05/31 21:21:30 jongfoster
286 * Permissionsfile / actions file changes:
287 * - Changed "permission" to "action" throughout
288 * - changes to file format to allow string parameters
289 * - Moved helper functions to actions.c
291 * Revision 1.12 2001/05/31 17:35:20 oes
293 * - Enhanced domain part globbing with infix and prefix asterisk
294 * matching and optional unanchored operation
296 * Revision 1.11 2001/05/29 11:53:23 oes
297 * "See why" link added to "blocked" page
299 * Revision 1.10 2001/05/29 09:50:24 jongfoster
300 * Unified blocklist/imagelist/permissionslist.
301 * File format is still under discussion, but the internal changes
304 * Also modified interceptor behaviour:
305 * - We now intercept all URLs beginning with one of the following
306 * prefixes (and *only* these prefixes):
308 * * http://ijbswa.sf.net/config/
309 * * http://ijbswa.sourceforge.net/config/
310 * - New interceptors "home page" - go to http://i.j.b/ to see it.
311 * - Internal changes so that intercepted and fast redirect pages
312 * are not replaced with an image.
313 * - Interceptors now have the option to send a binary page direct
314 * to the client. (i.e. ijb-send-banner uses this)
315 * - Implemented show-url-info interceptor. (Which is why I needed
316 * the above interceptors changes - a typical URL is
317 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
318 * The previous mechanism would not have intercepted that, and
319 * if it had been intercepted then it then it would have replaced
322 * Revision 1.9 2001/05/27 22:17:04 oes
324 * - re_process_buffer no longer writes the modified buffer
325 * to the client, which was very ugly. It now returns the
326 * buffer, which it is then written by chat.
328 * - content_length now adjusts the Content-Length: header
329 * for modified documents rather than crunch()ing it.
330 * (Length info in csp->content_length, which is 0 for
331 * unmodified documents)
333 * - For this to work, sed() is called twice when filtering.
335 * Revision 1.8 2001/05/26 17:13:28 jongfoster
336 * Filled in a function comment.
338 * Revision 1.7 2001/05/26 15:26:15 jongfoster
339 * ACL feature now provides more security by immediately dropping
340 * connections from untrusted hosts.
342 * Revision 1.6 2001/05/26 00:28:36 jongfoster
343 * Automatic reloading of config file.
344 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
345 * Most of the global variables have been moved to a new
346 * struct configuration_spec, accessed through csp->config->globalname
347 * Most of the globals remaining are used by the Win32 GUI.
349 * Revision 1.5 2001/05/25 22:34:30 jongfoster
352 * Revision 1.4 2001/05/22 18:46:04 oes
354 * - Enabled filtering banners by size rather than URL
355 * by adding patterns that replace all standard banner
356 * sizes with the "Junkbuster" gif to the re_filterfile
358 * - Enabled filtering WebBugs by providing a pattern
359 * which kills all 1x1 images
361 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
362 * which is selected by the (nonstandard and therefore
363 * capital) letter 'U' in the option string.
364 * It causes the quantifiers to be ungreedy by default.
365 * Appending a ? turns back to greedy (!).
367 * - Added a new interceptor ijb-send-banner, which
368 * sends back the "Junkbuster" gif. Without imagelist or
369 * MSIE detection support, or if tinygif = 1, or the
370 * URL isn't recognized as an imageurl, a lame HTML
371 * explanation is sent instead.
373 * - Added new feature, which permits blocking remote
374 * script redirects and firing back a local redirect
376 * The feature is conditionally compiled, i.e. it
377 * can be disabled with --disable-fast-redirects,
378 * plus it must be activated by a "fast-redirects"
379 * line in the config file, has its own log level
380 * and of course wants to be displayed by show-proxy-args
381 * Note: Boy, all the #ifdefs in 1001 locations and
382 * all the fumbling with configure.in and acconfig.h
383 * were *way* more work than the feature itself :-(
385 * - Because a generic redirect template was needed for
386 * this, tinygif = 3 now uses the same.
388 * - Moved GIFs, and other static HTTP response templates
393 * - Removed some >400 CRs again (Jon, you really worked
396 * Revision 1.3 2001/05/20 16:44:47 jongfoster
397 * Removing last hardcoded Junkbusters.com URLs.
399 * Revision 1.2 2001/05/20 01:21:20 jongfoster
400 * Version 2.9.4 checkin.
401 * - Merged popupfile and cookiefile, and added control over PCRS
402 * filtering, in new "permissionsfile".
403 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
404 * file error you now get a message box (in the Win32 GUI) rather
405 * than the program exiting with no explanation.
406 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
408 * - Removed tabs from "config"
409 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
410 * - Bumped up version number.
412 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
413 * Initial import of version 2.9.3 source tree
416 *********************************************************************/
422 #include <sys/types.h>
431 #endif /* ndef __OS2__ */
432 #include <netinet/in.h>
434 #include <winsock2.h>
435 #endif /* ndef _WIN32 */
439 #endif /* def __OS2__ */
447 #include "jbsockets.h"
448 #include "miscutil.h"
452 #include "deanimate.h"
453 #include "urlmatch.h"
454 #ifdef FEATURE_ACTIVITY_CONSOLE
456 #endif /* def FEATURE_ACTIVITY_CONSOLE */
461 const char filters_h_rcs[] = FILTERS_H_VERSION;
463 /* Fix a problem with Solaris. There should be no effect on other
465 * Solaris's isspace() is a macro which uses it's argument directly
466 * as an array index. Therefore we need to make sure that high-bit
467 * characters generate +ve values, and ideally we also want to make
468 * the argument match the declared parameter type of "int".
470 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
474 /*********************************************************************
476 * Function : block_acl
478 * Description : Block this request?
479 * Decide yes or no based on ACL file.
482 * 1 : dst = The proxy or gateway address this is going to.
483 * Or NULL to check all possible targets.
484 * 2 : csp = Current client state (buffers, headers, etc...)
485 * Also includes the client IP address.
487 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
489 *********************************************************************/
490 int block_acl(struct access_control_addr *dst, struct client_state *csp)
492 struct access_control_list *acl = csp->config->acl;
494 /* if not using an access control list, then permit the connection */
500 /* search the list */
503 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
507 /* Just want to check if they have any access */
508 if (acl->action == ACL_PERMIT)
513 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
514 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
516 if (acl->action == ACL_PERMIT)
522 #ifdef FEATURE_ACTIVITY_CONSOLE
523 accumulate_stats(STATS_ACL_RESTRICT, 1);
524 #endif /* def FEATURE_ACTIVITY_CONSOLE */
537 /*********************************************************************
539 * Function : acl_addr
541 * Description : Called from `load_config' to parse an ACL address.
544 * 1 : aspec = String specifying ACL address.
545 * 2 : aca = struct access_control_addr to fill in.
547 * Returns : 0 => Ok, everything else is an error.
549 *********************************************************************/
550 int acl_addr(char *aspec, struct access_control_addr *aca)
552 int i, masklength, port;
558 if ((p = strchr(aspec, '/')) != NULL)
562 if (ijb_isdigit(*p) == 0)
566 masklength = atoi(p);
569 if ((masklength < 0) || (masklength > 32))
574 if ((p = strchr(aspec, ':')) != NULL)
578 if (ijb_isdigit(*p) == 0)
587 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
589 if (aca->addr == INADDR_NONE)
594 /* build the netmask */
596 for (i=1; i <= masklength ; i++)
598 aca->mask |= (1 << (32 - i));
601 /* now mask off the host portion of the ip address
602 * (i.e. save on the network portion of the address).
604 aca->addr = aca->addr & aca->mask;
609 #endif /* def FEATURE_ACL */
612 /*********************************************************************
614 * Function : match_portlist
616 * Description : Check if a given number is covered by a comma
617 * separated list of numbers and ranges (a,b-c,d,..)
620 * 1 : portlist = String with list
621 * 2 : port = port to check
623 * Returns : 0 => no match
626 *********************************************************************/
627 int match_portlist(const char *portlist, int port)
629 char *min, *max, *next, *portlist_copy;
631 min = next = portlist_copy = strdup(portlist);
634 * Zero-terminate first item and remember offset for next
636 if (NULL != (next = strchr(portlist_copy, (int) ',')))
642 * Loop through all items, checking for match
646 if (NULL == (max = strchr(min, (int) '-')))
649 * No dash, check for equality
651 if (port == atoi(min))
660 * This is a range, so check if between min and max,
661 * or, if max was omitted, between min and 65K
664 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
678 * Zero-terminate next item and remember offset for n+1
680 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
692 /*********************************************************************
694 * Function : block_url
696 * Description : Called from `chat'. Check to see if we need to block this.
699 * 1 : csp = Current client state (buffers, headers, etc...)
701 * Returns : NULL => unblocked, else HTTP block response
703 *********************************************************************/
704 struct http_response *block_url(struct client_state *csp)
706 #ifdef FEATURE_IMAGE_BLOCKING
708 #endif /* def FEATURE_IMAGE_BLOCKING */
709 struct http_response *rsp;
712 * If it's not blocked, don't block it ;-)
714 if ((csp->action->flags & ACTION_BLOCK) == 0)
720 * Else, prepare a response
722 if (NULL == (rsp = alloc_http_response()))
724 return cgi_error_memory();
728 * If it's an image-url, send back an image or redirect
729 * as specified by the relevant +image action
731 #ifdef FEATURE_IMAGE_BLOCKING
732 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
735 /* determine HOW images should be blocked */
736 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
738 #if 1 /* Two alternative strategies, use this one for now: */
740 /* and handle accordingly: */
741 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
743 #ifdef FEATURE_ACTIVITY_CONSOLE
744 accumulate_stats(STATS_IMAGE_BLOCK, 1);
745 #endif /* def FEATURE_ACTIVITY_CONSOLE */
746 rsp->body = bindup(image_pattern_data, image_pattern_length);
747 if (rsp->body == NULL)
749 free_http_response(rsp);
750 return cgi_error_memory();
752 rsp->content_length = image_pattern_length;
754 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
756 free_http_response(rsp);
757 return cgi_error_memory();
761 else if (0 == strcmpic(p, "blank"))
763 #ifdef FEATURE_ACTIVITY_CONSOLE
764 accumulate_stats(STATS_IMAGE_BLOCK, 1);
765 #endif /* def FEATURE_ACTIVITY_CONSOLE */
766 rsp->body = bindup(image_blank_data, image_blank_length);
767 if (rsp->body == NULL)
769 free_http_response(rsp);
770 return cgi_error_memory();
772 rsp->content_length = image_blank_length;
774 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
776 free_http_response(rsp);
777 return cgi_error_memory();
783 rsp->status = strdup("302 Local Redirect from Privoxy");
784 if (rsp->status == NULL)
786 free_http_response(rsp);
787 return cgi_error_memory();
790 if (enlist_unique_header(rsp->headers, "Location", p))
792 free_http_response(rsp);
793 return cgi_error_memory();
797 #else /* Following code is disabled for now */
799 /* and handle accordingly: */
800 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
802 p = CGI_PREFIX "send-banner?type=pattern";
804 else if (0 == strcmpic(p, "blank"))
806 p = CGI_PREFIX "send-banner?type=blank";
808 rsp->status = strdup("302 Local Redirect from Privoxy");
809 if (rsp->status == NULL)
811 free_http_response(rsp);
812 return cgi_error_memory();
815 if (enlist_unique_header(rsp->headers, "Location", p))
817 free_http_response(rsp);
818 return cgi_error_memory();
820 #endif /* Preceeding code is disabled for now */
823 #endif /* def FEATURE_IMAGE_BLOCKING */
826 * Else, generate an HTML "blocked" message:
830 struct map * exports;
833 * Workaround for stupid Netscape bug which prevents
834 * pages from being displayed if loading a referenced
835 * JavaScript or style sheet fails. So make it appear
836 * as if it succeeded.
838 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
839 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
840 && !strstr(p, "Gecko") /* save Mozilla, */
841 && !strstr(p, "compatible") /* MSIE */
842 && !strstr(p, "Opera")) /* and Opera. */
844 rsp->status = strdup("200 Request for blocked URL");
848 rsp->status = strdup("404 Request for blocked URL");
851 if (rsp->status == NULL)
853 free_http_response(rsp);
854 return cgi_error_memory();
857 exports = default_exports(csp, NULL);
860 free_http_response(rsp);
861 return cgi_error_memory();
864 #ifdef FEATURE_FORCE_LOAD
865 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
866 if (csp->http->ssl != 0)
867 #endif /* ndef FEATURE_FORCE_LOAD */
869 err = map_block_killer(exports, "force-support");
872 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
873 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
874 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
879 free_http_response(rsp);
880 return cgi_error_memory();
883 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
886 free_http_response(rsp);
887 return cgi_error_memory();
891 return finish_http_response(rsp);
897 /*********************************************************************
899 * Function : trust_url FIXME: I should be called distrust_url
901 * Description : Calls is_untrusted_url to determine if the URL is trusted
902 * and if not, returns a HTTP 304 response with a reject message.
905 * 1 : csp = Current client state (buffers, headers, etc...)
907 * Returns : NULL => trusted, else http_response.
909 *********************************************************************/
910 struct http_response *trust_url(struct client_state *csp)
912 struct http_response *rsp;
913 struct map * exports;
914 char buf[BUFFER_SIZE];
916 struct url_spec **tl;
921 * Don't bother to work on trusted URLs
923 if (!is_untrusted_url(csp))
929 * Else, prepare a response:
931 if (NULL == (rsp = alloc_http_response()))
933 return cgi_error_memory();
936 exports = default_exports(csp, NULL);
939 free_http_response(rsp);
940 return cgi_error_memory();
944 * Export the protocol, host, port, and referrer information
946 err = map(exports, "hostport", 1, csp->http->hostport, 1);
947 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
948 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
950 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
952 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
956 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
962 free_http_response(rsp);
963 return cgi_error_memory();
967 * Export the trust list
970 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
972 sprintf(buf, "<li>%s</li>\n", t->spec);
973 string_append(&p, buf);
975 err = map(exports, "trusted-referrers", 1, p, 0);
980 free_http_response(rsp);
981 return cgi_error_memory();
985 * Export the trust info, if available
987 if (csp->config->trust_info->first)
989 struct list_entry *l;
992 for (l = csp->config->trust_info->first; l ; l = l->next)
994 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
995 string_append(&p, buf);
997 err = map(exports, "trust-info", 1, p, 0);
1001 err = map_block_killer(exports, "have-trust-info");
1007 free_http_response(rsp);
1008 return cgi_error_memory();
1012 * Export the force prefix or the force conditional block killer
1014 #ifdef FEATURE_FORCE_LOAD
1015 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1016 #else /* ifndef FEATURE_FORCE_LOAD */
1017 err = map_block_killer(exports, "force-support");
1018 #endif /* ndef FEATURE_FORCE_LOAD */
1023 free_http_response(rsp);
1024 return cgi_error_memory();
1028 * Build the response
1030 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1033 free_http_response(rsp);
1034 return cgi_error_memory();
1037 return finish_http_response(rsp);
1039 #endif /* def FEATURE_TRUST */
1042 #ifdef FEATURE_FAST_REDIRECTS
1043 /*********************************************************************
1045 * Function : redirect_url
1047 * Description : Checks for redirection URLs and returns a HTTP redirect
1048 * to the destination URL, if necessary
1051 * 1 : csp = Current client state (buffers, headers, etc...)
1053 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1055 *********************************************************************/
1056 struct http_response *redirect_url(struct client_state *csp)
1059 struct http_response *rsp;
1061 p = q = csp->http->path;
1062 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1065 * find the last URL encoded in the request
1067 while ((p = strstr(p, "http://")) != NULL)
1073 * if there was any, generate and return a HTTP redirect
1075 if (q != csp->http->path)
1077 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1079 if (NULL == (rsp = alloc_http_response()))
1081 return cgi_error_memory();
1084 if ( enlist_unique_header(rsp->headers, "Location", q)
1085 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1087 free_http_response(rsp);
1088 return cgi_error_memory();
1091 return finish_http_response(rsp);
1099 #endif /* def FEATURE_FAST_REDIRECTS */
1102 #ifdef FEATURE_IMAGE_BLOCKING
1103 /*********************************************************************
1105 * Function : is_imageurl
1107 * Description : Given a URL, decide whether it is an image or not,
1108 * using either the info from a previous +image action
1109 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1110 * is MSIE and not on a Mac, tell from the browser's accept
1114 * 1 : csp = Current client state (buffers, headers, etc...)
1116 * Returns : True (nonzero) if URL is an image, false (0)
1119 *********************************************************************/
1120 int is_imageurl(struct client_state *csp)
1122 #ifdef FEATURE_IMAGE_DETECT_MSIE
1125 tmp = get_header_value(csp->headers, "User-Agent:");
1126 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1128 tmp = get_header_value(csp->headers, "Accept:");
1129 if (tmp && strstr(tmp, "image/gif"))
1131 /* Client will accept HTML. If this seems counterintuitive,
1141 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1143 return ((csp->action->flags & ACTION_IMAGE) != 0);
1146 #endif /* def FEATURE_IMAGE_BLOCKING */
1149 #ifdef FEATURE_TRUST
1150 /*********************************************************************
1152 * Function : is_untrusted_url
1154 * Description : Should we "distrust" this URL (and block it)?
1156 * Yes if it matches a line in the trustfile, or if the
1157 * referrer matches a line starting with "+" in the
1162 * 1 : csp = Current client state (buffers, headers, etc...)
1164 * Returns : 0 => trusted, 1 => untrusted
1166 *********************************************************************/
1167 int is_untrusted_url(struct client_state *csp)
1169 struct file_list *fl;
1170 struct block_spec *b;
1171 struct url_spec **trusted_url;
1172 struct http_request rhttp[1];
1173 const char * referer;
1177 * If we don't have a trustlist, we trust everybody
1179 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1184 memset(rhttp, '\0', sizeof(*rhttp));
1187 * Do we trust the request URL itself?
1189 for (b = b->next; b ; b = b->next)
1191 if (url_match(b->url, csp->http))
1197 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1199 /* no referrer was supplied */
1205 * If not, do we maybe trust its referrer?
1207 err = parse_http_url(referer, rhttp, csp);
1213 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1215 if (url_match(*trusted_url, rhttp))
1217 /* if the URL's referrer is from a trusted referrer, then
1218 * add the target spec to the trustfile as an unblocked
1219 * domain and return NULL (which means it's OK).
1224 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1228 char * new_entry = strdup("~");
1230 string_append(&new_entry, csp->http->hostport);
1232 path = csp->http->path;
1233 if ( (path[0] == '/')
1235 && ((path_end = strchr(path + 2, '/')) != NULL))
1237 /* since this path points into a user's home space
1238 * be sure to include this spec in the trustfile.
1240 int path_len = path_end - path; /* save offset */
1241 path = strdup(path); /* Copy string */
1244 path_end = path + path_len; /* regenerate ptr to new buffer */
1245 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1247 string_join(&new_entry, path);
1250 if (new_entry != NULL)
1252 fprintf(fp, "%s\n", new_entry);
1257 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1258 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1268 #endif /* def FEATURE_TRUST */
1271 /*********************************************************************
1273 * Function : pcrs_filter_response
1275 * Description : Ecexute all text substitutions from all applying
1276 * +filter actions on the text buffer that's been accumulated
1277 * in csp->iob->buf. If this changes the contents, set
1278 * csp->content_length to the modified size and raise the
1279 * CSP_FLAG_MODIFIED flag.
1282 * 1 : csp = Current client state (buffers, headers, etc...)
1284 * Returns : a pointer to the (newly allocated) modified buffer.
1285 * or NULL if there were no hits or something went wrong
1287 *********************************************************************/
1288 char *pcrs_filter_response(struct client_state *csp)
1293 char *old = csp->iob->cur, *new = NULL;
1296 struct file_list *fl;
1297 struct re_filterfile_spec *b;
1298 struct list_entry *filtername;
1303 if (csp->iob->cur >= csp->iob->eod)
1307 size = csp->iob->eod - csp->iob->cur;
1309 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1311 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1316 * If the body has a "chunked" transfer-encoding,
1317 * get rid of it first, adjusting size and iob->eod
1319 if (csp->flags & CSP_FLAG_CHUNKED)
1321 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1322 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1326 csp->iob->eod = csp->iob->cur + size;
1327 csp->flags |= CSP_FLAG_MODIFIED;
1331 * For all applying +filter actions, look if a filter by that
1332 * name exists and if yes, execute it's pcrs_joblist on the
1335 for (b = fl->f; b; b = b->next)
1337 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1338 filtername ; filtername = filtername->next)
1340 if (strcmp(b->name, filtername->str) == 0)
1342 int current_hits = 0;
1344 if ( NULL == b->joblist )
1346 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1350 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1351 csp->http->hostport, csp->http->path, size, b->name);
1353 /* Apply all jobs from the joblist */
1354 for (job = b->joblist; NULL != job; job = job->next)
1356 current_hits += pcrs_execute(job, old, size, &new, &size);
1357 if (old != csp->iob->cur) free(old);
1361 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1362 hits += current_hits;
1368 * If there were no hits, destroy our copy and let
1369 * chat() use the original in csp->iob
1376 #ifdef FEATURE_ACTIVITY_CONSOLE
1378 accumulate_stats(STATS_FILTER, hits);
1379 #endif /* def FEATURE_ACTIVITY_CONSOLE */
1381 csp->flags |= CSP_FLAG_MODIFIED;
1382 csp->content_length = size;
1390 /*********************************************************************
1392 * Function : gif_deanimate_response
1394 * Description : Deanimate the GIF image that has been accumulated in
1395 * csp->iob->buf, set csp->content_length to the modified
1396 * size and raise the CSP_FLAG_MODIFIED flag.
1399 * 1 : csp = Current client state (buffers, headers, etc...)
1401 * Returns : a pointer to the (newly allocated) modified buffer.
1402 * or NULL in case something went wrong.
1404 *********************************************************************/
1405 char *gif_deanimate_response(struct client_state *csp)
1407 struct binbuffer *in, *out;
1409 size_t size = csp->iob->eod - csp->iob->cur;
1412 * If the body has a "chunked" transfer-encoding,
1413 * get rid of it first, adjusting size and iob->eod
1415 if (csp->flags & CSP_FLAG_CHUNKED)
1417 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1418 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1422 csp->iob->eod = csp->iob->cur + size;
1423 csp->flags |= CSP_FLAG_MODIFIED;
1426 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1427 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1429 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1433 in->buffer = csp->iob->cur;
1436 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1438 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1445 if ((int)size == out->offset)
1447 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1451 #ifdef FEATURE_ACTIVITY_CONSOLE
1452 accumulate_stats(STATS_GIF_DEANIMATE, 1);
1453 #endif /* def FEATURE_ACTIVITY_CONSOLE */
1454 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1456 csp->content_length = out->offset;
1457 csp->flags |= CSP_FLAG_MODIFIED;
1467 /*********************************************************************
1469 * Function : remove_chunked_transfer_coding
1471 * Description : In-situ remove the "chunked" transfer coding as defined
1472 * in rfc2616 from a buffer.
1475 * 1 : buffer = Pointer to the text buffer
1476 * 2 : size = Number of bytes to be processed
1478 * Returns : The new size, i.e. the number of bytes from buffer which
1479 * are occupied by the stripped body, or 0 in case something
1482 *********************************************************************/
1483 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1486 unsigned int chunksize = 0;
1487 char *from_p, *to_p;
1490 from_p = to_p = buffer;
1492 if (sscanf(buffer, "%x", &chunksize) != 1)
1494 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1498 while (chunksize > 0)
1500 if (NULL == (from_p = strstr(from_p, "\r\n")))
1502 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1506 if ((newsize += chunksize) >= size)
1508 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
1513 memmove(to_p, from_p, (size_t) chunksize);
1514 to_p = buffer + newsize;
1515 from_p += chunksize + 2;
1517 if (sscanf(from_p, "%x", &chunksize) != 1)
1519 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1524 /* FIXME: Should this get its own loglevel? */
1525 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1531 /*********************************************************************
1533 * Function : url_actions
1535 * Description : Gets the actions for this URL.
1538 * 1 : http = http_request request for blocked URLs
1539 * 2 : csp = Current client state (buffers, headers, etc...)
1543 *********************************************************************/
1544 void url_actions(struct http_request *http,
1545 struct client_state *csp)
1547 struct file_list *fl;
1548 struct url_actions *b;
1551 init_current_action(csp->action);
1553 for (i = 0; i < MAX_ACTION_FILES; i++)
1555 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1560 apply_url_actions(csp->action, http, b);
1567 /*********************************************************************
1569 * Function : apply_url_actions
1571 * Description : Applies a list of URL actions.
1574 * 1 : action = Destination.
1575 * 2 : http = Current URL
1576 * 3 : b = list of URL actions to apply
1580 *********************************************************************/
1581 void apply_url_actions(struct current_action_spec *action,
1582 struct http_request *http,
1583 struct url_actions *b)
1587 /* Should never happen */
1591 for (b = b->next; NULL != b; b = b->next)
1593 if (url_match(b->url, http))
1595 merge_current_action(action, b->action);
1601 /*********************************************************************
1603 * Function : forward_url
1605 * Description : Should we forward this to another proxy?
1608 * 1 : http = http_request request for current URL
1609 * 2 : csp = Current client state (buffers, headers, etc...)
1611 * Returns : Pointer to forwarding information.
1613 *********************************************************************/
1614 const struct forward_spec * forward_url(struct http_request *http,
1615 struct client_state *csp)
1617 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1618 struct forward_spec *fwd = csp->config->forward;
1627 if (url_match(fwd->url, http))