1 const char filters_rcs[] = "$Id: filters.c,v 1.58.2.9 2006/01/29 23:10:56 david__schmidt Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/Attic/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', `trust_url', `gif_deanimate_response',
13 * `jpeg_inspect_response'
15 * Copyright : Written by and Copyright (C) 2001, 2004 the SourceForge
16 * Privoxy team. http://www.privoxy.org/
18 * Based on the Internet Junkbuster originally written
19 * by and Copyright (C) 1997 Anonymous Coders and
20 * Junkbusters Corporation. http://www.junkbusters.com
22 * This program is free software; you can redistribute it
23 * and/or modify it under the terms of the GNU General
24 * Public License as published by the Free Software
25 * Foundation; either version 2 of the License, or (at
26 * your option) any later version.
28 * This program is distributed in the hope that it will
29 * be useful, but WITHOUT ANY WARRANTY; without even the
30 * implied warranty of MERCHANTABILITY or FITNESS FOR A
31 * PARTICULAR PURPOSE. See the GNU General Public
32 * License for more details.
34 * The GNU General Public License should be included with
35 * this file. If not, you can view it at
36 * http://www.gnu.org/copyleft/gpl.html
37 * or write to the Free Software Foundation, Inc., 59
38 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
42 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
43 * Multiple filter file support
45 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
46 * A few memory leaks plugged (mostly on error paths)
48 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
49 * Add the ability to check jpeg images for invalid
50 * lengths of comment blocks. Defensive strategy
51 * against the exploit:
52 * Microsoft Security Bulletin MS04-028
53 * Buffer Overrun in JPEG Processing (GDI+) Could
54 * Allow Code Execution (833987)
55 * Enabled with +inspect-jpegs in actions files.
57 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
58 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
60 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
61 * Fixed bug #839859: "See why" link URL now gets url-encoded.
63 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
66 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
67 * Added basic support for OPTIONS and TRACE HTTP methods:
68 * New function direct_response which handles OPTIONS and
69 * TRACE requests whose Max-Forwards header field is zero.
71 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
72 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
74 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
75 * - Bugfix: Executing a filters without jobs no longer results in
76 * turing off *all* filters.
77 * - Security fix: Malicious web servers can't cause a seg fault
78 * through bogus chunk sizes anymore
80 * Revision 1.58 2002/04/24 02:11:17 oes
81 * Jon's multiple AF patch: url_actions now evaluates rules
84 * Revision 1.57 2002/04/08 20:38:34 swa
87 * Revision 1.56 2002/04/05 15:51:24 oes
88 * - bugfix: error-pages now get correct request protocol
89 * - fix for invalid HTML in trust info
91 * Revision 1.55 2002/04/02 16:13:51 oes
92 * Fix: No "Go there anyway" for SSL
94 * Revision 1.54 2002/04/02 14:55:56 oes
95 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
97 * Revision 1.53 2002/03/26 22:29:54 swa
98 * we have a new homepage!
100 * Revision 1.52 2002/03/24 16:35:57 jongfoster
103 * Revision 1.51 2002/03/24 15:23:33 jongfoster
106 * Revision 1.50 2002/03/24 13:25:43 swa
107 * name change related issues
109 * Revision 1.49 2002/03/16 20:29:14 oes
112 * Revision 1.48 2002/03/13 20:25:34 oes
113 * Better logging for content filters
115 * Revision 1.47 2002/03/13 00:30:52 jongfoster
117 * Added option of always sending redirect for imageblock,
118 * currently disabled with #if 0.
120 * Revision 1.46 2002/03/12 01:42:49 oes
121 * Introduced modular filters
123 * Revision 1.45 2002/03/08 16:47:50 oes
124 * Added choice beween GIF and PNG built-in images
126 * Revision 1.44 2002/03/07 03:49:31 oes
127 * - Fixed compiler warnings etc
128 * - Changed built-in images from GIF to PNG
129 * (with regard to Unisys patent issue)
130 * - Added a 4x4 pattern PNG which is less intrusive
131 * than the logo but also clearly marks the deleted banners
133 * Revision 1.43 2002/01/22 23:51:59 jongfoster
134 * Replacing strsav() with the safer string_append().
136 * Adding missing html_encode() to error message generators. Where encoded
137 * and unencoded versions of a string were provided, removing the unencoded
140 * Revision 1.42 2002/01/17 21:00:32 jongfoster
141 * Moving all our URL and URL pattern parsing code to urlmatch.c.
143 * Using a single, simple url_match(pattern,url) function - rather than
144 * the 3-line match routine which was repeated all over the place.
146 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
148 * Using parse_http_url() to parse URLs without faking a HTTP
149 * request line for parse_http_request().
151 * Revision 1.41 2001/11/13 00:14:07 jongfoster
152 * Fixing stupid bug now I've figured out what || means.
153 * (It always returns 0 or 1, not one of it's paramaters.)
155 * Revision 1.40 2001/10/26 17:37:55 oes
156 * - Re-enabled Netscape 200/404 bug workaround in block_url():
157 * - Removed OS/2 special case
158 * - Made block_url() independant from sed() having been run
159 * - Made trust_url independant from sed() having been run
160 * - Made is_imageurl independant from sed() having been run.
161 * It now checks User-Agent: and Accept: by itself.
164 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
165 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
166 * threads to call select() simultaneously. So, it's time to do a real, live,
167 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
168 * (native). Both versions will work, but using __OS2__ offers multi-threading.
170 * Revision 1.38 2001/10/23 21:32:33 jongfoster
171 * Adding error-checking to selected functions
173 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
174 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
175 * filters.c. Added a FIXME in front of the offending code. I'll gladly
176 * put in a better/more robust fix for all parties if one is presented...
177 * It seems that just returning 200 instead of 404 would pretty much fix
178 * it for everyone, but I don't know all the history of the problem.
180 * Revision 1.36 2001/10/10 16:44:16 oes
181 * Added match_portlist function
183 * Revision 1.35 2001/10/07 15:41:23 oes
184 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
186 * New function remove_chunked_transfer_coding that strips chunked
187 * transfer coding to plain and is called by pcrs_filter_response
188 * and gif_deanimate_response if neccessary
190 * Improved handling of zero-change re_filter runs
192 * pcrs_filter_response and gif_deanimate_response now remove
193 * chunked transfer codeing before processing the body.
195 * Revision 1.34 2001/09/20 15:49:36 steudten
197 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
198 * See cgi.c fill_template().
200 * Revision 1.33 2001/09/16 17:05:14 jongfoster
201 * Removing unused #include showarg.h
203 * Revision 1.32 2001/09/16 13:21:27 jongfoster
204 * Changes to use new list functions.
206 * Revision 1.31 2001/09/16 11:38:02 jongfoster
207 * Splitting fill_template() into 2 functions:
208 * template_load() loads the file
209 * template_fill() performs the PCRS regexps.
210 * This is because the CGI edit interface has a "table row"
211 * template which is used many times in the page - this
212 * change means it's only loaded from disk once.
214 * Revision 1.30 2001/09/16 11:00:10 jongfoster
215 * New function alloc_http_response, for symmetry with free_http_response
217 * Revision 1.29 2001/09/13 23:32:40 jongfoster
218 * Moving image data to cgi.c rather than cgi.h
219 * Fixing a GPF under Win32 (and any other OS that protects global
220 * constants from being written to).
222 * Revision 1.28 2001/09/10 10:18:51 oes
223 * Silenced compiler warnings
225 * Revision 1.27 2001/08/05 16:06:20 jongfoster
226 * Modifiying "struct map" so that there are now separate header and
227 * "map_entry" structures. This means that functions which modify a
228 * map no longer need to return a pointer to the modified map.
229 * Also, it no longer reverses the order of the entries (which may be
230 * important with some advanced template substitutions).
232 * Revision 1.26 2001/07/30 22:08:36 jongfoster
233 * Tidying up #defines:
234 * - All feature #defines are now of the form FEATURE_xxx
235 * - Permanently turned off WIN_GUI_EDIT
236 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
238 * Revision 1.25 2001/07/26 10:09:46 oes
239 * Made browser detection a little less naive
241 * Revision 1.24 2001/07/25 17:22:51 oes
242 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
244 * Revision 1.23 2001/07/23 13:40:12 oes
245 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
247 * Revision 1.22 2001/07/18 12:29:34 oes
248 * - Made gif_deanimate_response respect
249 * csp->action->string[ACTION_STRING_DEANIMATE]
250 * - Logging cosmetics
252 * Revision 1.21 2001/07/13 13:59:53 oes
253 * - Introduced gif_deanimate_response which shares the
254 * generic content modification interface of pcrs_filter_response
255 * and acts as a wrapper to deanimate.c:gif_deanimate()
256 * - Renamed re_process_buffer to pcrs_filter_response
257 * - pcrs_filter_response now returns NULL on failiure
258 * - Removed all #ifdef PCRS
260 * Revision 1.20 2001/07/01 17:01:04 oes
261 * Added comments and missing return statement in is_untrusted_url()
263 * Revision 1.19 2001/06/29 21:45:41 oes
264 * Indentation, CRLF->LF, Tab-> Space
266 * Revision 1.18 2001/06/29 13:27:38 oes
267 * - Cleaned up, renamed and reorderd functions
268 * and improved comments
271 * - Ported to CGI platform. Now delivers
272 * http_response or NULL
273 * - Unified HTML and GIF generation (moved image detection
274 * and GIF generation here from jcc.c:chat())
275 * - Fixed HTTP status to:
276 * - 403 (Forbidden) for the "blocked" HTML message
277 * - 200 (OK) for GIF answers
278 * - 302 (Redirect) for redirect to GIF
281 * - Ported to CGI platform. Now delivers
282 * http_response or NULL
283 * - Separated detection of untrusted URL into
284 * (bool)is_untrusted_url
285 * - Added enforcement of untrusted requests
287 * - Moved redirect_url() from cgi.c to here
288 * and ported it to the CGI platform
290 * - Removed logentry from cancelled commit
292 * Revision 1.17 2001/06/09 10:55:28 jongfoster
293 * Changing BUFSIZ ==> BUFFER_SIZE
295 * Revision 1.16 2001/06/07 23:10:26 jongfoster
296 * Allowing unanchored domain patterns to back off and retry
297 * if they partially match. Optimized right-anchored patterns.
298 * Moving ACL and forward files into config file.
299 * Replacing struct gateway with struct forward_spec
301 * Revision 1.15 2001/06/03 19:12:00 oes
302 * extracted-CGI relevant stuff
304 * Revision 1.14 2001/06/01 10:30:55 oes
305 * Added optional left-anchoring to domaincmp
307 * Revision 1.13 2001/05/31 21:21:30 jongfoster
308 * Permissionsfile / actions file changes:
309 * - Changed "permission" to "action" throughout
310 * - changes to file format to allow string parameters
311 * - Moved helper functions to actions.c
313 * Revision 1.12 2001/05/31 17:35:20 oes
315 * - Enhanced domain part globbing with infix and prefix asterisk
316 * matching and optional unanchored operation
318 * Revision 1.11 2001/05/29 11:53:23 oes
319 * "See why" link added to "blocked" page
321 * Revision 1.10 2001/05/29 09:50:24 jongfoster
322 * Unified blocklist/imagelist/permissionslist.
323 * File format is still under discussion, but the internal changes
326 * Also modified interceptor behaviour:
327 * - We now intercept all URLs beginning with one of the following
328 * prefixes (and *only* these prefixes):
330 * * http://ijbswa.sf.net/config/
331 * * http://ijbswa.sourceforge.net/config/
332 * - New interceptors "home page" - go to http://i.j.b/ to see it.
333 * - Internal changes so that intercepted and fast redirect pages
334 * are not replaced with an image.
335 * - Interceptors now have the option to send a binary page direct
336 * to the client. (i.e. ijb-send-banner uses this)
337 * - Implemented show-url-info interceptor. (Which is why I needed
338 * the above interceptors changes - a typical URL is
339 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
340 * The previous mechanism would not have intercepted that, and
341 * if it had been intercepted then it then it would have replaced
344 * Revision 1.9 2001/05/27 22:17:04 oes
346 * - re_process_buffer no longer writes the modified buffer
347 * to the client, which was very ugly. It now returns the
348 * buffer, which it is then written by chat.
350 * - content_length now adjusts the Content-Length: header
351 * for modified documents rather than crunch()ing it.
352 * (Length info in csp->content_length, which is 0 for
353 * unmodified documents)
355 * - For this to work, sed() is called twice when filtering.
357 * Revision 1.8 2001/05/26 17:13:28 jongfoster
358 * Filled in a function comment.
360 * Revision 1.7 2001/05/26 15:26:15 jongfoster
361 * ACL feature now provides more security by immediately dropping
362 * connections from untrusted hosts.
364 * Revision 1.6 2001/05/26 00:28:36 jongfoster
365 * Automatic reloading of config file.
366 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
367 * Most of the global variables have been moved to a new
368 * struct configuration_spec, accessed through csp->config->globalname
369 * Most of the globals remaining are used by the Win32 GUI.
371 * Revision 1.5 2001/05/25 22:34:30 jongfoster
374 * Revision 1.4 2001/05/22 18:46:04 oes
376 * - Enabled filtering banners by size rather than URL
377 * by adding patterns that replace all standard banner
378 * sizes with the "Junkbuster" gif to the re_filterfile
380 * - Enabled filtering WebBugs by providing a pattern
381 * which kills all 1x1 images
383 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
384 * which is selected by the (nonstandard and therefore
385 * capital) letter 'U' in the option string.
386 * It causes the quantifiers to be ungreedy by default.
387 * Appending a ? turns back to greedy (!).
389 * - Added a new interceptor ijb-send-banner, which
390 * sends back the "Junkbuster" gif. Without imagelist or
391 * MSIE detection support, or if tinygif = 1, or the
392 * URL isn't recognized as an imageurl, a lame HTML
393 * explanation is sent instead.
395 * - Added new feature, which permits blocking remote
396 * script redirects and firing back a local redirect
398 * The feature is conditionally compiled, i.e. it
399 * can be disabled with --disable-fast-redirects,
400 * plus it must be activated by a "fast-redirects"
401 * line in the config file, has its own log level
402 * and of course wants to be displayed by show-proxy-args
403 * Note: Boy, all the #ifdefs in 1001 locations and
404 * all the fumbling with configure.in and acconfig.h
405 * were *way* more work than the feature itself :-(
407 * - Because a generic redirect template was needed for
408 * this, tinygif = 3 now uses the same.
410 * - Moved GIFs, and other static HTTP response templates
415 * - Removed some >400 CRs again (Jon, you really worked
418 * Revision 1.3 2001/05/20 16:44:47 jongfoster
419 * Removing last hardcoded Junkbusters.com URLs.
421 * Revision 1.2 2001/05/20 01:21:20 jongfoster
422 * Version 2.9.4 checkin.
423 * - Merged popupfile and cookiefile, and added control over PCRS
424 * filtering, in new "permissionsfile".
425 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
426 * file error you now get a message box (in the Win32 GUI) rather
427 * than the program exiting with no explanation.
428 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
430 * - Removed tabs from "config"
431 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
432 * - Bumped up version number.
434 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
435 * Initial import of version 2.9.3 source tree
438 *********************************************************************/
444 #include <sys/types.h>
453 #endif /* ndef __OS2__ */
454 #include <netinet/in.h>
456 #include <winsock2.h>
457 #endif /* ndef _WIN32 */
461 #endif /* def __OS2__ */
469 #include "jbsockets.h"
470 #include "miscutil.h"
474 #include "deanimate.h"
475 #include "urlmatch.h"
481 const char filters_h_rcs[] = FILTERS_H_VERSION;
483 /* Fix a problem with Solaris. There should be no effect on other
485 * Solaris's isspace() is a macro which uses it's argument directly
486 * as an array index. Therefore we need to make sure that high-bit
487 * characters generate +ve values, and ideally we also want to make
488 * the argument match the declared parameter type of "int".
490 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
494 /*********************************************************************
496 * Function : block_acl
498 * Description : Block this request?
499 * Decide yes or no based on ACL file.
502 * 1 : dst = The proxy or gateway address this is going to.
503 * Or NULL to check all possible targets.
504 * 2 : csp = Current client state (buffers, headers, etc...)
505 * Also includes the client IP address.
507 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
509 *********************************************************************/
510 int block_acl(struct access_control_addr *dst, struct client_state *csp)
512 struct access_control_list *acl = csp->config->acl;
514 /* if not using an access control list, then permit the connection */
520 /* search the list */
523 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
527 /* Just want to check if they have any access */
528 if (acl->action == ACL_PERMIT)
533 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
534 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
536 if (acl->action == ACL_PERMIT)
554 /*********************************************************************
556 * Function : acl_addr
558 * Description : Called from `load_config' to parse an ACL address.
561 * 1 : aspec = String specifying ACL address.
562 * 2 : aca = struct access_control_addr to fill in.
564 * Returns : 0 => Ok, everything else is an error.
566 *********************************************************************/
567 int acl_addr(char *aspec, struct access_control_addr *aca)
569 int i, masklength, port;
575 if ((p = strchr(aspec, '/')) != NULL)
579 if (ijb_isdigit(*p) == 0)
583 masklength = atoi(p);
586 if ((masklength < 0) || (masklength > 32))
591 if ((p = strchr(aspec, ':')) != NULL)
595 if (ijb_isdigit(*p) == 0)
604 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
606 if (aca->addr == INADDR_NONE)
611 /* build the netmask */
613 for (i=1; i <= masklength ; i++)
615 aca->mask |= (1 << (32 - i));
618 /* now mask off the host portion of the ip address
619 * (i.e. save on the network portion of the address).
621 aca->addr = aca->addr & aca->mask;
626 #endif /* def FEATURE_ACL */
629 /*********************************************************************
631 * Function : match_portlist
633 * Description : Check if a given number is covered by a comma
634 * separated list of numbers and ranges (a,b-c,d,..)
637 * 1 : portlist = String with list
638 * 2 : port = port to check
640 * Returns : 0 => no match
643 *********************************************************************/
644 int match_portlist(const char *portlist, int port)
646 char *min, *max, *next, *portlist_copy;
648 min = next = portlist_copy = strdup(portlist);
651 * Zero-terminate first item and remember offset for next
653 if (NULL != (next = strchr(portlist_copy, (int) ',')))
659 * Loop through all items, checking for match
663 if (NULL == (max = strchr(min, (int) '-')))
666 * No dash, check for equality
668 if (port == atoi(min))
677 * This is a range, so check if between min and max,
678 * or, if max was omitted, between min and 65K
681 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
695 * Zero-terminate next item and remember offset for n+1
697 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
709 /*********************************************************************
711 * Function : block_url
713 * Description : Called from `chat'. Check to see if we need to block this.
716 * 1 : csp = Current client state (buffers, headers, etc...)
718 * Returns : NULL => unblocked, else HTTP block response
720 *********************************************************************/
721 struct http_response *block_url(struct client_state *csp)
723 struct http_response *rsp;
726 * If it's not blocked, don't block it ;-)
728 if ((csp->action->flags & ACTION_BLOCK) == 0)
734 * Else, prepare a response
736 if (NULL == (rsp = alloc_http_response()))
738 return cgi_error_memory();
742 * If it's an image-url, send back an image or redirect
743 * as specified by the relevant +image action
745 #ifdef FEATURE_IMAGE_BLOCKING
746 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
750 /* determine HOW images should be blocked */
751 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
753 #if 1 /* Two alternative strategies, use this one for now: */
755 /* and handle accordingly: */
756 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
758 rsp->body = bindup(image_pattern_data, image_pattern_length);
759 if (rsp->body == NULL)
761 free_http_response(rsp);
762 return cgi_error_memory();
764 rsp->content_length = image_pattern_length;
766 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
768 free_http_response(rsp);
769 return cgi_error_memory();
773 else if (0 == strcmpic(p, "blank"))
775 rsp->body = bindup(image_blank_data, image_blank_length);
776 if (rsp->body == NULL)
778 free_http_response(rsp);
779 return cgi_error_memory();
781 rsp->content_length = image_blank_length;
783 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
785 free_http_response(rsp);
786 return cgi_error_memory();
792 rsp->status = strdup("302 Local Redirect from Privoxy");
793 if (rsp->status == NULL)
795 free_http_response(rsp);
796 return cgi_error_memory();
799 if (enlist_unique_header(rsp->headers, "Location", p))
801 free_http_response(rsp);
802 return cgi_error_memory();
806 #else /* Following code is disabled for now */
808 /* and handle accordingly: */
809 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
811 p = CGI_PREFIX "send-banner?type=pattern";
813 else if (0 == strcmpic(p, "blank"))
815 p = CGI_PREFIX "send-banner?type=blank";
817 rsp->status = strdup("302 Local Redirect from Privoxy");
818 if (rsp->status == NULL)
820 free_http_response(rsp);
821 return cgi_error_memory();
824 if (enlist_unique_header(rsp->headers, "Location", p))
826 free_http_response(rsp);
827 return cgi_error_memory();
829 #endif /* Preceeding code is disabled for now */
832 #endif /* def FEATURE_IMAGE_BLOCKING */
835 * Else, generate an HTML "blocked" message:
839 struct map * exports;
843 * Workaround for stupid Netscape bug which prevents
844 * pages from being displayed if loading a referenced
845 * JavaScript or style sheet fails. So make it appear
846 * as if it succeeded.
848 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
849 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
850 && !strstr(p, "Gecko") /* save Mozilla, */
851 && !strstr(p, "compatible") /* MSIE */
852 && !strstr(p, "Opera")) /* and Opera. */
854 rsp->status = strdup("200 Request for blocked URL");
858 rsp->status = strdup("404 Request for blocked URL");
861 if (rsp->status == NULL)
863 free_http_response(rsp);
864 return cgi_error_memory();
867 exports = default_exports(csp, NULL);
870 free_http_response(rsp);
871 return cgi_error_memory();
874 #ifdef FEATURE_FORCE_LOAD
875 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
876 if (csp->http->ssl != 0)
877 #endif /* ndef FEATURE_FORCE_LOAD */
879 err = map_block_killer(exports, "force-support");
882 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
883 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
884 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
885 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
890 free_http_response(rsp);
891 return cgi_error_memory();
894 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
897 free_http_response(rsp);
898 return cgi_error_memory();
902 return finish_http_response(rsp);
908 /*********************************************************************
910 * Function : trust_url FIXME: I should be called distrust_url
912 * Description : Calls is_untrusted_url to determine if the URL is trusted
913 * and if not, returns a HTTP 304 response with a reject message.
916 * 1 : csp = Current client state (buffers, headers, etc...)
918 * Returns : NULL => trusted, else http_response.
920 *********************************************************************/
921 struct http_response *trust_url(struct client_state *csp)
923 struct http_response *rsp;
924 struct map * exports;
925 char buf[BUFFER_SIZE];
927 struct url_spec **tl;
932 * Don't bother to work on trusted URLs
934 if (!is_untrusted_url(csp))
940 * Else, prepare a response:
942 if (NULL == (rsp = alloc_http_response()))
944 return cgi_error_memory();
947 exports = default_exports(csp, NULL);
950 free_http_response(rsp);
951 return cgi_error_memory();
955 * Export the protocol, host, port, and referrer information
957 err = map(exports, "hostport", 1, csp->http->hostport, 1);
958 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
959 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
961 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
963 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
967 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
973 free_http_response(rsp);
974 return cgi_error_memory();
978 * Export the trust list
981 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
983 sprintf(buf, "<li>%s</li>\n", t->spec);
984 string_append(&p, buf);
986 err = map(exports, "trusted-referrers", 1, p, 0);
992 free_http_response(rsp);
993 return cgi_error_memory();
997 * Export the trust info, if available
999 if (csp->config->trust_info->first)
1001 struct list_entry *l;
1004 for (l = csp->config->trust_info->first; l ; l = l->next)
1006 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
1007 string_append(&p, buf);
1009 err = map(exports, "trust-info", 1, p, 0);
1014 err = map_block_killer(exports, "have-trust-info");
1020 free_http_response(rsp);
1021 return cgi_error_memory();
1025 * Export the force prefix or the force conditional block killer
1027 #ifdef FEATURE_FORCE_LOAD
1028 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1029 #else /* ifndef FEATURE_FORCE_LOAD */
1030 err = map_block_killer(exports, "force-support");
1031 #endif /* ndef FEATURE_FORCE_LOAD */
1036 free_http_response(rsp);
1037 return cgi_error_memory();
1041 * Build the response
1043 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1046 free_http_response(rsp);
1047 return cgi_error_memory();
1050 return finish_http_response(rsp);
1052 #endif /* def FEATURE_TRUST */
1055 #ifdef FEATURE_FAST_REDIRECTS
1056 /*********************************************************************
1058 * Function : redirect_url
1060 * Description : Checks for redirection URLs and returns a HTTP redirect
1061 * to the destination URL, if necessary
1064 * 1 : csp = Current client state (buffers, headers, etc...)
1066 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1068 *********************************************************************/
1069 struct http_response *redirect_url(struct client_state *csp)
1072 struct http_response *rsp;
1074 p = q = csp->http->path;
1075 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1078 * find the last URL encoded in the request
1080 while ((p = strstr(p, "http://")) != NULL)
1086 * if there was any, generate and return a HTTP redirect
1088 if (q != csp->http->path)
1090 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1092 if (NULL == (rsp = alloc_http_response()))
1094 return cgi_error_memory();
1097 if ( enlist_unique_header(rsp->headers, "Location", q)
1098 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1100 free_http_response(rsp);
1101 return cgi_error_memory();
1104 return finish_http_response(rsp);
1112 #endif /* def FEATURE_FAST_REDIRECTS */
1115 #ifdef FEATURE_IMAGE_BLOCKING
1116 /*********************************************************************
1118 * Function : is_imageurl
1120 * Description : Given a URL, decide whether it is an image or not,
1121 * using either the info from a previous +image action
1122 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1123 * is MSIE and not on a Mac, tell from the browser's accept
1127 * 1 : csp = Current client state (buffers, headers, etc...)
1129 * Returns : True (nonzero) if URL is an image, false (0)
1132 *********************************************************************/
1133 int is_imageurl(struct client_state *csp)
1135 #ifdef FEATURE_IMAGE_DETECT_MSIE
1138 tmp = get_header_value(csp->headers, "User-Agent:");
1139 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1141 tmp = get_header_value(csp->headers, "Accept:");
1142 if (tmp && strstr(tmp, "image/gif"))
1144 /* Client will accept HTML. If this seems counterintuitive,
1154 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1156 return ((csp->action->flags & ACTION_IMAGE) != 0);
1159 #endif /* def FEATURE_IMAGE_BLOCKING */
1162 #ifdef FEATURE_TRUST
1163 /*********************************************************************
1165 * Function : is_untrusted_url
1167 * Description : Should we "distrust" this URL (and block it)?
1169 * Yes if it matches a line in the trustfile, or if the
1170 * referrer matches a line starting with "+" in the
1175 * 1 : csp = Current client state (buffers, headers, etc...)
1177 * Returns : 0 => trusted, 1 => untrusted
1179 *********************************************************************/
1180 int is_untrusted_url(struct client_state *csp)
1182 struct file_list *fl;
1183 struct block_spec *b;
1184 struct url_spec **trusted_url;
1185 struct http_request rhttp[1];
1186 const char * referer;
1190 * If we don't have a trustlist, we trust everybody
1192 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1197 memset(rhttp, '\0', sizeof(*rhttp));
1200 * Do we trust the request URL itself?
1202 for (b = b->next; b ; b = b->next)
1204 if (url_match(b->url, csp->http))
1210 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1212 /* no referrer was supplied */
1218 * If not, do we maybe trust its referrer?
1220 err = parse_http_url(referer, rhttp, csp);
1226 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1228 if (url_match(*trusted_url, rhttp))
1230 /* if the URL's referrer is from a trusted referrer, then
1231 * add the target spec to the trustfile as an unblocked
1232 * domain and return NULL (which means it's OK).
1237 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1241 char * new_entry = strdup("~");
1243 string_append(&new_entry, csp->http->hostport);
1245 path = csp->http->path;
1246 if ( (path[0] == '/')
1248 && ((path_end = strchr(path + 2, '/')) != NULL))
1250 /* since this path points into a user's home space
1251 * be sure to include this spec in the trustfile.
1253 int path_len = path_end - path; /* save offset */
1254 path = strdup(path); /* Copy string */
1257 path_end = path + path_len; /* regenerate ptr to new buffer */
1258 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1260 string_join(&new_entry, path);
1263 if (new_entry != NULL)
1265 fprintf(fp, "%s\n", new_entry);
1270 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1271 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1281 #endif /* def FEATURE_TRUST */
1284 /*********************************************************************
1286 * Function : pcrs_filter_response
1288 * Description : Execute all text substitutions from all applying
1289 * +filter actions on the text buffer that's been accumulated
1290 * in csp->iob->buf. If this changes the contents, set
1291 * csp->content_length to the modified size and raise the
1292 * CSP_FLAG_MODIFIED flag.
1295 * 1 : csp = Current client state (buffers, headers, etc...)
1297 * Returns : a pointer to the (newly allocated) modified buffer.
1298 * or NULL if there were no hits or something went wrong
1300 *********************************************************************/
1301 char *pcrs_filter_response(struct client_state *csp)
1306 char *old = csp->iob->cur, *new = NULL;
1309 struct file_list *fl;
1310 struct re_filterfile_spec *b;
1311 struct list_entry *filtername;
1313 int i, found_filters = 0;
1318 if (csp->iob->cur >= csp->iob->eod)
1322 size = csp->iob->eod - csp->iob->cur;
1325 * Need to check the set of re_filterfiles...
1327 for (i = 0; i < MAX_AF_FILES; i++)
1340 if (0 == found_filters)
1342 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1347 * If the body has a "chunked" transfer-encoding,
1348 * get rid of it first, adjusting size and iob->eod
1350 if (csp->flags & CSP_FLAG_CHUNKED)
1352 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1353 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1357 csp->iob->eod = csp->iob->cur + size;
1358 csp->flags |= CSP_FLAG_MODIFIED;
1361 for (i = 0; i < MAX_AF_FILES; i++)
1364 if ((NULL == fl) || (NULL == fl->f))
1367 * For all applying +filter actions, look if a filter by that
1368 * name exists and if yes, execute it's pcrs_joblist on the
1371 for (b = fl->f; b; b = b->next)
1373 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1374 filtername ; filtername = filtername->next)
1376 if (strcmp(b->name, filtername->str) == 0)
1378 int current_hits = 0;
1380 if ( NULL == b->joblist )
1382 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1386 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1387 csp->http->hostport, csp->http->path, size, b->name);
1389 /* Apply all jobs from the joblist */
1390 for (job = b->joblist; NULL != job; job = job->next)
1392 current_hits += pcrs_execute(job, old, size, &new, &size);
1393 if (old != csp->iob->cur) free(old);
1397 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1398 hits += current_hits;
1405 * If there were no hits, destroy our copy and let
1406 * chat() use the original in csp->iob
1414 csp->flags |= CSP_FLAG_MODIFIED;
1415 csp->content_length = size;
1423 /*********************************************************************
1425 * Function : gif_deanimate_response
1427 * Description : Deanimate the GIF image that has been accumulated in
1428 * csp->iob->buf, set csp->content_length to the modified
1429 * size and raise the CSP_FLAG_MODIFIED flag.
1432 * 1 : csp = Current client state (buffers, headers, etc...)
1434 * Returns : a pointer to the (newly allocated) modified buffer.
1435 * or NULL in case something went wrong.
1437 *********************************************************************/
1438 char *gif_deanimate_response(struct client_state *csp)
1440 struct binbuffer *in, *out;
1442 size_t size = csp->iob->eod - csp->iob->cur;
1445 * If the body has a "chunked" transfer-encoding,
1446 * get rid of it first, adjusting size and iob->eod
1448 if (csp->flags & CSP_FLAG_CHUNKED)
1450 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1451 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1455 csp->iob->eod = csp->iob->cur + size;
1456 csp->flags |= CSP_FLAG_MODIFIED;
1459 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1460 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1462 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1466 in->buffer = csp->iob->cur;
1469 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1471 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1478 if ((int)size == out->offset)
1480 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1484 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1486 csp->content_length = out->offset;
1487 csp->flags |= CSP_FLAG_MODIFIED;
1497 /*********************************************************************
1499 * Function : jpeg_inspect_response
1504 * 1 : csp = Current client state (buffers, headers, etc...)
1506 * Returns : a pointer to the (newly allocated) modified buffer
1507 * or NULL in case something went wrong.
1509 *********************************************************************/
1510 char *jpeg_inspect_response(struct client_state *csp)
1512 struct binbuffer *in = NULL, *out = NULL;
1514 size_t size = csp->iob->eod - csp->iob->cur;
1517 * If the body has a "chunked" transfer-encoding,
1518 * get rid of it first, adjusting size and iob->eod
1520 if (csp->flags & CSP_FLAG_CHUNKED)
1522 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1523 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1527 csp->iob->eod = csp->iob->cur + size;
1528 csp->flags |= CSP_FLAG_MODIFIED;
1531 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1533 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
1537 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
1539 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
1543 in->buffer = csp->iob->cur;
1547 * Calling jpeg_inspect has the side-effect of creating and
1548 * modifying the image buffer of "out" directly.
1550 if (jpeg_inspect(in, out))
1552 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
1560 csp->content_length = out->offset;
1561 csp->flags |= CSP_FLAG_MODIFIED;
1571 /*********************************************************************
1573 * Function : remove_chunked_transfer_coding
1575 * Description : In-situ remove the "chunked" transfer coding as defined
1576 * in rfc2616 from a buffer.
1579 * 1 : buffer = Pointer to the text buffer
1580 * 2 : size = Number of bytes to be processed
1582 * Returns : The new size, i.e. the number of bytes from buffer which
1583 * are occupied by the stripped body, or 0 in case something
1586 *********************************************************************/
1587 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1590 unsigned int chunksize = 0;
1591 char *from_p, *to_p;
1594 from_p = to_p = buffer;
1596 if (sscanf(buffer, "%x", &chunksize) != 1)
1598 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1602 while (chunksize > 0)
1604 if (NULL == (from_p = strstr(from_p, "\r\n")))
1606 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1610 if ((newsize += chunksize) >= size)
1612 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
1617 memmove(to_p, from_p, (size_t) chunksize);
1618 to_p = buffer + newsize;
1619 from_p += chunksize + 2;
1621 if (sscanf(from_p, "%x", &chunksize) != 1)
1623 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1628 /* FIXME: Should this get its own loglevel? */
1629 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1635 /*********************************************************************
1637 * Function : url_actions
1639 * Description : Gets the actions for this URL.
1642 * 1 : http = http_request request for blocked URLs
1643 * 2 : csp = Current client state (buffers, headers, etc...)
1647 *********************************************************************/
1648 void url_actions(struct http_request *http,
1649 struct client_state *csp)
1651 struct file_list *fl;
1652 struct url_actions *b;
1655 init_current_action(csp->action);
1657 for (i = 0; i < MAX_AF_FILES; i++)
1659 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1664 apply_url_actions(csp->action, http, b);
1671 /*********************************************************************
1673 * Function : apply_url_actions
1675 * Description : Applies a list of URL actions.
1678 * 1 : action = Destination.
1679 * 2 : http = Current URL
1680 * 3 : b = list of URL actions to apply
1684 *********************************************************************/
1685 void apply_url_actions(struct current_action_spec *action,
1686 struct http_request *http,
1687 struct url_actions *b)
1691 /* Should never happen */
1695 for (b = b->next; NULL != b; b = b->next)
1697 if (url_match(b->url, http))
1699 merge_current_action(action, b->action);
1705 /*********************************************************************
1707 * Function : forward_url
1709 * Description : Should we forward this to another proxy?
1712 * 1 : http = http_request request for current URL
1713 * 2 : csp = Current client state (buffers, headers, etc...)
1715 * Returns : Pointer to forwarding information.
1717 *********************************************************************/
1718 const struct forward_spec * forward_url(struct http_request *http,
1719 struct client_state *csp)
1721 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1722 struct forward_spec *fwd = csp->config->forward;
1731 if (url_match(fwd->url, http))
1742 /*********************************************************************
1744 * Function : direct_response
1746 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
1747 * request and if so, return a HTTP 501 to the client.
1749 * FIXME: I have a stupid name and I should handle the
1750 * requests properly. Still, what we do here is rfc-
1751 * compliant, whereas ignoring or forwarding are not.
1754 * 1 : csp = Current client state (buffers, headers, etc...)
1756 * Returns : http_response if , NULL if nonmatch or handler fail
1758 *********************************************************************/
1759 struct http_response *direct_response(struct client_state *csp)
1761 struct http_response *rsp;
1762 struct list_entry *p;
1764 if ((0 == strcmpic(csp->http->gpc, "trace"))
1765 || (0 == strcmpic(csp->http->gpc, "options")))
1767 for (p = csp->headers->first; (p != NULL) ; p = p->next)
1769 if (!strncmp("Max-Forwards:", p->str, 13)
1770 && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
1772 /* FIXME: We could handle at least TRACE here,
1773 but that would require a verbatim copy of
1774 the request which we don't have anymore */
1776 log_error(LOG_LEVEL_HEADER, "Found Max-Forwards:0 in OPTIONS or TRACE request -- Returning 501");
1778 /* Get mem for response or fail*/
1779 if (NULL == (rsp = alloc_http_response()))
1781 return cgi_error_memory();
1784 if (NULL == (rsp->status = strdup("501 Not Implemented")))
1786 free_http_response(rsp);
1787 return cgi_error_memory();
1791 return(finish_http_response(rsp));