1 const char filters_rcs[] = "$Id: filters.c,v 2.3 2002/12/28 03:58:19 david__schmidt 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 2.3 2002/12/28 03:58:19 david__schmidt
42 * Initial drop of dashboard instrumentation - enabled with
43 * --enable-activity-console
45 * Revision 2.2 2002/09/04 15:38:24 oes
46 * Synced with the stable branch:
47 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
48 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
50 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
51 * - Bugfix: Executing a filters without jobs no longer results in
52 * turing off *all* filters.
53 * - Security fix: Malicious web servers can't cause a seg fault
54 * through bogus chunk sizes anymore
56 * Revision 2.1 2002/08/26 11:08:18 sarantis
59 * Revision 2.0 2002/06/04 14:34:21 jongfoster
60 * Moving source files to src/
62 * Revision 1.58 2002/04/24 02:11:17 oes
63 * Jon's multiple AF patch: url_actions now evaluates rules
66 * Revision 1.57 2002/04/08 20:38:34 swa
69 * Revision 1.56 2002/04/05 15:51:24 oes
70 * - bugfix: error-pages now get correct request protocol
71 * - fix for invalid HTML in trust info
73 * Revision 1.55 2002/04/02 16:13:51 oes
74 * Fix: No "Go there anyway" for SSL
76 * Revision 1.54 2002/04/02 14:55:56 oes
77 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
79 * Revision 1.53 2002/03/26 22:29:54 swa
80 * we have a new homepage!
82 * Revision 1.52 2002/03/24 16:35:57 jongfoster
85 * Revision 1.51 2002/03/24 15:23:33 jongfoster
88 * Revision 1.50 2002/03/24 13:25:43 swa
89 * name change related issues
91 * Revision 1.49 2002/03/16 20:29:14 oes
94 * Revision 1.48 2002/03/13 20:25:34 oes
95 * Better logging for content filters
97 * Revision 1.47 2002/03/13 00:30:52 jongfoster
99 * Added option of always sending redirect for imageblock,
100 * currently disabled with #if 0.
102 * Revision 1.46 2002/03/12 01:42:49 oes
103 * Introduced modular filters
105 * Revision 1.45 2002/03/08 16:47:50 oes
106 * Added choice beween GIF and PNG built-in images
108 * Revision 1.44 2002/03/07 03:49:31 oes
109 * - Fixed compiler warnings etc
110 * - Changed built-in images from GIF to PNG
111 * (with regard to Unisys patent issue)
112 * - Added a 4x4 pattern PNG which is less intrusive
113 * than the logo but also clearly marks the deleted banners
115 * Revision 1.43 2002/01/22 23:51:59 jongfoster
116 * Replacing strsav() with the safer string_append().
118 * Adding missing html_encode() to error message generators. Where encoded
119 * and unencoded versions of a string were provided, removing the unencoded
122 * Revision 1.42 2002/01/17 21:00:32 jongfoster
123 * Moving all our URL and URL pattern parsing code to urlmatch.c.
125 * Using a single, simple url_match(pattern,url) function - rather than
126 * the 3-line match routine which was repeated all over the place.
128 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
130 * Using parse_http_url() to parse URLs without faking a HTTP
131 * request line for parse_http_request().
133 * Revision 1.41 2001/11/13 00:14:07 jongfoster
134 * Fixing stupid bug now I've figured out what || means.
135 * (It always returns 0 or 1, not one of it's paramaters.)
137 * Revision 1.40 2001/10/26 17:37:55 oes
138 * - Re-enabled Netscape 200/404 bug workaround in block_url():
139 * - Removed OS/2 special case
140 * - Made block_url() independant from sed() having been run
141 * - Made trust_url independant from sed() having been run
142 * - Made is_imageurl independant from sed() having been run.
143 * It now checks User-Agent: and Accept: by itself.
146 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
147 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
148 * threads to call select() simultaneously. So, it's time to do a real, live,
149 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
150 * (native). Both versions will work, but using __OS2__ offers multi-threading.
152 * Revision 1.38 2001/10/23 21:32:33 jongfoster
153 * Adding error-checking to selected functions
155 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
156 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
157 * filters.c. Added a FIXME in front of the offending code. I'll gladly
158 * put in a better/more robust fix for all parties if one is presented...
159 * It seems that just returning 200 instead of 404 would pretty much fix
160 * it for everyone, but I don't know all the history of the problem.
162 * Revision 1.36 2001/10/10 16:44:16 oes
163 * Added match_portlist function
165 * Revision 1.35 2001/10/07 15:41:23 oes
166 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
168 * New function remove_chunked_transfer_coding that strips chunked
169 * transfer coding to plain and is called by pcrs_filter_response
170 * and gif_deanimate_response if neccessary
172 * Improved handling of zero-change re_filter runs
174 * pcrs_filter_response and gif_deanimate_response now remove
175 * chunked transfer codeing before processing the body.
177 * Revision 1.34 2001/09/20 15:49:36 steudten
179 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
180 * See cgi.c fill_template().
182 * Revision 1.33 2001/09/16 17:05:14 jongfoster
183 * Removing unused #include showarg.h
185 * Revision 1.32 2001/09/16 13:21:27 jongfoster
186 * Changes to use new list functions.
188 * Revision 1.31 2001/09/16 11:38:02 jongfoster
189 * Splitting fill_template() into 2 functions:
190 * template_load() loads the file
191 * template_fill() performs the PCRS regexps.
192 * This is because the CGI edit interface has a "table row"
193 * template which is used many times in the page - this
194 * change means it's only loaded from disk once.
196 * Revision 1.30 2001/09/16 11:00:10 jongfoster
197 * New function alloc_http_response, for symmetry with free_http_response
199 * Revision 1.29 2001/09/13 23:32:40 jongfoster
200 * Moving image data to cgi.c rather than cgi.h
201 * Fixing a GPF under Win32 (and any other OS that protects global
202 * constants from being written to).
204 * Revision 1.28 2001/09/10 10:18:51 oes
205 * Silenced compiler warnings
207 * Revision 1.27 2001/08/05 16:06:20 jongfoster
208 * Modifiying "struct map" so that there are now separate header and
209 * "map_entry" structures. This means that functions which modify a
210 * map no longer need to return a pointer to the modified map.
211 * Also, it no longer reverses the order of the entries (which may be
212 * important with some advanced template substitutions).
214 * Revision 1.26 2001/07/30 22:08:36 jongfoster
215 * Tidying up #defines:
216 * - All feature #defines are now of the form FEATURE_xxx
217 * - Permanently turned off WIN_GUI_EDIT
218 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
220 * Revision 1.25 2001/07/26 10:09:46 oes
221 * Made browser detection a little less naive
223 * Revision 1.24 2001/07/25 17:22:51 oes
224 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
226 * Revision 1.23 2001/07/23 13:40:12 oes
227 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
229 * Revision 1.22 2001/07/18 12:29:34 oes
230 * - Made gif_deanimate_response respect
231 * csp->action->string[ACTION_STRING_DEANIMATE]
232 * - Logging cosmetics
234 * Revision 1.21 2001/07/13 13:59:53 oes
235 * - Introduced gif_deanimate_response which shares the
236 * generic content modification interface of pcrs_filter_response
237 * and acts as a wrapper to deanimate.c:gif_deanimate()
238 * - Renamed re_process_buffer to pcrs_filter_response
239 * - pcrs_filter_response now returns NULL on failure
240 * - Removed all #ifdef PCRS
242 * Revision 1.20 2001/07/01 17:01:04 oes
243 * Added comments and missing return statement in is_untrusted_url()
245 * Revision 1.19 2001/06/29 21:45:41 oes
246 * Indentation, CRLF->LF, Tab-> Space
248 * Revision 1.18 2001/06/29 13:27:38 oes
249 * - Cleaned up, renamed and reorderd functions
250 * and improved comments
253 * - Ported to CGI platform. Now delivers
254 * http_response or NULL
255 * - Unified HTML and GIF generation (moved image detection
256 * and GIF generation here from jcc.c:chat())
257 * - Fixed HTTP status to:
258 * - 403 (Forbidden) for the "blocked" HTML message
259 * - 200 (OK) for GIF answers
260 * - 302 (Redirect) for redirect to GIF
263 * - Ported to CGI platform. Now delivers
264 * http_response or NULL
265 * - Separated detection of untrusted URL into
266 * (bool)is_untrusted_url
267 * - Added enforcement of untrusted requests
269 * - Moved redirect_url() from cgi.c to here
270 * and ported it to the CGI platform
272 * - Removed logentry from cancelled commit
274 * Revision 1.17 2001/06/09 10:55:28 jongfoster
275 * Changing BUFSIZ ==> BUFFER_SIZE
277 * Revision 1.16 2001/06/07 23:10:26 jongfoster
278 * Allowing unanchored domain patterns to back off and retry
279 * if they partially match. Optimized right-anchored patterns.
280 * Moving ACL and forward files into config file.
281 * Replacing struct gateway with struct forward_spec
283 * Revision 1.15 2001/06/03 19:12:00 oes
284 * extracted-CGI relevant stuff
286 * Revision 1.14 2001/06/01 10:30:55 oes
287 * Added optional left-anchoring to domaincmp
289 * Revision 1.13 2001/05/31 21:21:30 jongfoster
290 * Permissionsfile / actions file changes:
291 * - Changed "permission" to "action" throughout
292 * - changes to file format to allow string parameters
293 * - Moved helper functions to actions.c
295 * Revision 1.12 2001/05/31 17:35:20 oes
297 * - Enhanced domain part globbing with infix and prefix asterisk
298 * matching and optional unanchored operation
300 * Revision 1.11 2001/05/29 11:53:23 oes
301 * "See why" link added to "blocked" page
303 * Revision 1.10 2001/05/29 09:50:24 jongfoster
304 * Unified blocklist/imagelist/permissionslist.
305 * File format is still under discussion, but the internal changes
308 * Also modified interceptor behaviour:
309 * - We now intercept all URLs beginning with one of the following
310 * prefixes (and *only* these prefixes):
312 * * http://ijbswa.sf.net/config/
313 * * http://ijbswa.sourceforge.net/config/
314 * - New interceptors "home page" - go to http://i.j.b/ to see it.
315 * - Internal changes so that intercepted and fast redirect pages
316 * are not replaced with an image.
317 * - Interceptors now have the option to send a binary page direct
318 * to the client. (i.e. ijb-send-banner uses this)
319 * - Implemented show-url-info interceptor. (Which is why I needed
320 * the above interceptors changes - a typical URL is
321 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
322 * The previous mechanism would not have intercepted that, and
323 * if it had been intercepted then it then it would have replaced
326 * Revision 1.9 2001/05/27 22:17:04 oes
328 * - re_process_buffer no longer writes the modified buffer
329 * to the client, which was very ugly. It now returns the
330 * buffer, which it is then written by chat.
332 * - content_length now adjusts the Content-Length: header
333 * for modified documents rather than crunch()ing it.
334 * (Length info in csp->content_length, which is 0 for
335 * unmodified documents)
337 * - For this to work, sed() is called twice when filtering.
339 * Revision 1.8 2001/05/26 17:13:28 jongfoster
340 * Filled in a function comment.
342 * Revision 1.7 2001/05/26 15:26:15 jongfoster
343 * ACL feature now provides more security by immediately dropping
344 * connections from untrusted hosts.
346 * Revision 1.6 2001/05/26 00:28:36 jongfoster
347 * Automatic reloading of config file.
348 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
349 * Most of the global variables have been moved to a new
350 * struct configuration_spec, accessed through csp->config->globalname
351 * Most of the globals remaining are used by the Win32 GUI.
353 * Revision 1.5 2001/05/25 22:34:30 jongfoster
356 * Revision 1.4 2001/05/22 18:46:04 oes
358 * - Enabled filtering banners by size rather than URL
359 * by adding patterns that replace all standard banner
360 * sizes with the "Junkbuster" gif to the re_filterfile
362 * - Enabled filtering WebBugs by providing a pattern
363 * which kills all 1x1 images
365 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
366 * which is selected by the (nonstandard and therefore
367 * capital) letter 'U' in the option string.
368 * It causes the quantifiers to be ungreedy by default.
369 * Appending a ? turns back to greedy (!).
371 * - Added a new interceptor ijb-send-banner, which
372 * sends back the "Junkbuster" gif. Without imagelist or
373 * MSIE detection support, or if tinygif = 1, or the
374 * URL isn't recognized as an imageurl, a lame HTML
375 * explanation is sent instead.
377 * - Added new feature, which permits blocking remote
378 * script redirects and firing back a local redirect
380 * The feature is conditionally compiled, i.e. it
381 * can be disabled with --disable-fast-redirects,
382 * plus it must be activated by a "fast-redirects"
383 * line in the config file, has its own log level
384 * and of course wants to be displayed by show-proxy-args
385 * Note: Boy, all the #ifdefs in 1001 locations and
386 * all the fumbling with configure.in and acconfig.h
387 * were *way* more work than the feature itself :-(
389 * - Because a generic redirect template was needed for
390 * this, tinygif = 3 now uses the same.
392 * - Moved GIFs, and other static HTTP response templates
397 * - Removed some >400 CRs again (Jon, you really worked
400 * Revision 1.3 2001/05/20 16:44:47 jongfoster
401 * Removing last hardcoded Junkbusters.com URLs.
403 * Revision 1.2 2001/05/20 01:21:20 jongfoster
404 * Version 2.9.4 checkin.
405 * - Merged popupfile and cookiefile, and added control over PCRS
406 * filtering, in new "permissionsfile".
407 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
408 * file error you now get a message box (in the Win32 GUI) rather
409 * than the program exiting with no explanation.
410 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
412 * - Removed tabs from "config"
413 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
414 * - Bumped up version number.
416 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
417 * Initial import of version 2.9.3 source tree
420 *********************************************************************/
426 #include <sys/types.h>
435 #endif /* ndef __OS2__ */
436 #include <netinet/in.h>
438 #include <winsock2.h>
439 #endif /* ndef _WIN32 */
443 #endif /* def __OS2__ */
451 #include "jbsockets.h"
452 #include "miscutil.h"
456 #include "deanimate.h"
457 #include "urlmatch.h"
458 #ifdef FEATURE_ACTIVITY_CONSOLE
460 #endif /* def FEATURE_ACTIVITY_CONSOLE */
465 const char filters_h_rcs[] = FILTERS_H_VERSION;
467 /* Fix a problem with Solaris. There should be no effect on other
469 * Solaris's isspace() is a macro which uses it's argument directly
470 * as an array index. Therefore we need to make sure that high-bit
471 * characters generate +ve values, and ideally we also want to make
472 * the argument match the declared parameter type of "int".
474 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
478 /*********************************************************************
480 * Function : block_acl
482 * Description : Block this request?
483 * Decide yes or no based on ACL file.
486 * 1 : dst = The proxy or gateway address this is going to.
487 * Or NULL to check all possible targets.
488 * 2 : csp = Current client state (buffers, headers, etc...)
489 * Also includes the client IP address.
491 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
493 *********************************************************************/
494 int block_acl(struct access_control_addr *dst, struct client_state *csp)
496 struct access_control_list *acl = csp->config->acl;
498 /* if not using an access control list, then permit the connection */
504 /* search the list */
507 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
511 /* Just want to check if they have any access */
512 if (acl->action == ACL_PERMIT)
517 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
518 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
520 if (acl->action == ACL_PERMIT)
526 #ifdef FEATURE_ACTIVITY_CONSOLE
527 accumulate_stats(STATS_ACL_RESTRICT, 1);
528 #endif /* def FEATURE_ACTIVITY_CONSOLE */
541 /*********************************************************************
543 * Function : acl_addr
545 * Description : Called from `load_config' to parse an ACL address.
548 * 1 : aspec = String specifying ACL address.
549 * 2 : aca = struct access_control_addr to fill in.
551 * Returns : 0 => Ok, everything else is an error.
553 *********************************************************************/
554 int acl_addr(char *aspec, struct access_control_addr *aca)
556 int i, masklength, port;
562 if ((p = strchr(aspec, '/')) != NULL)
566 if (ijb_isdigit(*p) == 0)
570 masklength = atoi(p);
573 if ((masklength < 0) || (masklength > 32))
578 if ((p = strchr(aspec, ':')) != NULL)
582 if (ijb_isdigit(*p) == 0)
591 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
593 if (aca->addr == INADDR_NONE)
598 /* build the netmask */
600 for (i=1; i <= masklength ; i++)
602 aca->mask |= (1 << (32 - i));
605 /* now mask off the host portion of the ip address
606 * (i.e. save on the network portion of the address).
608 aca->addr = aca->addr & aca->mask;
613 #endif /* def FEATURE_ACL */
616 /*********************************************************************
618 * Function : match_portlist
620 * Description : Check if a given number is covered by a comma
621 * separated list of numbers and ranges (a,b-c,d,..)
624 * 1 : portlist = String with list
625 * 2 : port = port to check
627 * Returns : 0 => no match
630 *********************************************************************/
631 int match_portlist(const char *portlist, int port)
633 char *min, *max, *next, *portlist_copy;
635 min = next = portlist_copy = strdup(portlist);
638 * Zero-terminate first item and remember offset for next
640 if (NULL != (next = strchr(portlist_copy, (int) ',')))
646 * Loop through all items, checking for match
650 if (NULL == (max = strchr(min, (int) '-')))
653 * No dash, check for equality
655 if (port == atoi(min))
664 * This is a range, so check if between min and max,
665 * or, if max was omitted, between min and 65K
668 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
682 * Zero-terminate next item and remember offset for n+1
684 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
696 /*********************************************************************
698 * Function : block_url
700 * Description : Called from `chat'. Check to see if we need to block this.
703 * 1 : csp = Current client state (buffers, headers, etc...)
705 * Returns : NULL => unblocked, else HTTP block response
707 *********************************************************************/
708 struct http_response *block_url(struct client_state *csp)
710 #ifdef FEATURE_IMAGE_BLOCKING
712 #endif /* def FEATURE_IMAGE_BLOCKING */
713 struct http_response *rsp;
716 * If it's not blocked, don't block it ;-)
718 if ((csp->action->flags & ACTION_BLOCK) == 0)
724 * Else, prepare a response
726 if (NULL == (rsp = alloc_http_response()))
728 return cgi_error_memory();
732 * If it's an image-url, send back an image or redirect
733 * as specified by the relevant +image action
735 #ifdef FEATURE_IMAGE_BLOCKING
736 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
739 /* determine HOW images should be blocked */
740 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
742 #if 1 /* Two alternative strategies, use this one for now: */
744 /* and handle accordingly: */
745 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
747 #ifdef FEATURE_ACTIVITY_CONSOLE
748 accumulate_stats(STATS_IMAGE_BLOCK, 1);
749 #endif /* def FEATURE_ACTIVITY_CONSOLE */
750 rsp->body = bindup(image_pattern_data, image_pattern_length);
751 if (rsp->body == NULL)
753 free_http_response(rsp);
754 return cgi_error_memory();
756 rsp->content_length = image_pattern_length;
758 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
760 free_http_response(rsp);
761 return cgi_error_memory();
765 else if (0 == strcmpic(p, "blank"))
767 #ifdef FEATURE_ACTIVITY_CONSOLE
768 accumulate_stats(STATS_IMAGE_BLOCK, 1);
769 #endif /* def FEATURE_ACTIVITY_CONSOLE */
770 rsp->body = bindup(image_blank_data, image_blank_length);
771 if (rsp->body == NULL)
773 free_http_response(rsp);
774 return cgi_error_memory();
776 rsp->content_length = image_blank_length;
778 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
780 free_http_response(rsp);
781 return cgi_error_memory();
787 rsp->status = strdup("302 Local Redirect from Privoxy");
788 if (rsp->status == NULL)
790 free_http_response(rsp);
791 return cgi_error_memory();
794 if (enlist_unique_header(rsp->headers, "Location", p))
796 free_http_response(rsp);
797 return cgi_error_memory();
801 #else /* Following code is disabled for now */
803 /* and handle accordingly: */
804 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
806 p = CGI_PREFIX "send-banner?type=pattern";
808 else if (0 == strcmpic(p, "blank"))
810 p = CGI_PREFIX "send-banner?type=blank";
812 rsp->status = strdup("302 Local Redirect from Privoxy");
813 if (rsp->status == NULL)
815 free_http_response(rsp);
816 return cgi_error_memory();
819 if (enlist_unique_header(rsp->headers, "Location", p))
821 free_http_response(rsp);
822 return cgi_error_memory();
824 #endif /* Preceeding code is disabled for now */
827 #endif /* def FEATURE_IMAGE_BLOCKING */
830 * Else, generate an HTML "blocked" message:
834 struct map * exports;
837 * Workaround for stupid Netscape bug which prevents
838 * pages from being displayed if loading a referenced
839 * JavaScript or style sheet fails. So make it appear
840 * as if it succeeded.
842 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
843 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
844 && !strstr(p, "Gecko") /* save Mozilla, */
845 && !strstr(p, "compatible") /* MSIE */
846 && !strstr(p, "Opera")) /* and Opera. */
848 rsp->status = strdup("200 Request for blocked URL");
852 rsp->status = strdup("404 Request for blocked URL");
855 if (rsp->status == NULL)
857 free_http_response(rsp);
858 return cgi_error_memory();
861 exports = default_exports(csp, NULL);
864 free_http_response(rsp);
865 return cgi_error_memory();
868 #ifdef FEATURE_FORCE_LOAD
869 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
870 if (csp->http->ssl != 0)
871 #endif /* ndef FEATURE_FORCE_LOAD */
873 err = map_block_killer(exports, "force-support");
876 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
877 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
878 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
883 free_http_response(rsp);
884 return cgi_error_memory();
887 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
890 free_http_response(rsp);
891 return cgi_error_memory();
895 return finish_http_response(rsp);
901 /*********************************************************************
903 * Function : trust_url FIXME: I should be called distrust_url
905 * Description : Calls is_untrusted_url to determine if the URL is trusted
906 * and if not, returns a HTTP 304 response with a reject message.
909 * 1 : csp = Current client state (buffers, headers, etc...)
911 * Returns : NULL => trusted, else http_response.
913 *********************************************************************/
914 struct http_response *trust_url(struct client_state *csp)
916 struct http_response *rsp;
917 struct map * exports;
918 char buf[BUFFER_SIZE];
920 struct url_spec **tl;
925 * Don't bother to work on trusted URLs
927 if (!is_untrusted_url(csp))
933 * Else, prepare a response:
935 if (NULL == (rsp = alloc_http_response()))
937 return cgi_error_memory();
940 exports = default_exports(csp, NULL);
943 free_http_response(rsp);
944 return cgi_error_memory();
948 * Export the protocol, host, port, and referrer information
950 err = map(exports, "hostport", 1, csp->http->hostport, 1);
951 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
952 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
954 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
956 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
960 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
966 free_http_response(rsp);
967 return cgi_error_memory();
971 * Export the trust list
974 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
976 sprintf(buf, "<li>%s</li>\n", t->spec);
977 string_append(&p, buf);
979 err = map(exports, "trusted-referrers", 1, p, 0);
984 free_http_response(rsp);
985 return cgi_error_memory();
989 * Export the trust info, if available
991 if (csp->config->trust_info->first)
993 struct list_entry *l;
996 for (l = csp->config->trust_info->first; l ; l = l->next)
998 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
999 string_append(&p, buf);
1001 err = map(exports, "trust-info", 1, p, 0);
1005 err = map_block_killer(exports, "have-trust-info");
1011 free_http_response(rsp);
1012 return cgi_error_memory();
1016 * Export the force prefix or the force conditional block killer
1018 #ifdef FEATURE_FORCE_LOAD
1019 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1020 #else /* ifndef FEATURE_FORCE_LOAD */
1021 err = map_block_killer(exports, "force-support");
1022 #endif /* ndef FEATURE_FORCE_LOAD */
1027 free_http_response(rsp);
1028 return cgi_error_memory();
1032 * Build the response
1034 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1037 free_http_response(rsp);
1038 return cgi_error_memory();
1041 return finish_http_response(rsp);
1043 #endif /* def FEATURE_TRUST */
1046 #ifdef FEATURE_FAST_REDIRECTS
1047 /*********************************************************************
1049 * Function : redirect_url
1051 * Description : Checks for redirection URLs and returns a HTTP redirect
1052 * to the destination URL, if necessary
1055 * 1 : csp = Current client state (buffers, headers, etc...)
1057 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1059 *********************************************************************/
1060 struct http_response *redirect_url(struct client_state *csp)
1063 struct http_response *rsp;
1065 p = q = csp->http->path;
1066 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1069 * find the last URL encoded in the request
1071 while ((p = strstr(p, "http://")) != NULL)
1077 * if there was any, generate and return a HTTP redirect
1079 if (q != csp->http->path)
1081 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1083 if (NULL == (rsp = alloc_http_response()))
1085 return cgi_error_memory();
1088 if ( enlist_unique_header(rsp->headers, "Location", q)
1089 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1091 free_http_response(rsp);
1092 return cgi_error_memory();
1095 return finish_http_response(rsp);
1103 #endif /* def FEATURE_FAST_REDIRECTS */
1106 #ifdef FEATURE_IMAGE_BLOCKING
1107 /*********************************************************************
1109 * Function : is_imageurl
1111 * Description : Given a URL, decide whether it is an image or not,
1112 * using either the info from a previous +image action
1113 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1114 * is MSIE and not on a Mac, tell from the browser's accept
1118 * 1 : csp = Current client state (buffers, headers, etc...)
1120 * Returns : True (nonzero) if URL is an image, false (0)
1123 *********************************************************************/
1124 int is_imageurl(struct client_state *csp)
1126 #ifdef FEATURE_IMAGE_DETECT_MSIE
1129 tmp = get_header_value(csp->headers, "User-Agent:");
1130 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1132 tmp = get_header_value(csp->headers, "Accept:");
1133 if (tmp && strstr(tmp, "image/gif"))
1135 /* Client will accept HTML. If this seems counterintuitive,
1145 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1147 return ((csp->action->flags & ACTION_IMAGE) != 0);
1150 #endif /* def FEATURE_IMAGE_BLOCKING */
1153 #ifdef FEATURE_TRUST
1154 /*********************************************************************
1156 * Function : is_untrusted_url
1158 * Description : Should we "distrust" this URL (and block it)?
1160 * Yes if it matches a line in the trustfile, or if the
1161 * referrer matches a line starting with "+" in the
1166 * 1 : csp = Current client state (buffers, headers, etc...)
1168 * Returns : 0 => trusted, 1 => untrusted
1170 *********************************************************************/
1171 int is_untrusted_url(struct client_state *csp)
1173 struct file_list *fl;
1174 struct block_spec *b;
1175 struct url_spec **trusted_url;
1176 struct http_request rhttp[1];
1177 const char * referer;
1181 * If we don't have a trustlist, we trust everybody
1183 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1188 memset(rhttp, '\0', sizeof(*rhttp));
1191 * Do we trust the request URL itself?
1193 for (b = b->next; b ; b = b->next)
1195 if (url_match(b->url, csp->http))
1201 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1203 /* no referrer was supplied */
1209 * If not, do we maybe trust its referrer?
1211 err = parse_http_url(referer, rhttp, csp);
1217 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1219 if (url_match(*trusted_url, rhttp))
1221 /* if the URL's referrer is from a trusted referrer, then
1222 * add the target spec to the trustfile as an unblocked
1223 * domain and return NULL (which means it's OK).
1228 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1232 char * new_entry = strdup("~");
1234 string_append(&new_entry, csp->http->hostport);
1236 path = csp->http->path;
1237 if ( (path[0] == '/')
1239 && ((path_end = strchr(path + 2, '/')) != NULL))
1241 /* since this path points into a user's home space
1242 * be sure to include this spec in the trustfile.
1244 int path_len = path_end - path; /* save offset */
1245 path = strdup(path); /* Copy string */
1248 path_end = path + path_len; /* regenerate ptr to new buffer */
1249 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1251 string_join(&new_entry, path);
1254 if (new_entry != NULL)
1256 fprintf(fp, "%s\n", new_entry);
1261 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1262 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1272 #endif /* def FEATURE_TRUST */
1275 /*********************************************************************
1277 * Function : pcrs_filter_response
1279 * Description : Ecexute all text substitutions from all applying
1280 * +filter actions on the text buffer that's been accumulated
1281 * in csp->iob->buf. If this changes the contents, set
1282 * csp->content_length to the modified size and raise the
1283 * CSP_FLAG_MODIFIED flag.
1286 * 1 : csp = Current client state (buffers, headers, etc...)
1288 * Returns : a pointer to the (newly allocated) modified buffer.
1289 * or NULL if there were no hits or something went wrong
1291 *********************************************************************/
1292 char *pcrs_filter_response(struct client_state *csp)
1297 char *old_buf = csp->iob->cur, *new_buf = NULL;
1300 struct file_list *fl;
1301 struct re_filterfile_spec *b;
1302 struct list_entry *filtername;
1307 if (csp->iob->cur >= csp->iob->eod)
1311 size = csp->iob->eod - csp->iob->cur;
1313 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1315 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1320 * If the body has a "chunked" transfer-encoding,
1321 * get rid of it first, adjusting size and iob->eod
1323 if (csp->flags & CSP_FLAG_CHUNKED)
1325 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1326 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1330 csp->iob->eod = csp->iob->cur + size;
1331 csp->flags |= CSP_FLAG_MODIFIED;
1335 * For all applying +filter actions, look if a filter by that
1336 * name exists and if yes, execute it's pcrs_joblist on the
1339 for (b = fl->f; b; b = b->next)
1341 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1342 filtername ; filtername = filtername->next)
1344 if (strcmp(b->name, filtername->str) == 0)
1346 int current_hits = 0;
1348 if ( NULL == b->joblist )
1350 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1354 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1355 csp->http->hostport, csp->http->path, size, b->name);
1357 /* Apply all jobs from the joblist */
1358 for (job = b->joblist; NULL != job; job = job->next)
1360 current_hits += pcrs_execute(job, old_buf, size, &new_buf, &size);
1361 if (old_buf != csp->iob->cur) free(old_buf);
1365 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1366 hits += current_hits;
1372 * If there were no hits, destroy our copy and let
1373 * chat() use the original in csp->iob
1380 #ifdef FEATURE_ACTIVITY_CONSOLE
1382 accumulate_stats(STATS_FILTER, hits);
1383 #endif /* def FEATURE_ACTIVITY_CONSOLE */
1385 csp->flags |= CSP_FLAG_MODIFIED;
1386 csp->content_length = size;
1394 /*********************************************************************
1396 * Function : gif_deanimate_response
1398 * Description : Deanimate the GIF image that has been accumulated in
1399 * csp->iob->buf, set csp->content_length to the modified
1400 * size and raise the CSP_FLAG_MODIFIED flag.
1403 * 1 : csp = Current client state (buffers, headers, etc...)
1405 * Returns : a pointer to the (newly allocated) modified buffer.
1406 * or NULL in case something went wrong.
1408 *********************************************************************/
1409 char *gif_deanimate_response(struct client_state *csp)
1411 struct binbuffer *in, *out;
1413 size_t size = csp->iob->eod - csp->iob->cur;
1416 * If the body has a "chunked" transfer-encoding,
1417 * get rid of it first, adjusting size and iob->eod
1419 if (csp->flags & CSP_FLAG_CHUNKED)
1421 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1422 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1426 csp->iob->eod = csp->iob->cur + size;
1427 csp->flags |= CSP_FLAG_MODIFIED;
1430 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1431 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1433 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1437 in->buffer = csp->iob->cur;
1440 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1442 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1449 if ((int)size == out->offset)
1451 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1455 #ifdef FEATURE_ACTIVITY_CONSOLE
1456 accumulate_stats(STATS_GIF_DEANIMATE, 1);
1457 #endif /* def FEATURE_ACTIVITY_CONSOLE */
1458 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1460 csp->content_length = out->offset;
1461 csp->flags |= CSP_FLAG_MODIFIED;
1471 /*********************************************************************
1473 * Function : remove_chunked_transfer_coding
1475 * Description : In-situ remove the "chunked" transfer coding as defined
1476 * in rfc2616 from a buffer.
1479 * 1 : buffer = Pointer to the text buffer
1480 * 2 : size = Number of bytes to be processed
1482 * Returns : The new size, i.e. the number of bytes from buffer which
1483 * are occupied by the stripped body, or 0 in case something
1486 *********************************************************************/
1487 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1490 unsigned int chunksize = 0;
1491 char *from_p, *to_p;
1494 from_p = to_p = buffer;
1496 if (sscanf(buffer, "%x", &chunksize) != 1)
1498 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1502 while (chunksize > 0)
1504 if (NULL == (from_p = strstr(from_p, "\r\n")))
1506 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1510 if ((newsize += chunksize) >= size)
1512 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
1517 memmove(to_p, from_p, (size_t) chunksize);
1518 to_p = buffer + newsize;
1519 from_p += chunksize + 2;
1521 if (sscanf(from_p, "%x", &chunksize) != 1)
1523 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1528 /* FIXME: Should this get its own loglevel? */
1529 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1535 /*********************************************************************
1537 * Function : url_actions
1539 * Description : Gets the actions for this URL.
1542 * 1 : http = http_request request for blocked URLs
1543 * 2 : csp = Current client state (buffers, headers, etc...)
1547 *********************************************************************/
1548 void url_actions(struct http_request *http,
1549 struct client_state *csp)
1551 struct file_list *fl;
1552 struct url_actions *b;
1555 init_current_action(csp->action);
1557 for (i = 0; i < MAX_ACTION_FILES; i++)
1559 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1564 apply_url_actions(csp->action, http, b);
1571 /*********************************************************************
1573 * Function : apply_url_actions
1575 * Description : Applies a list of URL actions.
1578 * 1 : action = Destination.
1579 * 2 : http = Current URL
1580 * 3 : b = list of URL actions to apply
1584 *********************************************************************/
1585 void apply_url_actions(struct current_action_spec *action,
1586 struct http_request *http,
1587 struct url_actions *b)
1591 /* Should never happen */
1595 for (b = b->next; NULL != b; b = b->next)
1597 if (url_match(b->url, http))
1599 merge_current_action(action, b->action);
1605 /*********************************************************************
1607 * Function : forward_url
1609 * Description : Should we forward this to another proxy?
1612 * 1 : http = http_request request for current URL
1613 * 2 : csp = Current client state (buffers, headers, etc...)
1615 * Returns : Pointer to forwarding information.
1617 *********************************************************************/
1618 const struct forward_spec * forward_url(struct http_request *http,
1619 struct client_state *csp)
1621 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1622 struct forward_spec *fwd = csp->config->forward;
1631 if (url_match(fwd->url, http))