1 const char filters_rcs[] = "$Id: filters.c,v 2.4 2003/01/21 02:49:27 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 2.4 2003/01/21 02:49:27 david__schmidt
42 * Developer TODO 612294: src: C++ keyword as variable name
43 * I changed all ocurrences of 'new' to 'new_something' wherever I found
44 * one. I also brought up all the source files in MSDEV to see if I could
45 * spot any highlighted keywords that really were variables. Non-scientific,
46 * but at least I tried. :-)
48 * Revision 2.3 2002/12/28 03:58:19 david__schmidt
49 * Initial drop of dashboard instrumentation - enabled with
50 * --enable-activity-console
52 * Revision 2.2 2002/09/04 15:38:24 oes
53 * Synced with the stable branch:
54 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
55 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
57 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
58 * - Bugfix: Executing a filters without jobs no longer results in
59 * turing off *all* filters.
60 * - Security fix: Malicious web servers can't cause a seg fault
61 * through bogus chunk sizes anymore
63 * Revision 2.1 2002/08/26 11:08:18 sarantis
66 * Revision 2.0 2002/06/04 14:34:21 jongfoster
67 * Moving source files to src/
69 * Revision 1.58 2002/04/24 02:11:17 oes
70 * Jon's multiple AF patch: url_actions now evaluates rules
73 * Revision 1.57 2002/04/08 20:38:34 swa
76 * Revision 1.56 2002/04/05 15:51:24 oes
77 * - bugfix: error-pages now get correct request protocol
78 * - fix for invalid HTML in trust info
80 * Revision 1.55 2002/04/02 16:13:51 oes
81 * Fix: No "Go there anyway" for SSL
83 * Revision 1.54 2002/04/02 14:55:56 oes
84 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
86 * Revision 1.53 2002/03/26 22:29:54 swa
87 * we have a new homepage!
89 * Revision 1.52 2002/03/24 16:35:57 jongfoster
92 * Revision 1.51 2002/03/24 15:23:33 jongfoster
95 * Revision 1.50 2002/03/24 13:25:43 swa
96 * name change related issues
98 * Revision 1.49 2002/03/16 20:29:14 oes
101 * Revision 1.48 2002/03/13 20:25:34 oes
102 * Better logging for content filters
104 * Revision 1.47 2002/03/13 00:30:52 jongfoster
106 * Added option of always sending redirect for imageblock,
107 * currently disabled with #if 0.
109 * Revision 1.46 2002/03/12 01:42:49 oes
110 * Introduced modular filters
112 * Revision 1.45 2002/03/08 16:47:50 oes
113 * Added choice beween GIF and PNG built-in images
115 * Revision 1.44 2002/03/07 03:49:31 oes
116 * - Fixed compiler warnings etc
117 * - Changed built-in images from GIF to PNG
118 * (with regard to Unisys patent issue)
119 * - Added a 4x4 pattern PNG which is less intrusive
120 * than the logo but also clearly marks the deleted banners
122 * Revision 1.43 2002/01/22 23:51:59 jongfoster
123 * Replacing strsav() with the safer string_append().
125 * Adding missing html_encode() to error message generators. Where encoded
126 * and unencoded versions of a string were provided, removing the unencoded
129 * Revision 1.42 2002/01/17 21:00:32 jongfoster
130 * Moving all our URL and URL pattern parsing code to urlmatch.c.
132 * Using a single, simple url_match(pattern,url) function - rather than
133 * the 3-line match routine which was repeated all over the place.
135 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
137 * Using parse_http_url() to parse URLs without faking a HTTP
138 * request line for parse_http_request().
140 * Revision 1.41 2001/11/13 00:14:07 jongfoster
141 * Fixing stupid bug now I've figured out what || means.
142 * (It always returns 0 or 1, not one of it's paramaters.)
144 * Revision 1.40 2001/10/26 17:37:55 oes
145 * - Re-enabled Netscape 200/404 bug workaround in block_url():
146 * - Removed OS/2 special case
147 * - Made block_url() independant from sed() having been run
148 * - Made trust_url independant from sed() having been run
149 * - Made is_imageurl independant from sed() having been run.
150 * It now checks User-Agent: and Accept: by itself.
153 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
154 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
155 * threads to call select() simultaneously. So, it's time to do a real, live,
156 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
157 * (native). Both versions will work, but using __OS2__ offers multi-threading.
159 * Revision 1.38 2001/10/23 21:32:33 jongfoster
160 * Adding error-checking to selected functions
162 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
163 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
164 * filters.c. Added a FIXME in front of the offending code. I'll gladly
165 * put in a better/more robust fix for all parties if one is presented...
166 * It seems that just returning 200 instead of 404 would pretty much fix
167 * it for everyone, but I don't know all the history of the problem.
169 * Revision 1.36 2001/10/10 16:44:16 oes
170 * Added match_portlist function
172 * Revision 1.35 2001/10/07 15:41:23 oes
173 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
175 * New function remove_chunked_transfer_coding that strips chunked
176 * transfer coding to plain and is called by pcrs_filter_response
177 * and gif_deanimate_response if neccessary
179 * Improved handling of zero-change re_filter runs
181 * pcrs_filter_response and gif_deanimate_response now remove
182 * chunked transfer codeing before processing the body.
184 * Revision 1.34 2001/09/20 15:49:36 steudten
186 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
187 * See cgi.c fill_template().
189 * Revision 1.33 2001/09/16 17:05:14 jongfoster
190 * Removing unused #include showarg.h
192 * Revision 1.32 2001/09/16 13:21:27 jongfoster
193 * Changes to use new list functions.
195 * Revision 1.31 2001/09/16 11:38:02 jongfoster
196 * Splitting fill_template() into 2 functions:
197 * template_load() loads the file
198 * template_fill() performs the PCRS regexps.
199 * This is because the CGI edit interface has a "table row"
200 * template which is used many times in the page - this
201 * change means it's only loaded from disk once.
203 * Revision 1.30 2001/09/16 11:00:10 jongfoster
204 * New function alloc_http_response, for symmetry with free_http_response
206 * Revision 1.29 2001/09/13 23:32:40 jongfoster
207 * Moving image data to cgi.c rather than cgi.h
208 * Fixing a GPF under Win32 (and any other OS that protects global
209 * constants from being written to).
211 * Revision 1.28 2001/09/10 10:18:51 oes
212 * Silenced compiler warnings
214 * Revision 1.27 2001/08/05 16:06:20 jongfoster
215 * Modifiying "struct map" so that there are now separate header and
216 * "map_entry" structures. This means that functions which modify a
217 * map no longer need to return a pointer to the modified map.
218 * Also, it no longer reverses the order of the entries (which may be
219 * important with some advanced template substitutions).
221 * Revision 1.26 2001/07/30 22:08:36 jongfoster
222 * Tidying up #defines:
223 * - All feature #defines are now of the form FEATURE_xxx
224 * - Permanently turned off WIN_GUI_EDIT
225 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
227 * Revision 1.25 2001/07/26 10:09:46 oes
228 * Made browser detection a little less naive
230 * Revision 1.24 2001/07/25 17:22:51 oes
231 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
233 * Revision 1.23 2001/07/23 13:40:12 oes
234 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
236 * Revision 1.22 2001/07/18 12:29:34 oes
237 * - Made gif_deanimate_response respect
238 * csp->action->string[ACTION_STRING_DEANIMATE]
239 * - Logging cosmetics
241 * Revision 1.21 2001/07/13 13:59:53 oes
242 * - Introduced gif_deanimate_response which shares the
243 * generic content modification interface of pcrs_filter_response
244 * and acts as a wrapper to deanimate.c:gif_deanimate()
245 * - Renamed re_process_buffer to pcrs_filter_response
246 * - pcrs_filter_response now returns NULL on failure
247 * - Removed all #ifdef PCRS
249 * Revision 1.20 2001/07/01 17:01:04 oes
250 * Added comments and missing return statement in is_untrusted_url()
252 * Revision 1.19 2001/06/29 21:45:41 oes
253 * Indentation, CRLF->LF, Tab-> Space
255 * Revision 1.18 2001/06/29 13:27:38 oes
256 * - Cleaned up, renamed and reorderd functions
257 * and improved comments
260 * - Ported to CGI platform. Now delivers
261 * http_response or NULL
262 * - Unified HTML and GIF generation (moved image detection
263 * and GIF generation here from jcc.c:chat())
264 * - Fixed HTTP status to:
265 * - 403 (Forbidden) for the "blocked" HTML message
266 * - 200 (OK) for GIF answers
267 * - 302 (Redirect) for redirect to GIF
270 * - Ported to CGI platform. Now delivers
271 * http_response or NULL
272 * - Separated detection of untrusted URL into
273 * (bool)is_untrusted_url
274 * - Added enforcement of untrusted requests
276 * - Moved redirect_url() from cgi.c to here
277 * and ported it to the CGI platform
279 * - Removed logentry from cancelled commit
281 * Revision 1.17 2001/06/09 10:55:28 jongfoster
282 * Changing BUFSIZ ==> BUFFER_SIZE
284 * Revision 1.16 2001/06/07 23:10:26 jongfoster
285 * Allowing unanchored domain patterns to back off and retry
286 * if they partially match. Optimized right-anchored patterns.
287 * Moving ACL and forward files into config file.
288 * Replacing struct gateway with struct forward_spec
290 * Revision 1.15 2001/06/03 19:12:00 oes
291 * extracted-CGI relevant stuff
293 * Revision 1.14 2001/06/01 10:30:55 oes
294 * Added optional left-anchoring to domaincmp
296 * Revision 1.13 2001/05/31 21:21:30 jongfoster
297 * Permissionsfile / actions file changes:
298 * - Changed "permission" to "action" throughout
299 * - changes to file format to allow string parameters
300 * - Moved helper functions to actions.c
302 * Revision 1.12 2001/05/31 17:35:20 oes
304 * - Enhanced domain part globbing with infix and prefix asterisk
305 * matching and optional unanchored operation
307 * Revision 1.11 2001/05/29 11:53:23 oes
308 * "See why" link added to "blocked" page
310 * Revision 1.10 2001/05/29 09:50:24 jongfoster
311 * Unified blocklist/imagelist/permissionslist.
312 * File format is still under discussion, but the internal changes
315 * Also modified interceptor behaviour:
316 * - We now intercept all URLs beginning with one of the following
317 * prefixes (and *only* these prefixes):
319 * * http://ijbswa.sf.net/config/
320 * * http://ijbswa.sourceforge.net/config/
321 * - New interceptors "home page" - go to http://i.j.b/ to see it.
322 * - Internal changes so that intercepted and fast redirect pages
323 * are not replaced with an image.
324 * - Interceptors now have the option to send a binary page direct
325 * to the client. (i.e. ijb-send-banner uses this)
326 * - Implemented show-url-info interceptor. (Which is why I needed
327 * the above interceptors changes - a typical URL is
328 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
329 * The previous mechanism would not have intercepted that, and
330 * if it had been intercepted then it then it would have replaced
333 * Revision 1.9 2001/05/27 22:17:04 oes
335 * - re_process_buffer no longer writes the modified buffer
336 * to the client, which was very ugly. It now returns the
337 * buffer, which it is then written by chat.
339 * - content_length now adjusts the Content-Length: header
340 * for modified documents rather than crunch()ing it.
341 * (Length info in csp->content_length, which is 0 for
342 * unmodified documents)
344 * - For this to work, sed() is called twice when filtering.
346 * Revision 1.8 2001/05/26 17:13:28 jongfoster
347 * Filled in a function comment.
349 * Revision 1.7 2001/05/26 15:26:15 jongfoster
350 * ACL feature now provides more security by immediately dropping
351 * connections from untrusted hosts.
353 * Revision 1.6 2001/05/26 00:28:36 jongfoster
354 * Automatic reloading of config file.
355 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
356 * Most of the global variables have been moved to a new
357 * struct configuration_spec, accessed through csp->config->globalname
358 * Most of the globals remaining are used by the Win32 GUI.
360 * Revision 1.5 2001/05/25 22:34:30 jongfoster
363 * Revision 1.4 2001/05/22 18:46:04 oes
365 * - Enabled filtering banners by size rather than URL
366 * by adding patterns that replace all standard banner
367 * sizes with the "Junkbuster" gif to the re_filterfile
369 * - Enabled filtering WebBugs by providing a pattern
370 * which kills all 1x1 images
372 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
373 * which is selected by the (nonstandard and therefore
374 * capital) letter 'U' in the option string.
375 * It causes the quantifiers to be ungreedy by default.
376 * Appending a ? turns back to greedy (!).
378 * - Added a new interceptor ijb-send-banner, which
379 * sends back the "Junkbuster" gif. Without imagelist or
380 * MSIE detection support, or if tinygif = 1, or the
381 * URL isn't recognized as an imageurl, a lame HTML
382 * explanation is sent instead.
384 * - Added new feature, which permits blocking remote
385 * script redirects and firing back a local redirect
387 * The feature is conditionally compiled, i.e. it
388 * can be disabled with --disable-fast-redirects,
389 * plus it must be activated by a "fast-redirects"
390 * line in the config file, has its own log level
391 * and of course wants to be displayed by show-proxy-args
392 * Note: Boy, all the #ifdefs in 1001 locations and
393 * all the fumbling with configure.in and acconfig.h
394 * were *way* more work than the feature itself :-(
396 * - Because a generic redirect template was needed for
397 * this, tinygif = 3 now uses the same.
399 * - Moved GIFs, and other static HTTP response templates
404 * - Removed some >400 CRs again (Jon, you really worked
407 * Revision 1.3 2001/05/20 16:44:47 jongfoster
408 * Removing last hardcoded Junkbusters.com URLs.
410 * Revision 1.2 2001/05/20 01:21:20 jongfoster
411 * Version 2.9.4 checkin.
412 * - Merged popupfile and cookiefile, and added control over PCRS
413 * filtering, in new "permissionsfile".
414 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
415 * file error you now get a message box (in the Win32 GUI) rather
416 * than the program exiting with no explanation.
417 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
419 * - Removed tabs from "config"
420 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
421 * - Bumped up version number.
423 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
424 * Initial import of version 2.9.3 source tree
427 *********************************************************************/
433 #include <sys/types.h>
442 #endif /* ndef __OS2__ */
443 #include <netinet/in.h>
445 #include <winsock2.h>
446 #endif /* ndef _WIN32 */
450 #endif /* def __OS2__ */
458 #include "jbsockets.h"
459 #include "miscutil.h"
463 #include "deanimate.h"
464 #include "urlmatch.h"
465 #ifdef FEATURE_ACTIVITY_CONSOLE
467 #endif /* def FEATURE_ACTIVITY_CONSOLE */
472 const char filters_h_rcs[] = FILTERS_H_VERSION;
474 /* Fix a problem with Solaris. There should be no effect on other
476 * Solaris's isspace() is a macro which uses it's argument directly
477 * as an array index. Therefore we need to make sure that high-bit
478 * characters generate +ve values, and ideally we also want to make
479 * the argument match the declared parameter type of "int".
481 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
485 /*********************************************************************
487 * Function : block_acl
489 * Description : Block this request?
490 * Decide yes or no based on ACL file.
493 * 1 : dst = The proxy or gateway address this is going to.
494 * Or NULL to check all possible targets.
495 * 2 : csp = Current client state (buffers, headers, etc...)
496 * Also includes the client IP address.
498 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
500 *********************************************************************/
501 int block_acl(struct access_control_addr *dst, struct client_state *csp)
503 struct access_control_list *acl = csp->config->acl;
505 /* if not using an access control list, then permit the connection */
511 /* search the list */
514 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
518 /* Just want to check if they have any access */
519 if (acl->action == ACL_PERMIT)
524 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
525 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
527 if (acl->action == ACL_PERMIT)
533 #ifdef FEATURE_ACTIVITY_CONSOLE
534 accumulate_stats(STATS_ACL_RESTRICT, 1);
535 #endif /* def FEATURE_ACTIVITY_CONSOLE */
548 /*********************************************************************
550 * Function : acl_addr
552 * Description : Called from `load_config' to parse an ACL address.
555 * 1 : aspec = String specifying ACL address.
556 * 2 : aca = struct access_control_addr to fill in.
558 * Returns : 0 => Ok, everything else is an error.
560 *********************************************************************/
561 int acl_addr(char *aspec, struct access_control_addr *aca)
563 int i, masklength, port;
569 if ((p = strchr(aspec, '/')) != NULL)
573 if (ijb_isdigit(*p) == 0)
577 masklength = atoi(p);
580 if ((masklength < 0) || (masklength > 32))
585 if ((p = strchr(aspec, ':')) != NULL)
589 if (ijb_isdigit(*p) == 0)
598 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
600 if (aca->addr == INADDR_NONE)
605 /* build the netmask */
607 for (i=1; i <= masklength ; i++)
609 aca->mask |= (1 << (32 - i));
612 /* now mask off the host portion of the ip address
613 * (i.e. save on the network portion of the address).
615 aca->addr = aca->addr & aca->mask;
620 #endif /* def FEATURE_ACL */
623 /*********************************************************************
625 * Function : match_portlist
627 * Description : Check if a given number is covered by a comma
628 * separated list of numbers and ranges (a,b-c,d,..)
631 * 1 : portlist = String with list
632 * 2 : port = port to check
634 * Returns : 0 => no match
637 *********************************************************************/
638 int match_portlist(const char *portlist, int port)
640 char *min, *max, *next, *portlist_copy;
642 min = next = portlist_copy = strdup(portlist);
645 * Zero-terminate first item and remember offset for next
647 if (NULL != (next = strchr(portlist_copy, (int) ',')))
653 * Loop through all items, checking for match
657 if (NULL == (max = strchr(min, (int) '-')))
660 * No dash, check for equality
662 if (port == atoi(min))
671 * This is a range, so check if between min and max,
672 * or, if max was omitted, between min and 65K
675 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
689 * Zero-terminate next item and remember offset for n+1
691 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
703 /*********************************************************************
705 * Function : block_url
707 * Description : Called from `chat'. Check to see if we need to block this.
710 * 1 : csp = Current client state (buffers, headers, etc...)
712 * Returns : NULL => unblocked, else HTTP block response
714 *********************************************************************/
715 struct http_response *block_url(struct client_state *csp)
717 #ifdef FEATURE_IMAGE_BLOCKING
719 #endif /* def FEATURE_IMAGE_BLOCKING */
720 struct http_response *rsp;
723 * If it's not blocked, don't block it ;-)
725 if ((csp->action->flags & ACTION_BLOCK) == 0)
731 * Else, prepare a response
733 if (NULL == (rsp = alloc_http_response()))
735 return cgi_error_memory();
739 * If it's an image-url, send back an image or redirect
740 * as specified by the relevant +image action
742 #ifdef FEATURE_IMAGE_BLOCKING
743 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
746 /* determine HOW images should be blocked */
747 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
749 #if 1 /* Two alternative strategies, use this one for now: */
751 /* and handle accordingly: */
752 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
754 #ifdef FEATURE_ACTIVITY_CONSOLE
755 accumulate_stats(STATS_IMAGE_BLOCK, 1);
756 #endif /* def FEATURE_ACTIVITY_CONSOLE */
757 rsp->body = bindup(image_pattern_data, image_pattern_length);
758 if (rsp->body == NULL)
760 free_http_response(rsp);
761 return cgi_error_memory();
763 rsp->content_length = image_pattern_length;
765 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
767 free_http_response(rsp);
768 return cgi_error_memory();
772 else if (0 == strcmpic(p, "blank"))
774 #ifdef FEATURE_ACTIVITY_CONSOLE
775 accumulate_stats(STATS_IMAGE_BLOCK, 1);
776 #endif /* def FEATURE_ACTIVITY_CONSOLE */
777 rsp->body = bindup(image_blank_data, image_blank_length);
778 if (rsp->body == NULL)
780 free_http_response(rsp);
781 return cgi_error_memory();
783 rsp->content_length = image_blank_length;
785 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
787 free_http_response(rsp);
788 return cgi_error_memory();
792 else if (0 == strcmpic(p, "custom"))
794 rsp->body = bindup(csp->config->image_blocker_data, csp->config->image_blocker_length);
795 if (rsp->body == NULL)
797 free_http_response(rsp);
798 return cgi_error_memory();
800 rsp->content_length = csp->config->image_blocker_length;
802 if (enlist_unique_header(rsp->headers, "Content-Type", csp->config->image_blocker_format))
804 free_http_response(rsp);
805 return cgi_error_memory();
811 rsp->status = strdup("302 Local Redirect from Privoxy");
812 if (rsp->status == NULL)
814 free_http_response(rsp);
815 return cgi_error_memory();
818 if (enlist_unique_header(rsp->headers, "Location", p))
820 free_http_response(rsp);
821 return cgi_error_memory();
825 #else /* Following code is disabled for now */
827 /* and handle accordingly: */
828 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
830 p = CGI_PREFIX "send-banner?type=pattern";
832 else if (0 == strcmpic(p, "blank"))
834 p = CGI_PREFIX "send-banner?type=blank";
836 rsp->status = strdup("302 Local Redirect from Privoxy");
837 if (rsp->status == NULL)
839 free_http_response(rsp);
840 return cgi_error_memory();
843 if (enlist_unique_header(rsp->headers, "Location", p))
845 free_http_response(rsp);
846 return cgi_error_memory();
848 #endif /* Preceeding code is disabled for now */
851 #endif /* def FEATURE_IMAGE_BLOCKING */
854 * Else, generate an HTML "blocked" message:
858 struct map * exports;
861 * Workaround for stupid Netscape bug which prevents
862 * pages from being displayed if loading a referenced
863 * JavaScript or style sheet fails. So make it appear
864 * as if it succeeded.
866 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
867 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
868 && !strstr(p, "Gecko") /* save Mozilla, */
869 && !strstr(p, "compatible") /* MSIE */
870 && !strstr(p, "Opera")) /* and Opera. */
872 rsp->status = strdup("200 Request for blocked URL");
876 rsp->status = strdup("404 Request for blocked URL");
879 if (rsp->status == NULL)
881 free_http_response(rsp);
882 return cgi_error_memory();
885 exports = default_exports(csp, NULL);
888 free_http_response(rsp);
889 return cgi_error_memory();
892 #ifdef FEATURE_FORCE_LOAD
893 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
894 if (csp->http->ssl != 0)
895 #endif /* ndef FEATURE_FORCE_LOAD */
897 err = map_block_killer(exports, "force-support");
900 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
901 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
902 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
907 free_http_response(rsp);
908 return cgi_error_memory();
911 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
914 free_http_response(rsp);
915 return cgi_error_memory();
919 return finish_http_response(rsp);
925 /*********************************************************************
927 * Function : trust_url FIXME: I should be called distrust_url
929 * Description : Calls is_untrusted_url to determine if the URL is trusted
930 * and if not, returns a HTTP 304 response with a reject message.
933 * 1 : csp = Current client state (buffers, headers, etc...)
935 * Returns : NULL => trusted, else http_response.
937 *********************************************************************/
938 struct http_response *trust_url(struct client_state *csp)
940 struct http_response *rsp;
941 struct map * exports;
942 char buf[BUFFER_SIZE];
944 struct url_spec **tl;
949 * Don't bother to work on trusted URLs
951 if (!is_untrusted_url(csp))
957 * Else, prepare a response:
959 if (NULL == (rsp = alloc_http_response()))
961 return cgi_error_memory();
964 exports = default_exports(csp, NULL);
967 free_http_response(rsp);
968 return cgi_error_memory();
972 * Export the protocol, host, port, and referrer information
974 err = map(exports, "hostport", 1, csp->http->hostport, 1);
975 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
976 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
978 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
980 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
984 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
990 free_http_response(rsp);
991 return cgi_error_memory();
995 * Export the trust list
998 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1000 sprintf(buf, "<li>%s</li>\n", t->spec);
1001 string_append(&p, buf);
1003 err = map(exports, "trusted-referrers", 1, p, 0);
1008 free_http_response(rsp);
1009 return cgi_error_memory();
1013 * Export the trust info, if available
1015 if (csp->config->trust_info->first)
1017 struct list_entry *l;
1020 for (l = csp->config->trust_info->first; l ; l = l->next)
1022 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
1023 string_append(&p, buf);
1025 err = map(exports, "trust-info", 1, p, 0);
1029 err = map_block_killer(exports, "have-trust-info");
1035 free_http_response(rsp);
1036 return cgi_error_memory();
1040 * Export the force prefix or the force conditional block killer
1042 #ifdef FEATURE_FORCE_LOAD
1043 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1044 #else /* ifndef FEATURE_FORCE_LOAD */
1045 err = map_block_killer(exports, "force-support");
1046 #endif /* ndef FEATURE_FORCE_LOAD */
1051 free_http_response(rsp);
1052 return cgi_error_memory();
1056 * Build the response
1058 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1061 free_http_response(rsp);
1062 return cgi_error_memory();
1065 return finish_http_response(rsp);
1067 #endif /* def FEATURE_TRUST */
1070 #ifdef FEATURE_FAST_REDIRECTS
1071 /*********************************************************************
1073 * Function : redirect_url
1075 * Description : Checks for redirection URLs and returns a HTTP redirect
1076 * to the destination URL, if necessary
1079 * 1 : csp = Current client state (buffers, headers, etc...)
1081 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1083 *********************************************************************/
1084 struct http_response *redirect_url(struct client_state *csp)
1087 struct http_response *rsp;
1089 p = q = csp->http->path;
1090 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1093 * find the last URL encoded in the request
1095 while ((p = strstr(p, "http://")) != NULL)
1101 * if there was any, generate and return a HTTP redirect
1103 if (q != csp->http->path)
1105 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1107 if (NULL == (rsp = alloc_http_response()))
1109 return cgi_error_memory();
1112 if ( enlist_unique_header(rsp->headers, "Location", q)
1113 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1115 free_http_response(rsp);
1116 return cgi_error_memory();
1119 return finish_http_response(rsp);
1127 #endif /* def FEATURE_FAST_REDIRECTS */
1130 #ifdef FEATURE_IMAGE_BLOCKING
1131 /*********************************************************************
1133 * Function : is_imageurl
1135 * Description : Given a URL, decide whether it is an image or not,
1136 * using either the info from a previous +image action
1137 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1138 * is MSIE and not on a Mac, tell from the browser's accept
1142 * 1 : csp = Current client state (buffers, headers, etc...)
1144 * Returns : True (nonzero) if URL is an image, false (0)
1147 *********************************************************************/
1148 int is_imageurl(struct client_state *csp)
1150 #ifdef FEATURE_IMAGE_DETECT_MSIE
1153 tmp = get_header_value(csp->headers, "User-Agent:");
1154 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1156 tmp = get_header_value(csp->headers, "Accept:");
1157 if (tmp && strstr(tmp, "image/gif"))
1159 /* Client will accept HTML. If this seems counterintuitive,
1169 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1171 return ((csp->action->flags & ACTION_IMAGE) != 0);
1174 #endif /* def FEATURE_IMAGE_BLOCKING */
1177 #ifdef FEATURE_TRUST
1178 /*********************************************************************
1180 * Function : is_untrusted_url
1182 * Description : Should we "distrust" this URL (and block it)?
1184 * Yes if it matches a line in the trustfile, or if the
1185 * referrer matches a line starting with "+" in the
1190 * 1 : csp = Current client state (buffers, headers, etc...)
1192 * Returns : 0 => trusted, 1 => untrusted
1194 *********************************************************************/
1195 int is_untrusted_url(struct client_state *csp)
1197 struct file_list *fl;
1198 struct block_spec *b;
1199 struct url_spec **trusted_url;
1200 struct http_request rhttp[1];
1201 const char * referer;
1205 * If we don't have a trustlist, we trust everybody
1207 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1212 memset(rhttp, '\0', sizeof(*rhttp));
1215 * Do we trust the request URL itself?
1217 for (b = b->next; b ; b = b->next)
1219 if (url_match(b->url, csp->http))
1225 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1227 /* no referrer was supplied */
1233 * If not, do we maybe trust its referrer?
1235 err = parse_http_url(referer, rhttp, csp);
1241 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1243 if (url_match(*trusted_url, rhttp))
1245 /* if the URL's referrer is from a trusted referrer, then
1246 * add the target spec to the trustfile as an unblocked
1247 * domain and return NULL (which means it's OK).
1252 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1256 char * new_entry = strdup("~");
1258 string_append(&new_entry, csp->http->hostport);
1260 path = csp->http->path;
1261 if ( (path[0] == '/')
1263 && ((path_end = strchr(path + 2, '/')) != NULL))
1265 /* since this path points into a user's home space
1266 * be sure to include this spec in the trustfile.
1268 int path_len = path_end - path; /* save offset */
1269 path = strdup(path); /* Copy string */
1272 path_end = path + path_len; /* regenerate ptr to new buffer */
1273 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1275 string_join(&new_entry, path);
1278 if (new_entry != NULL)
1280 fprintf(fp, "%s\n", new_entry);
1285 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1286 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1296 #endif /* def FEATURE_TRUST */
1299 /*********************************************************************
1301 * Function : pcrs_filter_response
1303 * Description : Ecexute all text substitutions from all applying
1304 * +filter actions on the text buffer that's been accumulated
1305 * in csp->iob->buf. If this changes the contents, set
1306 * csp->content_length to the modified size and raise the
1307 * CSP_FLAG_MODIFIED flag.
1310 * 1 : csp = Current client state (buffers, headers, etc...)
1312 * Returns : a pointer to the (newly allocated) modified buffer.
1313 * or NULL if there were no hits or something went wrong
1315 *********************************************************************/
1316 char *pcrs_filter_response(struct client_state *csp)
1321 char *old_buf = csp->iob->cur, *new_buf = NULL;
1324 struct file_list *fl;
1325 struct re_filterfile_spec *b;
1326 struct list_entry *filtername;
1331 if (csp->iob->cur >= csp->iob->eod)
1335 size = csp->iob->eod - csp->iob->cur;
1337 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1339 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1344 * If the body has a "chunked" transfer-encoding,
1345 * get rid of it first, adjusting size and iob->eod
1347 if (csp->flags & CSP_FLAG_CHUNKED)
1349 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1350 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1354 csp->iob->eod = csp->iob->cur + size;
1355 csp->flags |= CSP_FLAG_MODIFIED;
1359 * For all applying +filter actions, look if a filter by that
1360 * name exists and if yes, execute it's pcrs_joblist on the
1363 for (b = fl->f; b; b = b->next)
1365 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1366 filtername ; filtername = filtername->next)
1368 if (strcmp(b->name, filtername->str) == 0)
1370 int current_hits = 0;
1372 if ( NULL == b->joblist )
1374 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1378 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1379 csp->http->hostport, csp->http->path, size, b->name);
1381 /* Apply all jobs from the joblist */
1382 for (job = b->joblist; NULL != job; job = job->next)
1384 current_hits += pcrs_execute(job, old_buf, size, &new_buf, &size);
1385 if (old_buf != csp->iob->cur) free(old_buf);
1389 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1390 hits += current_hits;
1396 * If there were no hits, destroy our copy and let
1397 * chat() use the original in csp->iob
1404 #ifdef FEATURE_ACTIVITY_CONSOLE
1406 accumulate_stats(STATS_FILTER, hits);
1407 #endif /* def FEATURE_ACTIVITY_CONSOLE */
1409 csp->flags |= CSP_FLAG_MODIFIED;
1410 csp->content_length = size;
1418 /*********************************************************************
1420 * Function : gif_deanimate_response
1422 * Description : Deanimate the GIF image that has been accumulated in
1423 * csp->iob->buf, set csp->content_length to the modified
1424 * size and raise the CSP_FLAG_MODIFIED flag.
1427 * 1 : csp = Current client state (buffers, headers, etc...)
1429 * Returns : a pointer to the (newly allocated) modified buffer.
1430 * or NULL in case something went wrong.
1432 *********************************************************************/
1433 char *gif_deanimate_response(struct client_state *csp)
1435 struct binbuffer *in, *out;
1437 size_t size = csp->iob->eod - csp->iob->cur;
1440 * If the body has a "chunked" transfer-encoding,
1441 * get rid of it first, adjusting size and iob->eod
1443 if (csp->flags & CSP_FLAG_CHUNKED)
1445 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1446 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1450 csp->iob->eod = csp->iob->cur + size;
1451 csp->flags |= CSP_FLAG_MODIFIED;
1454 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1455 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1457 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1461 in->buffer = csp->iob->cur;
1464 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1466 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1473 if ((int)size == out->offset)
1475 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1479 #ifdef FEATURE_ACTIVITY_CONSOLE
1480 accumulate_stats(STATS_GIF_DEANIMATE, 1);
1481 #endif /* def FEATURE_ACTIVITY_CONSOLE */
1482 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1484 csp->content_length = out->offset;
1485 csp->flags |= CSP_FLAG_MODIFIED;
1495 /*********************************************************************
1497 * Function : remove_chunked_transfer_coding
1499 * Description : In-situ remove the "chunked" transfer coding as defined
1500 * in rfc2616 from a buffer.
1503 * 1 : buffer = Pointer to the text buffer
1504 * 2 : size = Number of bytes to be processed
1506 * Returns : The new size, i.e. the number of bytes from buffer which
1507 * are occupied by the stripped body, or 0 in case something
1510 *********************************************************************/
1511 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1514 unsigned int chunksize = 0;
1515 char *from_p, *to_p;
1518 from_p = to_p = buffer;
1520 if (sscanf(buffer, "%x", &chunksize) != 1)
1522 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1526 while (chunksize > 0)
1528 if (NULL == (from_p = strstr(from_p, "\r\n")))
1530 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1534 if ((newsize += chunksize) >= size)
1536 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
1541 memmove(to_p, from_p, (size_t) chunksize);
1542 to_p = buffer + newsize;
1543 from_p += chunksize + 2;
1545 if (sscanf(from_p, "%x", &chunksize) != 1)
1547 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1552 /* FIXME: Should this get its own loglevel? */
1553 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1559 /*********************************************************************
1561 * Function : url_actions
1563 * Description : Gets the actions for this URL.
1566 * 1 : http = http_request request for blocked URLs
1567 * 2 : csp = Current client state (buffers, headers, etc...)
1571 *********************************************************************/
1572 void url_actions(struct http_request *http,
1573 struct client_state *csp)
1575 struct file_list *fl;
1576 struct url_actions *b;
1579 init_current_action(csp->action);
1581 for (i = 0; i < MAX_ACTION_FILES; i++)
1583 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1588 apply_url_actions(csp->action, http, b);
1595 /*********************************************************************
1597 * Function : apply_url_actions
1599 * Description : Applies a list of URL actions.
1602 * 1 : action = Destination.
1603 * 2 : http = Current URL
1604 * 3 : b = list of URL actions to apply
1608 *********************************************************************/
1609 void apply_url_actions(struct current_action_spec *action,
1610 struct http_request *http,
1611 struct url_actions *b)
1615 /* Should never happen */
1619 for (b = b->next; NULL != b; b = b->next)
1621 if (url_match(b->url, http))
1623 merge_current_action(action, b->action);
1629 /*********************************************************************
1631 * Function : forward_url
1633 * Description : Should we forward this to another proxy?
1636 * 1 : http = http_request request for current URL
1637 * 2 : csp = Current client state (buffers, headers, etc...)
1639 * Returns : Pointer to forwarding information.
1641 *********************************************************************/
1642 const struct forward_spec * forward_url(struct http_request *http,
1643 struct client_state *csp)
1645 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1646 struct forward_spec *fwd = csp->config->forward;
1655 if (url_match(fwd->url, http))