1 const char filters_rcs[] = "$Id: filters.c,v 1.60 2006/07/18 14:48:46 david__schmidt Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/filters.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
7 * Functions declared include:
8 * `acl_addr', `add_stats', `block_acl', `block_imageurl',
9 * `block_url', `url_actions', `domain_split',
10 * `filter_popups', `forward_url', 'redirect_url',
11 * `ij_untrusted_url', `intercept_url', `pcrs_filter_respose',
12 * `ijb_send_banner', `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 1.60 2006/07/18 14:48:46 david__schmidt
43 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
44 * with what was really the latest development (the v_3_0_branch branch)
46 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
47 * Multiple filter file support
49 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
50 * A few memory leaks plugged (mostly on error paths)
52 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
53 * Add the ability to check jpeg images for invalid
54 * lengths of comment blocks. Defensive strategy
55 * against the exploit:
56 * Microsoft Security Bulletin MS04-028
57 * Buffer Overrun in JPEG Processing (GDI+) Could
58 * Allow Code Execution (833987)
59 * Enabled with +inspect-jpegs in actions files.
61 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
62 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
64 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
65 * Fixed bug #839859: "See why" link URL now gets url-encoded.
67 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
70 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
71 * Added basic support for OPTIONS and TRACE HTTP methods:
72 * New function direct_response which handles OPTIONS and
73 * TRACE requests whose Max-Forwards header field is zero.
75 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
76 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
78 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
79 * - Bugfix: Executing a filters without jobs no longer results in
80 * turing off *all* filters.
81 * - Security fix: Malicious web servers can't cause a seg fault
82 * through bogus chunk sizes anymore
84 * Revision 1.58 2002/04/24 02:11:17 oes
85 * Jon's multiple AF patch: url_actions now evaluates rules
88 * Revision 1.57 2002/04/08 20:38:34 swa
91 * Revision 1.56 2002/04/05 15:51:24 oes
92 * - bugfix: error-pages now get correct request protocol
93 * - fix for invalid HTML in trust info
95 * Revision 1.55 2002/04/02 16:13:51 oes
96 * Fix: No "Go there anyway" for SSL
98 * Revision 1.54 2002/04/02 14:55:56 oes
99 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
101 * Revision 1.53 2002/03/26 22:29:54 swa
102 * we have a new homepage!
104 * Revision 1.52 2002/03/24 16:35:57 jongfoster
107 * Revision 1.51 2002/03/24 15:23:33 jongfoster
110 * Revision 1.50 2002/03/24 13:25:43 swa
111 * name change related issues
113 * Revision 1.49 2002/03/16 20:29:14 oes
116 * Revision 1.48 2002/03/13 20:25:34 oes
117 * Better logging for content filters
119 * Revision 1.47 2002/03/13 00:30:52 jongfoster
121 * Added option of always sending redirect for imageblock,
122 * currently disabled with #if 0.
124 * Revision 1.46 2002/03/12 01:42:49 oes
125 * Introduced modular filters
127 * Revision 1.45 2002/03/08 16:47:50 oes
128 * Added choice beween GIF and PNG built-in images
130 * Revision 1.44 2002/03/07 03:49:31 oes
131 * - Fixed compiler warnings etc
132 * - Changed built-in images from GIF to PNG
133 * (with regard to Unisys patent issue)
134 * - Added a 4x4 pattern PNG which is less intrusive
135 * than the logo but also clearly marks the deleted banners
137 * Revision 1.43 2002/01/22 23:51:59 jongfoster
138 * Replacing strsav() with the safer string_append().
140 * Adding missing html_encode() to error message generators. Where encoded
141 * and unencoded versions of a string were provided, removing the unencoded
144 * Revision 1.42 2002/01/17 21:00:32 jongfoster
145 * Moving all our URL and URL pattern parsing code to urlmatch.c.
147 * Using a single, simple url_match(pattern,url) function - rather than
148 * the 3-line match routine which was repeated all over the place.
150 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
152 * Using parse_http_url() to parse URLs without faking a HTTP
153 * request line for parse_http_request().
155 * Revision 1.41 2001/11/13 00:14:07 jongfoster
156 * Fixing stupid bug now I've figured out what || means.
157 * (It always returns 0 or 1, not one of it's paramaters.)
159 * Revision 1.40 2001/10/26 17:37:55 oes
160 * - Re-enabled Netscape 200/404 bug workaround in block_url():
161 * - Removed OS/2 special case
162 * - Made block_url() independant from sed() having been run
163 * - Made trust_url independant from sed() having been run
164 * - Made is_imageurl independant from sed() having been run.
165 * It now checks User-Agent: and Accept: by itself.
168 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
169 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
170 * threads to call select() simultaneously. So, it's time to do a real, live,
171 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
172 * (native). Both versions will work, but using __OS2__ offers multi-threading.
174 * Revision 1.38 2001/10/23 21:32:33 jongfoster
175 * Adding error-checking to selected functions
177 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
178 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
179 * filters.c. Added a FIXME in front of the offending code. I'll gladly
180 * put in a better/more robust fix for all parties if one is presented...
181 * It seems that just returning 200 instead of 404 would pretty much fix
182 * it for everyone, but I don't know all the history of the problem.
184 * Revision 1.36 2001/10/10 16:44:16 oes
185 * Added match_portlist function
187 * Revision 1.35 2001/10/07 15:41:23 oes
188 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
190 * New function remove_chunked_transfer_coding that strips chunked
191 * transfer coding to plain and is called by pcrs_filter_response
192 * and gif_deanimate_response if neccessary
194 * Improved handling of zero-change re_filter runs
196 * pcrs_filter_response and gif_deanimate_response now remove
197 * chunked transfer codeing before processing the body.
199 * Revision 1.34 2001/09/20 15:49:36 steudten
201 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
202 * See cgi.c fill_template().
204 * Revision 1.33 2001/09/16 17:05:14 jongfoster
205 * Removing unused #include showarg.h
207 * Revision 1.32 2001/09/16 13:21:27 jongfoster
208 * Changes to use new list functions.
210 * Revision 1.31 2001/09/16 11:38:02 jongfoster
211 * Splitting fill_template() into 2 functions:
212 * template_load() loads the file
213 * template_fill() performs the PCRS regexps.
214 * This is because the CGI edit interface has a "table row"
215 * template which is used many times in the page - this
216 * change means it's only loaded from disk once.
218 * Revision 1.30 2001/09/16 11:00:10 jongfoster
219 * New function alloc_http_response, for symmetry with free_http_response
221 * Revision 1.29 2001/09/13 23:32:40 jongfoster
222 * Moving image data to cgi.c rather than cgi.h
223 * Fixing a GPF under Win32 (and any other OS that protects global
224 * constants from being written to).
226 * Revision 1.28 2001/09/10 10:18:51 oes
227 * Silenced compiler warnings
229 * Revision 1.27 2001/08/05 16:06:20 jongfoster
230 * Modifiying "struct map" so that there are now separate header and
231 * "map_entry" structures. This means that functions which modify a
232 * map no longer need to return a pointer to the modified map.
233 * Also, it no longer reverses the order of the entries (which may be
234 * important with some advanced template substitutions).
236 * Revision 1.26 2001/07/30 22:08:36 jongfoster
237 * Tidying up #defines:
238 * - All feature #defines are now of the form FEATURE_xxx
239 * - Permanently turned off WIN_GUI_EDIT
240 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
242 * Revision 1.25 2001/07/26 10:09:46 oes
243 * Made browser detection a little less naive
245 * Revision 1.24 2001/07/25 17:22:51 oes
246 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
248 * Revision 1.23 2001/07/23 13:40:12 oes
249 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
251 * Revision 1.22 2001/07/18 12:29:34 oes
252 * - Made gif_deanimate_response respect
253 * csp->action->string[ACTION_STRING_DEANIMATE]
254 * - Logging cosmetics
256 * Revision 1.21 2001/07/13 13:59:53 oes
257 * - Introduced gif_deanimate_response which shares the
258 * generic content modification interface of pcrs_filter_response
259 * and acts as a wrapper to deanimate.c:gif_deanimate()
260 * - Renamed re_process_buffer to pcrs_filter_response
261 * - pcrs_filter_response now returns NULL on failiure
262 * - Removed all #ifdef PCRS
264 * Revision 1.20 2001/07/01 17:01:04 oes
265 * Added comments and missing return statement in is_untrusted_url()
267 * Revision 1.19 2001/06/29 21:45:41 oes
268 * Indentation, CRLF->LF, Tab-> Space
270 * Revision 1.18 2001/06/29 13:27:38 oes
271 * - Cleaned up, renamed and reorderd functions
272 * and improved comments
275 * - Ported to CGI platform. Now delivers
276 * http_response or NULL
277 * - Unified HTML and GIF generation (moved image detection
278 * and GIF generation here from jcc.c:chat())
279 * - Fixed HTTP status to:
280 * - 403 (Forbidden) for the "blocked" HTML message
281 * - 200 (OK) for GIF answers
282 * - 302 (Redirect) for redirect to GIF
285 * - Ported to CGI platform. Now delivers
286 * http_response or NULL
287 * - Separated detection of untrusted URL into
288 * (bool)is_untrusted_url
289 * - Added enforcement of untrusted requests
291 * - Moved redirect_url() from cgi.c to here
292 * and ported it to the CGI platform
294 * - Removed logentry from cancelled commit
296 * Revision 1.17 2001/06/09 10:55:28 jongfoster
297 * Changing BUFSIZ ==> BUFFER_SIZE
299 * Revision 1.16 2001/06/07 23:10:26 jongfoster
300 * Allowing unanchored domain patterns to back off and retry
301 * if they partially match. Optimized right-anchored patterns.
302 * Moving ACL and forward files into config file.
303 * Replacing struct gateway with struct forward_spec
305 * Revision 1.15 2001/06/03 19:12:00 oes
306 * extracted-CGI relevant stuff
308 * Revision 1.14 2001/06/01 10:30:55 oes
309 * Added optional left-anchoring to domaincmp
311 * Revision 1.13 2001/05/31 21:21:30 jongfoster
312 * Permissionsfile / actions file changes:
313 * - Changed "permission" to "action" throughout
314 * - changes to file format to allow string parameters
315 * - Moved helper functions to actions.c
317 * Revision 1.12 2001/05/31 17:35:20 oes
319 * - Enhanced domain part globbing with infix and prefix asterisk
320 * matching and optional unanchored operation
322 * Revision 1.11 2001/05/29 11:53:23 oes
323 * "See why" link added to "blocked" page
325 * Revision 1.10 2001/05/29 09:50:24 jongfoster
326 * Unified blocklist/imagelist/permissionslist.
327 * File format is still under discussion, but the internal changes
330 * Also modified interceptor behaviour:
331 * - We now intercept all URLs beginning with one of the following
332 * prefixes (and *only* these prefixes):
334 * * http://ijbswa.sf.net/config/
335 * * http://ijbswa.sourceforge.net/config/
336 * - New interceptors "home page" - go to http://i.j.b/ to see it.
337 * - Internal changes so that intercepted and fast redirect pages
338 * are not replaced with an image.
339 * - Interceptors now have the option to send a binary page direct
340 * to the client. (i.e. ijb-send-banner uses this)
341 * - Implemented show-url-info interceptor. (Which is why I needed
342 * the above interceptors changes - a typical URL is
343 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
344 * The previous mechanism would not have intercepted that, and
345 * if it had been intercepted then it then it would have replaced
348 * Revision 1.9 2001/05/27 22:17:04 oes
350 * - re_process_buffer no longer writes the modified buffer
351 * to the client, which was very ugly. It now returns the
352 * buffer, which it is then written by chat.
354 * - content_length now adjusts the Content-Length: header
355 * for modified documents rather than crunch()ing it.
356 * (Length info in csp->content_length, which is 0 for
357 * unmodified documents)
359 * - For this to work, sed() is called twice when filtering.
361 * Revision 1.8 2001/05/26 17:13:28 jongfoster
362 * Filled in a function comment.
364 * Revision 1.7 2001/05/26 15:26:15 jongfoster
365 * ACL feature now provides more security by immediately dropping
366 * connections from untrusted hosts.
368 * Revision 1.6 2001/05/26 00:28:36 jongfoster
369 * Automatic reloading of config file.
370 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
371 * Most of the global variables have been moved to a new
372 * struct configuration_spec, accessed through csp->config->globalname
373 * Most of the globals remaining are used by the Win32 GUI.
375 * Revision 1.5 2001/05/25 22:34:30 jongfoster
378 * Revision 1.4 2001/05/22 18:46:04 oes
380 * - Enabled filtering banners by size rather than URL
381 * by adding patterns that replace all standard banner
382 * sizes with the "Junkbuster" gif to the re_filterfile
384 * - Enabled filtering WebBugs by providing a pattern
385 * which kills all 1x1 images
387 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
388 * which is selected by the (nonstandard and therefore
389 * capital) letter 'U' in the option string.
390 * It causes the quantifiers to be ungreedy by default.
391 * Appending a ? turns back to greedy (!).
393 * - Added a new interceptor ijb-send-banner, which
394 * sends back the "Junkbuster" gif. Without imagelist or
395 * MSIE detection support, or if tinygif = 1, or the
396 * URL isn't recognized as an imageurl, a lame HTML
397 * explanation is sent instead.
399 * - Added new feature, which permits blocking remote
400 * script redirects and firing back a local redirect
402 * The feature is conditionally compiled, i.e. it
403 * can be disabled with --disable-fast-redirects,
404 * plus it must be activated by a "fast-redirects"
405 * line in the config file, has its own log level
406 * and of course wants to be displayed by show-proxy-args
407 * Note: Boy, all the #ifdefs in 1001 locations and
408 * all the fumbling with configure.in and acconfig.h
409 * were *way* more work than the feature itself :-(
411 * - Because a generic redirect template was needed for
412 * this, tinygif = 3 now uses the same.
414 * - Moved GIFs, and other static HTTP response templates
419 * - Removed some >400 CRs again (Jon, you really worked
422 * Revision 1.3 2001/05/20 16:44:47 jongfoster
423 * Removing last hardcoded Junkbusters.com URLs.
425 * Revision 1.2 2001/05/20 01:21:20 jongfoster
426 * Version 2.9.4 checkin.
427 * - Merged popupfile and cookiefile, and added control over PCRS
428 * filtering, in new "permissionsfile".
429 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
430 * file error you now get a message box (in the Win32 GUI) rather
431 * than the program exiting with no explanation.
432 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
434 * - Removed tabs from "config"
435 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
436 * - Bumped up version number.
438 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
439 * Initial import of version 2.9.3 source tree
442 *********************************************************************/
448 #include <sys/types.h>
457 #endif /* ndef __OS2__ */
458 #include <netinet/in.h>
460 #include <winsock2.h>
461 #endif /* ndef _WIN32 */
465 #endif /* def __OS2__ */
473 #include "jbsockets.h"
474 #include "miscutil.h"
478 #include "deanimate.h"
479 #include "urlmatch.h"
485 const char filters_h_rcs[] = FILTERS_H_VERSION;
487 /* Fix a problem with Solaris. There should be no effect on other
489 * Solaris's isspace() is a macro which uses it's argument directly
490 * as an array index. Therefore we need to make sure that high-bit
491 * characters generate +ve values, and ideally we also want to make
492 * the argument match the declared parameter type of "int".
494 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
498 /*********************************************************************
500 * Function : block_acl
502 * Description : Block this request?
503 * Decide yes or no based on ACL file.
506 * 1 : dst = The proxy or gateway address this is going to.
507 * Or NULL to check all possible targets.
508 * 2 : csp = Current client state (buffers, headers, etc...)
509 * Also includes the client IP address.
511 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
513 *********************************************************************/
514 int block_acl(struct access_control_addr *dst, struct client_state *csp)
516 struct access_control_list *acl = csp->config->acl;
518 /* if not using an access control list, then permit the connection */
524 /* search the list */
527 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
531 /* Just want to check if they have any access */
532 if (acl->action == ACL_PERMIT)
537 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
538 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
540 if (acl->action == ACL_PERMIT)
558 /*********************************************************************
560 * Function : acl_addr
562 * Description : Called from `load_config' to parse an ACL address.
565 * 1 : aspec = String specifying ACL address.
566 * 2 : aca = struct access_control_addr to fill in.
568 * Returns : 0 => Ok, everything else is an error.
570 *********************************************************************/
571 int acl_addr(char *aspec, struct access_control_addr *aca)
573 int i, masklength, port;
579 if ((p = strchr(aspec, '/')) != NULL)
583 if (ijb_isdigit(*p) == 0)
587 masklength = atoi(p);
590 if ((masklength < 0) || (masklength > 32))
595 if ((p = strchr(aspec, ':')) != NULL)
599 if (ijb_isdigit(*p) == 0)
608 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
610 if (aca->addr == INADDR_NONE)
615 /* build the netmask */
617 for (i=1; i <= masklength ; i++)
619 aca->mask |= (1 << (32 - i));
622 /* now mask off the host portion of the ip address
623 * (i.e. save on the network portion of the address).
625 aca->addr = aca->addr & aca->mask;
630 #endif /* def FEATURE_ACL */
633 /*********************************************************************
635 * Function : match_portlist
637 * Description : Check if a given number is covered by a comma
638 * separated list of numbers and ranges (a,b-c,d,..)
641 * 1 : portlist = String with list
642 * 2 : port = port to check
644 * Returns : 0 => no match
647 *********************************************************************/
648 int match_portlist(const char *portlist, int port)
650 char *min, *max, *next, *portlist_copy;
652 min = next = portlist_copy = strdup(portlist);
655 * Zero-terminate first item and remember offset for next
657 if (NULL != (next = strchr(portlist_copy, (int) ',')))
663 * Loop through all items, checking for match
667 if (NULL == (max = strchr(min, (int) '-')))
670 * No dash, check for equality
672 if (port == atoi(min))
681 * This is a range, so check if between min and max,
682 * or, if max was omitted, between min and 65K
685 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
699 * Zero-terminate next item and remember offset for n+1
701 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
713 /*********************************************************************
715 * Function : block_url
717 * Description : Called from `chat'. Check to see if we need to block this.
720 * 1 : csp = Current client state (buffers, headers, etc...)
722 * Returns : NULL => unblocked, else HTTP block response
724 *********************************************************************/
725 struct http_response *block_url(struct client_state *csp)
727 struct http_response *rsp;
728 const char *new_content_type = NULL;
731 * If it's not blocked, don't block it ;-)
733 if ((csp->action->flags & ACTION_BLOCK) == 0)
737 if (csp->action->flags & ACTION_REDIRECT)
739 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
742 * Else, prepare a response
744 if (NULL == (rsp = alloc_http_response()))
746 return cgi_error_memory();
750 * If it's an image-url, send back an image or redirect
751 * as specified by the relevant +image action
753 #ifdef FEATURE_IMAGE_BLOCKING
754 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
758 /* determine HOW images should be blocked */
759 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
761 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
763 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
765 #if 1 /* Two alternative strategies, use this one for now: */
767 /* and handle accordingly: */
768 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
770 rsp->status = strdup("403 Request blocked by Privoxy");
771 if (rsp->status == NULL)
773 free_http_response(rsp);
774 return cgi_error_memory();
776 rsp->body = bindup(image_pattern_data, image_pattern_length);
777 if (rsp->body == NULL)
779 free_http_response(rsp);
780 return cgi_error_memory();
782 rsp->content_length = image_pattern_length;
784 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
786 free_http_response(rsp);
787 return cgi_error_memory();
791 else if (0 == strcmpic(p, "blank"))
793 rsp->status = strdup("403 Request blocked by Privoxy");
794 if (rsp->status == NULL)
796 free_http_response(rsp);
797 return cgi_error_memory();
799 rsp->body = bindup(image_blank_data, image_blank_length);
800 if (rsp->body == NULL)
802 free_http_response(rsp);
803 return cgi_error_memory();
805 rsp->content_length = image_blank_length;
807 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
809 free_http_response(rsp);
810 return cgi_error_memory();
816 rsp->status = strdup("302 Local Redirect from Privoxy");
817 if (rsp->status == NULL)
819 free_http_response(rsp);
820 return cgi_error_memory();
823 if (enlist_unique_header(rsp->headers, "Location", p))
825 free_http_response(rsp);
826 return cgi_error_memory();
830 #else /* Following code is disabled for now */
832 /* and handle accordingly: */
833 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
835 p = CGI_PREFIX "send-banner?type=pattern";
837 else if (0 == strcmpic(p, "blank"))
839 p = CGI_PREFIX "send-banner?type=blank";
841 rsp->status = strdup("302 Local Redirect from Privoxy");
842 if (rsp->status == NULL)
844 free_http_response(rsp);
845 return cgi_error_memory();
848 if (enlist_unique_header(rsp->headers, "Location", p))
850 free_http_response(rsp);
851 return cgi_error_memory();
853 #endif /* Preceeding code is disabled for now */
855 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
858 * Send empty document.
860 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
863 rsp->body = strdup(" ");
864 rsp->content_length = 1;
866 rsp->status = strdup("403 Request blocked by Privoxy");
867 if (rsp->status == NULL)
869 free_http_response(rsp);
870 return cgi_error_memory();
872 if (new_content_type != 0)
874 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
875 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
877 free_http_response(rsp);
878 return cgi_error_memory();
884 #endif /* def FEATURE_IMAGE_BLOCKING */
887 * Else, generate an HTML "blocked" message:
891 struct map * exports;
895 * Workaround for stupid Netscape bug which prevents
896 * pages from being displayed if loading a referenced
897 * JavaScript or style sheet fails. So make it appear
898 * as if it succeeded.
900 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
901 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
902 && !strstr(p, "Gecko") /* save Mozilla, */
903 && !strstr(p, "compatible") /* MSIE */
904 && !strstr(p, "Opera")) /* and Opera. */
906 rsp->status = strdup("200 Request for blocked URL");
910 rsp->status = strdup("404 Request for blocked URL");
913 if (rsp->status == NULL)
915 free_http_response(rsp);
916 return cgi_error_memory();
919 exports = default_exports(csp, NULL);
922 free_http_response(rsp);
923 return cgi_error_memory();
926 #ifdef FEATURE_FORCE_LOAD
927 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
928 if (csp->http->ssl != 0)
929 #endif /* ndef FEATURE_FORCE_LOAD */
931 err = map_block_killer(exports, "force-support");
934 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
935 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
936 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
937 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
942 free_http_response(rsp);
943 return cgi_error_memory();
946 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
949 free_http_response(rsp);
950 return cgi_error_memory();
954 return finish_http_response(rsp);
960 /*********************************************************************
962 * Function : trust_url FIXME: I should be called distrust_url
964 * Description : Calls is_untrusted_url to determine if the URL is trusted
965 * and if not, returns a HTTP 304 response with a reject message.
968 * 1 : csp = Current client state (buffers, headers, etc...)
970 * Returns : NULL => trusted, else http_response.
972 *********************************************************************/
973 struct http_response *trust_url(struct client_state *csp)
975 struct http_response *rsp;
976 struct map * exports;
977 char buf[BUFFER_SIZE];
979 struct url_spec **tl;
984 * Don't bother to work on trusted URLs
986 if (!is_untrusted_url(csp))
992 * Else, prepare a response:
994 if (NULL == (rsp = alloc_http_response()))
996 return cgi_error_memory();
999 exports = default_exports(csp, NULL);
1000 if (exports == NULL)
1002 free_http_response(rsp);
1003 return cgi_error_memory();
1007 * Export the protocol, host, port, and referrer information
1009 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1010 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1011 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1013 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1015 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1019 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
1025 free_http_response(rsp);
1026 return cgi_error_memory();
1030 * Export the trust list
1033 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1035 sprintf(buf, "<li>%s</li>\n", t->spec);
1036 string_append(&p, buf);
1038 err = map(exports, "trusted-referrers", 1, p, 0);
1044 free_http_response(rsp);
1045 return cgi_error_memory();
1049 * Export the trust info, if available
1051 if (csp->config->trust_info->first)
1053 struct list_entry *l;
1056 for (l = csp->config->trust_info->first; l ; l = l->next)
1058 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
1059 string_append(&p, buf);
1061 err = map(exports, "trust-info", 1, p, 0);
1066 err = map_block_killer(exports, "have-trust-info");
1072 free_http_response(rsp);
1073 return cgi_error_memory();
1077 * Export the force prefix or the force conditional block killer
1079 #ifdef FEATURE_FORCE_LOAD
1080 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1081 #else /* ifndef FEATURE_FORCE_LOAD */
1082 err = map_block_killer(exports, "force-support");
1083 #endif /* ndef FEATURE_FORCE_LOAD */
1088 free_http_response(rsp);
1089 return cgi_error_memory();
1093 * Build the response
1095 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1098 free_http_response(rsp);
1099 return cgi_error_memory();
1102 return finish_http_response(rsp);
1104 #endif /* def FEATURE_TRUST */
1107 #ifdef FEATURE_FAST_REDIRECTS
1108 /*********************************************************************
1110 * Function : redirect_url
1112 * Description : Checks for redirection URLs and returns a HTTP redirect
1113 * to the destination URL, if necessary
1116 * 1 : csp = Current client state (buffers, headers, etc...)
1118 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1120 *********************************************************************/
1121 struct http_response *redirect_url(struct client_state *csp)
1124 struct http_response *rsp;
1125 char *redirect_mode = NULL;
1128 if ((csp->action->flags & ACTION_REDIRECT))
1130 q = csp->action->string[ACTION_STRING_REDIRECT];
1134 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1135 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1137 p = q = csp->http->path;
1138 log_error(LOG_LEVEL_REDIRECTS, "Decoding path: %s if necessary.", p);
1141 if (*p == '%') /* Escape sequence? */
1143 /* Yes, translate from hexadecimal to decimal */
1157 /* No, forward character. */
1163 p = q = csp->http->path;
1164 log_error(LOG_LEVEL_REDIRECTS, "Checking path for redirects: %s", p);
1167 * find the last URL encoded in the request
1169 while ((p = strstr(p, "http://")) != NULL)
1175 * if there was any, generate and return a HTTP redirect
1177 if (q != csp->http->path)
1179 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1181 if (NULL == (rsp = alloc_http_response()))
1183 return cgi_error_memory();
1186 if ( enlist_unique_header(rsp->headers, "Location", q)
1187 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1189 free_http_response(rsp);
1190 return cgi_error_memory();
1193 return finish_http_response(rsp);
1201 #endif /* def FEATURE_FAST_REDIRECTS */
1204 #ifdef FEATURE_IMAGE_BLOCKING
1205 /*********************************************************************
1207 * Function : is_imageurl
1209 * Description : Given a URL, decide whether it is an image or not,
1210 * using either the info from a previous +image action
1211 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1212 * is MSIE and not on a Mac, tell from the browser's accept
1216 * 1 : csp = Current client state (buffers, headers, etc...)
1218 * Returns : True (nonzero) if URL is an image, false (0)
1221 *********************************************************************/
1222 int is_imageurl(struct client_state *csp)
1224 #ifdef FEATURE_IMAGE_DETECT_MSIE
1227 tmp = get_header_value(csp->headers, "User-Agent:");
1228 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1230 tmp = get_header_value(csp->headers, "Accept:");
1231 if (tmp && strstr(tmp, "image/gif"))
1233 /* Client will accept HTML. If this seems counterintuitive,
1243 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1245 return ((csp->action->flags & ACTION_IMAGE) != 0);
1248 #endif /* def FEATURE_IMAGE_BLOCKING */
1251 #ifdef FEATURE_TRUST
1252 /*********************************************************************
1254 * Function : is_untrusted_url
1256 * Description : Should we "distrust" this URL (and block it)?
1258 * Yes if it matches a line in the trustfile, or if the
1259 * referrer matches a line starting with "+" in the
1264 * 1 : csp = Current client state (buffers, headers, etc...)
1266 * Returns : 0 => trusted, 1 => untrusted
1268 *********************************************************************/
1269 int is_untrusted_url(struct client_state *csp)
1271 struct file_list *fl;
1272 struct block_spec *b;
1273 struct url_spec **trusted_url;
1274 struct http_request rhttp[1];
1275 const char * referer;
1279 * If we don't have a trustlist, we trust everybody
1281 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1286 memset(rhttp, '\0', sizeof(*rhttp));
1289 * Do we trust the request URL itself?
1291 for (b = b->next; b ; b = b->next)
1293 if (url_match(b->url, csp->http))
1299 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1301 /* no referrer was supplied */
1307 * If not, do we maybe trust its referrer?
1309 err = parse_http_url(referer, rhttp, csp);
1315 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1317 if (url_match(*trusted_url, rhttp))
1319 /* if the URL's referrer is from a trusted referrer, then
1320 * add the target spec to the trustfile as an unblocked
1321 * domain and return NULL (which means it's OK).
1326 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1330 char * new_entry = strdup("~");
1332 string_append(&new_entry, csp->http->hostport);
1334 path = csp->http->path;
1335 if ( (path[0] == '/')
1337 && ((path_end = strchr(path + 2, '/')) != NULL))
1339 /* since this path points into a user's home space
1340 * be sure to include this spec in the trustfile.
1342 int path_len = path_end - path; /* save offset */
1343 path = strdup(path); /* Copy string */
1346 path_end = path + path_len; /* regenerate ptr to new buffer */
1347 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1349 string_join(&new_entry, path);
1352 if (new_entry != NULL)
1354 fprintf(fp, "%s\n", new_entry);
1359 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1360 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1370 #endif /* def FEATURE_TRUST */
1373 /*********************************************************************
1375 * Function : pcrs_filter_response
1377 * Description : Execute all text substitutions from all applying
1378 * +filter actions on the text buffer that's been accumulated
1379 * in csp->iob->buf. If this changes the contents, set
1380 * csp->content_length to the modified size and raise the
1381 * CSP_FLAG_MODIFIED flag.
1384 * 1 : csp = Current client state (buffers, headers, etc...)
1386 * Returns : a pointer to the (newly allocated) modified buffer.
1387 * or NULL if there were no hits or something went wrong
1389 *********************************************************************/
1390 char *pcrs_filter_response(struct client_state *csp)
1395 char *old = csp->iob->cur, *new = NULL;
1398 struct file_list *fl;
1399 struct re_filterfile_spec *b;
1400 struct list_entry *filtername;
1402 int i, found_filters = 0;
1407 if (csp->iob->cur >= csp->iob->eod)
1411 size = csp->iob->eod - csp->iob->cur;
1414 * Need to check the set of re_filterfiles...
1416 for (i = 0; i < MAX_AF_FILES; i++)
1429 if (0 == found_filters)
1431 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1436 * If the body has a "chunked" transfer-encoding,
1437 * get rid of it first, adjusting size and iob->eod
1439 if (csp->flags & CSP_FLAG_CHUNKED)
1441 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1442 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1446 csp->iob->eod = csp->iob->cur + size;
1447 csp->flags |= CSP_FLAG_MODIFIED;
1450 for (i = 0; i < MAX_AF_FILES; i++)
1453 if ((NULL == fl) || (NULL == fl->f))
1456 * For all applying +filter actions, look if a filter by that
1457 * name exists and if yes, execute it's pcrs_joblist on the
1460 for (b = fl->f; b; b = b->next)
1462 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1463 filtername ; filtername = filtername->next)
1465 if (strcmp(b->name, filtername->str) == 0)
1467 int current_hits = 0;
1469 if ( NULL == b->joblist )
1471 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1475 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1476 csp->http->hostport, csp->http->path, size, b->name);
1478 /* Apply all jobs from the joblist */
1479 for (job = b->joblist; NULL != job; job = job->next)
1481 current_hits += pcrs_execute(job, old, size, &new, &size);
1482 if (old != csp->iob->cur) free(old);
1486 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1487 hits += current_hits;
1494 * If there were no hits, destroy our copy and let
1495 * chat() use the original in csp->iob
1503 csp->flags |= CSP_FLAG_MODIFIED;
1504 csp->content_length = size;
1512 /*********************************************************************
1514 * Function : gif_deanimate_response
1516 * Description : Deanimate the GIF image that has been accumulated in
1517 * csp->iob->buf, set csp->content_length to the modified
1518 * size and raise the CSP_FLAG_MODIFIED flag.
1521 * 1 : csp = Current client state (buffers, headers, etc...)
1523 * Returns : a pointer to the (newly allocated) modified buffer.
1524 * or NULL in case something went wrong.
1526 *********************************************************************/
1527 char *gif_deanimate_response(struct client_state *csp)
1529 struct binbuffer *in, *out;
1531 size_t size = csp->iob->eod - csp->iob->cur;
1534 * If the body has a "chunked" transfer-encoding,
1535 * get rid of it first, adjusting size and iob->eod
1537 if (csp->flags & CSP_FLAG_CHUNKED)
1539 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1540 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1544 csp->iob->eod = csp->iob->cur + size;
1545 csp->flags |= CSP_FLAG_MODIFIED;
1548 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1549 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1551 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1555 in->buffer = csp->iob->cur;
1558 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1560 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1567 if ((int)size == out->offset)
1569 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1573 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1575 csp->content_length = out->offset;
1576 csp->flags |= CSP_FLAG_MODIFIED;
1586 /*********************************************************************
1588 * Function : jpeg_inspect_response
1593 * 1 : csp = Current client state (buffers, headers, etc...)
1595 * Returns : a pointer to the (newly allocated) modified buffer
1596 * or NULL in case something went wrong.
1598 *********************************************************************/
1599 char *jpeg_inspect_response(struct client_state *csp)
1601 struct binbuffer *in = NULL, *out = NULL;
1603 size_t size = csp->iob->eod - csp->iob->cur;
1606 * If the body has a "chunked" transfer-encoding,
1607 * get rid of it first, adjusting size and iob->eod
1609 if (csp->flags & CSP_FLAG_CHUNKED)
1611 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1612 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1616 csp->iob->eod = csp->iob->cur + size;
1617 csp->flags |= CSP_FLAG_MODIFIED;
1620 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1622 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
1626 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
1628 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
1632 in->buffer = csp->iob->cur;
1636 * Calling jpeg_inspect has the side-effect of creating and
1637 * modifying the image buffer of "out" directly.
1639 if (jpeg_inspect(in, out))
1641 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
1649 csp->content_length = out->offset;
1650 csp->flags |= CSP_FLAG_MODIFIED;
1660 /*********************************************************************
1662 * Function : remove_chunked_transfer_coding
1664 * Description : In-situ remove the "chunked" transfer coding as defined
1665 * in rfc2616 from a buffer.
1668 * 1 : buffer = Pointer to the text buffer
1669 * 2 : size = Number of bytes to be processed
1671 * Returns : The new size, i.e. the number of bytes from buffer which
1672 * are occupied by the stripped body, or 0 in case something
1675 *********************************************************************/
1676 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1679 unsigned int chunksize = 0;
1680 char *from_p, *to_p;
1683 from_p = to_p = buffer;
1685 if (sscanf(buffer, "%x", &chunksize) != 1)
1687 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1691 while (chunksize > 0)
1693 if (NULL == (from_p = strstr(from_p, "\r\n")))
1695 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1699 if ((newsize += chunksize) >= size)
1701 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
1706 memmove(to_p, from_p, (size_t) chunksize);
1707 to_p = buffer + newsize;
1708 from_p += chunksize + 2;
1710 if (sscanf(from_p, "%x", &chunksize) != 1)
1712 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1717 /* FIXME: Should this get its own loglevel? */
1718 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1724 /*********************************************************************
1726 * Function : url_actions
1728 * Description : Gets the actions for this URL.
1731 * 1 : http = http_request request for blocked URLs
1732 * 2 : csp = Current client state (buffers, headers, etc...)
1736 *********************************************************************/
1737 void url_actions(struct http_request *http,
1738 struct client_state *csp)
1740 struct file_list *fl;
1741 struct url_actions *b;
1744 init_current_action(csp->action);
1746 for (i = 0; i < MAX_AF_FILES; i++)
1748 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1753 apply_url_actions(csp->action, http, b);
1760 /*********************************************************************
1762 * Function : apply_url_actions
1764 * Description : Applies a list of URL actions.
1767 * 1 : action = Destination.
1768 * 2 : http = Current URL
1769 * 3 : b = list of URL actions to apply
1773 *********************************************************************/
1774 void apply_url_actions(struct current_action_spec *action,
1775 struct http_request *http,
1776 struct url_actions *b)
1780 /* Should never happen */
1784 for (b = b->next; NULL != b; b = b->next)
1786 if (url_match(b->url, http))
1788 merge_current_action(action, b->action);
1794 /*********************************************************************
1796 * Function : forward_url
1798 * Description : Should we forward this to another proxy?
1801 * 1 : http = http_request request for current URL
1802 * 2 : csp = Current client state (buffers, headers, etc...)
1804 * Returns : Pointer to forwarding information.
1806 *********************************************************************/
1807 const struct forward_spec * forward_url(struct http_request *http,
1808 struct client_state *csp)
1810 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1811 struct forward_spec *fwd = csp->config->forward;
1820 if (url_match(fwd->url, http))
1831 /*********************************************************************
1833 * Function : direct_response
1835 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
1836 * request and if so, return a HTTP 501 to the client.
1838 * FIXME: I have a stupid name and I should handle the
1839 * requests properly. Still, what we do here is rfc-
1840 * compliant, whereas ignoring or forwarding are not.
1843 * 1 : csp = Current client state (buffers, headers, etc...)
1845 * Returns : http_response if , NULL if nonmatch or handler fail
1847 *********************************************************************/
1848 struct http_response *direct_response(struct client_state *csp)
1850 struct http_response *rsp;
1851 struct list_entry *p;
1853 if ((0 == strcmpic(csp->http->gpc, "trace"))
1854 || (0 == strcmpic(csp->http->gpc, "options")))
1856 for (p = csp->headers->first; (p != NULL) ; p = p->next)
1858 if (!strncmp("Max-Forwards:", p->str, 13)
1859 && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
1861 /* FIXME: We could handle at least TRACE here,
1862 but that would require a verbatim copy of
1863 the request which we don't have anymore */
1865 log_error(LOG_LEVEL_HEADER, "Found Max-Forwards:0 in OPTIONS or TRACE request -- Returning 501");
1867 /* Get mem for response or fail*/
1868 if (NULL == (rsp = alloc_http_response()))
1870 return cgi_error_memory();
1873 if (NULL == (rsp->status = strdup("501 Not Implemented")))
1875 free_http_response(rsp);
1876 return cgi_error_memory();
1880 return(finish_http_response(rsp));