1 const char filters_rcs[] = "$Id: filters.c,v 1.58.2.3 2002/09/25 14:51:51 oes Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/Attic/filters.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
7 * Functions declared include:
8 * `acl_addr', `add_stats', `block_acl', `block_imageurl',
9 * `block_url', `url_actions', `domain_split',
10 * `filter_popups', `forward_url', 'redirect_url',
11 * `ij_untrusted_url', `intercept_url', `pcrs_filter_respose',
12 * 'ijb_send_banner', 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 1.58.2.3 2002/09/25 14:51:51 oes
42 * Added basic support for OPTIONS and TRACE HTTP methods:
43 * New function direct_response which handles OPTIONS and
44 * TRACE requests whose Max-Forwards header field is zero.
46 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
47 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
49 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
50 * - Bugfix: Executing a filters without jobs no longer results in
51 * turing off *all* filters.
52 * - Security fix: Malicious web servers can't cause a seg fault
53 * through bogus chunk sizes anymore
55 * Revision 1.58 2002/04/24 02:11:17 oes
56 * Jon's multiple AF patch: url_actions now evaluates rules
59 * Revision 1.57 2002/04/08 20:38:34 swa
62 * Revision 1.56 2002/04/05 15:51:24 oes
63 * - bugfix: error-pages now get correct request protocol
64 * - fix for invalid HTML in trust info
66 * Revision 1.55 2002/04/02 16:13:51 oes
67 * Fix: No "Go there anyway" for SSL
69 * Revision 1.54 2002/04/02 14:55:56 oes
70 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
72 * Revision 1.53 2002/03/26 22:29:54 swa
73 * we have a new homepage!
75 * Revision 1.52 2002/03/24 16:35:57 jongfoster
78 * Revision 1.51 2002/03/24 15:23:33 jongfoster
81 * Revision 1.50 2002/03/24 13:25:43 swa
82 * name change related issues
84 * Revision 1.49 2002/03/16 20:29:14 oes
87 * Revision 1.48 2002/03/13 20:25:34 oes
88 * Better logging for content filters
90 * Revision 1.47 2002/03/13 00:30:52 jongfoster
92 * Added option of always sending redirect for imageblock,
93 * currently disabled with #if 0.
95 * Revision 1.46 2002/03/12 01:42:49 oes
96 * Introduced modular filters
98 * Revision 1.45 2002/03/08 16:47:50 oes
99 * Added choice beween GIF and PNG built-in images
101 * Revision 1.44 2002/03/07 03:49:31 oes
102 * - Fixed compiler warnings etc
103 * - Changed built-in images from GIF to PNG
104 * (with regard to Unisys patent issue)
105 * - Added a 4x4 pattern PNG which is less intrusive
106 * than the logo but also clearly marks the deleted banners
108 * Revision 1.43 2002/01/22 23:51:59 jongfoster
109 * Replacing strsav() with the safer string_append().
111 * Adding missing html_encode() to error message generators. Where encoded
112 * and unencoded versions of a string were provided, removing the unencoded
115 * Revision 1.42 2002/01/17 21:00:32 jongfoster
116 * Moving all our URL and URL pattern parsing code to urlmatch.c.
118 * Using a single, simple url_match(pattern,url) function - rather than
119 * the 3-line match routine which was repeated all over the place.
121 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
123 * Using parse_http_url() to parse URLs without faking a HTTP
124 * request line for parse_http_request().
126 * Revision 1.41 2001/11/13 00:14:07 jongfoster
127 * Fixing stupid bug now I've figured out what || means.
128 * (It always returns 0 or 1, not one of it's paramaters.)
130 * Revision 1.40 2001/10/26 17:37:55 oes
131 * - Re-enabled Netscape 200/404 bug workaround in block_url():
132 * - Removed OS/2 special case
133 * - Made block_url() independant from sed() having been run
134 * - Made trust_url independant from sed() having been run
135 * - Made is_imageurl independant from sed() having been run.
136 * It now checks User-Agent: and Accept: by itself.
139 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
140 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
141 * threads to call select() simultaneously. So, it's time to do a real, live,
142 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
143 * (native). Both versions will work, but using __OS2__ offers multi-threading.
145 * Revision 1.38 2001/10/23 21:32:33 jongfoster
146 * Adding error-checking to selected functions
148 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
149 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
150 * filters.c. Added a FIXME in front of the offending code. I'll gladly
151 * put in a better/more robust fix for all parties if one is presented...
152 * It seems that just returning 200 instead of 404 would pretty much fix
153 * it for everyone, but I don't know all the history of the problem.
155 * Revision 1.36 2001/10/10 16:44:16 oes
156 * Added match_portlist function
158 * Revision 1.35 2001/10/07 15:41:23 oes
159 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
161 * New function remove_chunked_transfer_coding that strips chunked
162 * transfer coding to plain and is called by pcrs_filter_response
163 * and gif_deanimate_response if neccessary
165 * Improved handling of zero-change re_filter runs
167 * pcrs_filter_response and gif_deanimate_response now remove
168 * chunked transfer codeing before processing the body.
170 * Revision 1.34 2001/09/20 15:49:36 steudten
172 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
173 * See cgi.c fill_template().
175 * Revision 1.33 2001/09/16 17:05:14 jongfoster
176 * Removing unused #include showarg.h
178 * Revision 1.32 2001/09/16 13:21:27 jongfoster
179 * Changes to use new list functions.
181 * Revision 1.31 2001/09/16 11:38:02 jongfoster
182 * Splitting fill_template() into 2 functions:
183 * template_load() loads the file
184 * template_fill() performs the PCRS regexps.
185 * This is because the CGI edit interface has a "table row"
186 * template which is used many times in the page - this
187 * change means it's only loaded from disk once.
189 * Revision 1.30 2001/09/16 11:00:10 jongfoster
190 * New function alloc_http_response, for symmetry with free_http_response
192 * Revision 1.29 2001/09/13 23:32:40 jongfoster
193 * Moving image data to cgi.c rather than cgi.h
194 * Fixing a GPF under Win32 (and any other OS that protects global
195 * constants from being written to).
197 * Revision 1.28 2001/09/10 10:18:51 oes
198 * Silenced compiler warnings
200 * Revision 1.27 2001/08/05 16:06:20 jongfoster
201 * Modifiying "struct map" so that there are now separate header and
202 * "map_entry" structures. This means that functions which modify a
203 * map no longer need to return a pointer to the modified map.
204 * Also, it no longer reverses the order of the entries (which may be
205 * important with some advanced template substitutions).
207 * Revision 1.26 2001/07/30 22:08:36 jongfoster
208 * Tidying up #defines:
209 * - All feature #defines are now of the form FEATURE_xxx
210 * - Permanently turned off WIN_GUI_EDIT
211 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
213 * Revision 1.25 2001/07/26 10:09:46 oes
214 * Made browser detection a little less naive
216 * Revision 1.24 2001/07/25 17:22:51 oes
217 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
219 * Revision 1.23 2001/07/23 13:40:12 oes
220 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
222 * Revision 1.22 2001/07/18 12:29:34 oes
223 * - Made gif_deanimate_response respect
224 * csp->action->string[ACTION_STRING_DEANIMATE]
225 * - Logging cosmetics
227 * Revision 1.21 2001/07/13 13:59:53 oes
228 * - Introduced gif_deanimate_response which shares the
229 * generic content modification interface of pcrs_filter_response
230 * and acts as a wrapper to deanimate.c:gif_deanimate()
231 * - Renamed re_process_buffer to pcrs_filter_response
232 * - pcrs_filter_response now returns NULL on failiure
233 * - Removed all #ifdef PCRS
235 * Revision 1.20 2001/07/01 17:01:04 oes
236 * Added comments and missing return statement in is_untrusted_url()
238 * Revision 1.19 2001/06/29 21:45:41 oes
239 * Indentation, CRLF->LF, Tab-> Space
241 * Revision 1.18 2001/06/29 13:27:38 oes
242 * - Cleaned up, renamed and reorderd functions
243 * and improved comments
246 * - Ported to CGI platform. Now delivers
247 * http_response or NULL
248 * - Unified HTML and GIF generation (moved image detection
249 * and GIF generation here from jcc.c:chat())
250 * - Fixed HTTP status to:
251 * - 403 (Forbidden) for the "blocked" HTML message
252 * - 200 (OK) for GIF answers
253 * - 302 (Redirect) for redirect to GIF
256 * - Ported to CGI platform. Now delivers
257 * http_response or NULL
258 * - Separated detection of untrusted URL into
259 * (bool)is_untrusted_url
260 * - Added enforcement of untrusted requests
262 * - Moved redirect_url() from cgi.c to here
263 * and ported it to the CGI platform
265 * - Removed logentry from cancelled commit
267 * Revision 1.17 2001/06/09 10:55:28 jongfoster
268 * Changing BUFSIZ ==> BUFFER_SIZE
270 * Revision 1.16 2001/06/07 23:10:26 jongfoster
271 * Allowing unanchored domain patterns to back off and retry
272 * if they partially match. Optimized right-anchored patterns.
273 * Moving ACL and forward files into config file.
274 * Replacing struct gateway with struct forward_spec
276 * Revision 1.15 2001/06/03 19:12:00 oes
277 * extracted-CGI relevant stuff
279 * Revision 1.14 2001/06/01 10:30:55 oes
280 * Added optional left-anchoring to domaincmp
282 * Revision 1.13 2001/05/31 21:21:30 jongfoster
283 * Permissionsfile / actions file changes:
284 * - Changed "permission" to "action" throughout
285 * - changes to file format to allow string parameters
286 * - Moved helper functions to actions.c
288 * Revision 1.12 2001/05/31 17:35:20 oes
290 * - Enhanced domain part globbing with infix and prefix asterisk
291 * matching and optional unanchored operation
293 * Revision 1.11 2001/05/29 11:53:23 oes
294 * "See why" link added to "blocked" page
296 * Revision 1.10 2001/05/29 09:50:24 jongfoster
297 * Unified blocklist/imagelist/permissionslist.
298 * File format is still under discussion, but the internal changes
301 * Also modified interceptor behaviour:
302 * - We now intercept all URLs beginning with one of the following
303 * prefixes (and *only* these prefixes):
305 * * http://ijbswa.sf.net/config/
306 * * http://ijbswa.sourceforge.net/config/
307 * - New interceptors "home page" - go to http://i.j.b/ to see it.
308 * - Internal changes so that intercepted and fast redirect pages
309 * are not replaced with an image.
310 * - Interceptors now have the option to send a binary page direct
311 * to the client. (i.e. ijb-send-banner uses this)
312 * - Implemented show-url-info interceptor. (Which is why I needed
313 * the above interceptors changes - a typical URL is
314 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
315 * The previous mechanism would not have intercepted that, and
316 * if it had been intercepted then it then it would have replaced
319 * Revision 1.9 2001/05/27 22:17:04 oes
321 * - re_process_buffer no longer writes the modified buffer
322 * to the client, which was very ugly. It now returns the
323 * buffer, which it is then written by chat.
325 * - content_length now adjusts the Content-Length: header
326 * for modified documents rather than crunch()ing it.
327 * (Length info in csp->content_length, which is 0 for
328 * unmodified documents)
330 * - For this to work, sed() is called twice when filtering.
332 * Revision 1.8 2001/05/26 17:13:28 jongfoster
333 * Filled in a function comment.
335 * Revision 1.7 2001/05/26 15:26:15 jongfoster
336 * ACL feature now provides more security by immediately dropping
337 * connections from untrusted hosts.
339 * Revision 1.6 2001/05/26 00:28:36 jongfoster
340 * Automatic reloading of config file.
341 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
342 * Most of the global variables have been moved to a new
343 * struct configuration_spec, accessed through csp->config->globalname
344 * Most of the globals remaining are used by the Win32 GUI.
346 * Revision 1.5 2001/05/25 22:34:30 jongfoster
349 * Revision 1.4 2001/05/22 18:46:04 oes
351 * - Enabled filtering banners by size rather than URL
352 * by adding patterns that replace all standard banner
353 * sizes with the "Junkbuster" gif to the re_filterfile
355 * - Enabled filtering WebBugs by providing a pattern
356 * which kills all 1x1 images
358 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
359 * which is selected by the (nonstandard and therefore
360 * capital) letter 'U' in the option string.
361 * It causes the quantifiers to be ungreedy by default.
362 * Appending a ? turns back to greedy (!).
364 * - Added a new interceptor ijb-send-banner, which
365 * sends back the "Junkbuster" gif. Without imagelist or
366 * MSIE detection support, or if tinygif = 1, or the
367 * URL isn't recognized as an imageurl, a lame HTML
368 * explanation is sent instead.
370 * - Added new feature, which permits blocking remote
371 * script redirects and firing back a local redirect
373 * The feature is conditionally compiled, i.e. it
374 * can be disabled with --disable-fast-redirects,
375 * plus it must be activated by a "fast-redirects"
376 * line in the config file, has its own log level
377 * and of course wants to be displayed by show-proxy-args
378 * Note: Boy, all the #ifdefs in 1001 locations and
379 * all the fumbling with configure.in and acconfig.h
380 * were *way* more work than the feature itself :-(
382 * - Because a generic redirect template was needed for
383 * this, tinygif = 3 now uses the same.
385 * - Moved GIFs, and other static HTTP response templates
390 * - Removed some >400 CRs again (Jon, you really worked
393 * Revision 1.3 2001/05/20 16:44:47 jongfoster
394 * Removing last hardcoded Junkbusters.com URLs.
396 * Revision 1.2 2001/05/20 01:21:20 jongfoster
397 * Version 2.9.4 checkin.
398 * - Merged popupfile and cookiefile, and added control over PCRS
399 * filtering, in new "permissionsfile".
400 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
401 * file error you now get a message box (in the Win32 GUI) rather
402 * than the program exiting with no explanation.
403 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
405 * - Removed tabs from "config"
406 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
407 * - Bumped up version number.
409 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
410 * Initial import of version 2.9.3 source tree
413 *********************************************************************/
419 #include <sys/types.h>
428 #endif /* ndef __OS2__ */
429 #include <netinet/in.h>
431 #include <winsock2.h>
432 #endif /* ndef _WIN32 */
436 #endif /* def __OS2__ */
444 #include "jbsockets.h"
445 #include "miscutil.h"
449 #include "deanimate.h"
450 #include "urlmatch.h"
456 const char filters_h_rcs[] = FILTERS_H_VERSION;
458 /* Fix a problem with Solaris. There should be no effect on other
460 * Solaris's isspace() is a macro which uses it's argument directly
461 * as an array index. Therefore we need to make sure that high-bit
462 * characters generate +ve values, and ideally we also want to make
463 * the argument match the declared parameter type of "int".
465 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
469 /*********************************************************************
471 * Function : block_acl
473 * Description : Block this request?
474 * Decide yes or no based on ACL file.
477 * 1 : dst = The proxy or gateway address this is going to.
478 * Or NULL to check all possible targets.
479 * 2 : csp = Current client state (buffers, headers, etc...)
480 * Also includes the client IP address.
482 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
484 *********************************************************************/
485 int block_acl(struct access_control_addr *dst, struct client_state *csp)
487 struct access_control_list *acl = csp->config->acl;
489 /* if not using an access control list, then permit the connection */
495 /* search the list */
498 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
502 /* Just want to check if they have any access */
503 if (acl->action == ACL_PERMIT)
508 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
509 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
511 if (acl->action == ACL_PERMIT)
529 /*********************************************************************
531 * Function : acl_addr
533 * Description : Called from `load_config' to parse an ACL address.
536 * 1 : aspec = String specifying ACL address.
537 * 2 : aca = struct access_control_addr to fill in.
539 * Returns : 0 => Ok, everything else is an error.
541 *********************************************************************/
542 int acl_addr(char *aspec, struct access_control_addr *aca)
544 int i, masklength, port;
550 if ((p = strchr(aspec, '/')) != NULL)
554 if (ijb_isdigit(*p) == 0)
558 masklength = atoi(p);
561 if ((masklength < 0) || (masklength > 32))
566 if ((p = strchr(aspec, ':')) != NULL)
570 if (ijb_isdigit(*p) == 0)
579 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
581 if (aca->addr == INADDR_NONE)
586 /* build the netmask */
588 for (i=1; i <= masklength ; i++)
590 aca->mask |= (1 << (32 - i));
593 /* now mask off the host portion of the ip address
594 * (i.e. save on the network portion of the address).
596 aca->addr = aca->addr & aca->mask;
601 #endif /* def FEATURE_ACL */
604 /*********************************************************************
606 * Function : match_portlist
608 * Description : Check if a given number is covered by a comma
609 * separated list of numbers and ranges (a,b-c,d,..)
612 * 1 : portlist = String with list
613 * 2 : port = port to check
615 * Returns : 0 => no match
618 *********************************************************************/
619 int match_portlist(const char *portlist, int port)
621 char *min, *max, *next, *portlist_copy;
623 min = next = portlist_copy = strdup(portlist);
626 * Zero-terminate first item and remember offset for next
628 if (NULL != (next = strchr(portlist_copy, (int) ',')))
634 * Loop through all items, checking for match
638 if (NULL == (max = strchr(min, (int) '-')))
641 * No dash, check for equality
643 if (port == atoi(min))
652 * This is a range, so check if between min and max,
653 * or, if max was omitted, between min and 65K
656 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
670 * Zero-terminate next item and remember offset for n+1
672 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
684 /*********************************************************************
686 * Function : block_url
688 * Description : Called from `chat'. Check to see if we need to block this.
691 * 1 : csp = Current client state (buffers, headers, etc...)
693 * Returns : NULL => unblocked, else HTTP block response
695 *********************************************************************/
696 struct http_response *block_url(struct client_state *csp)
698 #ifdef FEATURE_IMAGE_BLOCKING
700 #endif /* def FEATURE_IMAGE_BLOCKING */
701 struct http_response *rsp;
704 * If it's not blocked, don't block it ;-)
706 if ((csp->action->flags & ACTION_BLOCK) == 0)
712 * Else, prepare a response
714 if (NULL == (rsp = alloc_http_response()))
716 return cgi_error_memory();
720 * If it's an image-url, send back an image or redirect
721 * as specified by the relevant +image action
723 #ifdef FEATURE_IMAGE_BLOCKING
724 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
727 /* determine HOW images should be blocked */
728 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
730 #if 1 /* Two alternative strategies, use this one for now: */
732 /* and handle accordingly: */
733 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
735 rsp->body = bindup(image_pattern_data, image_pattern_length);
736 if (rsp->body == NULL)
738 free_http_response(rsp);
739 return cgi_error_memory();
741 rsp->content_length = image_pattern_length;
743 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
745 free_http_response(rsp);
746 return cgi_error_memory();
750 else if (0 == strcmpic(p, "blank"))
752 rsp->body = bindup(image_blank_data, image_blank_length);
753 if (rsp->body == NULL)
755 free_http_response(rsp);
756 return cgi_error_memory();
758 rsp->content_length = image_blank_length;
760 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
762 free_http_response(rsp);
763 return cgi_error_memory();
769 rsp->status = strdup("302 Local Redirect from Privoxy");
770 if (rsp->status == NULL)
772 free_http_response(rsp);
773 return cgi_error_memory();
776 if (enlist_unique_header(rsp->headers, "Location", p))
778 free_http_response(rsp);
779 return cgi_error_memory();
783 #else /* Following code is disabled for now */
785 /* and handle accordingly: */
786 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
788 p = CGI_PREFIX "send-banner?type=pattern";
790 else if (0 == strcmpic(p, "blank"))
792 p = CGI_PREFIX "send-banner?type=blank";
794 rsp->status = strdup("302 Local Redirect from Privoxy");
795 if (rsp->status == NULL)
797 free_http_response(rsp);
798 return cgi_error_memory();
801 if (enlist_unique_header(rsp->headers, "Location", p))
803 free_http_response(rsp);
804 return cgi_error_memory();
806 #endif /* Preceeding code is disabled for now */
809 #endif /* def FEATURE_IMAGE_BLOCKING */
812 * Else, generate an HTML "blocked" message:
816 struct map * exports;
819 * Workaround for stupid Netscape bug which prevents
820 * pages from being displayed if loading a referenced
821 * JavaScript or style sheet fails. So make it appear
822 * as if it succeeded.
824 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
825 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
826 && !strstr(p, "Gecko") /* save Mozilla, */
827 && !strstr(p, "compatible") /* MSIE */
828 && !strstr(p, "Opera")) /* and Opera. */
830 rsp->status = strdup("200 Request for blocked URL");
834 rsp->status = strdup("404 Request for blocked URL");
837 if (rsp->status == NULL)
839 free_http_response(rsp);
840 return cgi_error_memory();
843 exports = default_exports(csp, NULL);
846 free_http_response(rsp);
847 return cgi_error_memory();
850 #ifdef FEATURE_FORCE_LOAD
851 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
852 if (csp->http->ssl != 0)
853 #endif /* ndef FEATURE_FORCE_LOAD */
855 err = map_block_killer(exports, "force-support");
858 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
859 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
860 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
865 free_http_response(rsp);
866 return cgi_error_memory();
869 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
872 free_http_response(rsp);
873 return cgi_error_memory();
877 return finish_http_response(rsp);
883 /*********************************************************************
885 * Function : trust_url FIXME: I should be called distrust_url
887 * Description : Calls is_untrusted_url to determine if the URL is trusted
888 * and if not, returns a HTTP 304 response with a reject message.
891 * 1 : csp = Current client state (buffers, headers, etc...)
893 * Returns : NULL => trusted, else http_response.
895 *********************************************************************/
896 struct http_response *trust_url(struct client_state *csp)
898 struct http_response *rsp;
899 struct map * exports;
900 char buf[BUFFER_SIZE];
902 struct url_spec **tl;
907 * Don't bother to work on trusted URLs
909 if (!is_untrusted_url(csp))
915 * Else, prepare a response:
917 if (NULL == (rsp = alloc_http_response()))
919 return cgi_error_memory();
922 exports = default_exports(csp, NULL);
925 free_http_response(rsp);
926 return cgi_error_memory();
930 * Export the protocol, host, port, and referrer information
932 err = map(exports, "hostport", 1, csp->http->hostport, 1);
933 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
934 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
936 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
938 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
942 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
948 free_http_response(rsp);
949 return cgi_error_memory();
953 * Export the trust list
956 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
958 sprintf(buf, "<li>%s</li>\n", t->spec);
959 string_append(&p, buf);
961 err = map(exports, "trusted-referrers", 1, p, 0);
966 free_http_response(rsp);
967 return cgi_error_memory();
971 * Export the trust info, if available
973 if (csp->config->trust_info->first)
975 struct list_entry *l;
978 for (l = csp->config->trust_info->first; l ; l = l->next)
980 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
981 string_append(&p, buf);
983 err = map(exports, "trust-info", 1, p, 0);
987 err = map_block_killer(exports, "have-trust-info");
993 free_http_response(rsp);
994 return cgi_error_memory();
998 * Export the force prefix or the force conditional block killer
1000 #ifdef FEATURE_FORCE_LOAD
1001 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1002 #else /* ifndef FEATURE_FORCE_LOAD */
1003 err = map_block_killer(exports, "force-support");
1004 #endif /* ndef FEATURE_FORCE_LOAD */
1009 free_http_response(rsp);
1010 return cgi_error_memory();
1014 * Build the response
1016 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1019 free_http_response(rsp);
1020 return cgi_error_memory();
1023 return finish_http_response(rsp);
1025 #endif /* def FEATURE_TRUST */
1028 #ifdef FEATURE_FAST_REDIRECTS
1029 /*********************************************************************
1031 * Function : redirect_url
1033 * Description : Checks for redirection URLs and returns a HTTP redirect
1034 * to the destination URL, if necessary
1037 * 1 : csp = Current client state (buffers, headers, etc...)
1039 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1041 *********************************************************************/
1042 struct http_response *redirect_url(struct client_state *csp)
1045 struct http_response *rsp;
1047 p = q = csp->http->path;
1048 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1051 * find the last URL encoded in the request
1053 while ((p = strstr(p, "http://")) != NULL)
1059 * if there was any, generate and return a HTTP redirect
1061 if (q != csp->http->path)
1063 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1065 if (NULL == (rsp = alloc_http_response()))
1067 return cgi_error_memory();
1070 if ( enlist_unique_header(rsp->headers, "Location", q)
1071 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1073 free_http_response(rsp);
1074 return cgi_error_memory();
1077 return finish_http_response(rsp);
1085 #endif /* def FEATURE_FAST_REDIRECTS */
1088 #ifdef FEATURE_IMAGE_BLOCKING
1089 /*********************************************************************
1091 * Function : is_imageurl
1093 * Description : Given a URL, decide whether it is an image or not,
1094 * using either the info from a previous +image action
1095 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1096 * is MSIE and not on a Mac, tell from the browser's accept
1100 * 1 : csp = Current client state (buffers, headers, etc...)
1102 * Returns : True (nonzero) if URL is an image, false (0)
1105 *********************************************************************/
1106 int is_imageurl(struct client_state *csp)
1108 #ifdef FEATURE_IMAGE_DETECT_MSIE
1111 tmp = get_header_value(csp->headers, "User-Agent:");
1112 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1114 tmp = get_header_value(csp->headers, "Accept:");
1115 if (tmp && strstr(tmp, "image/gif"))
1117 /* Client will accept HTML. If this seems counterintuitive,
1127 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1129 return ((csp->action->flags & ACTION_IMAGE) != 0);
1132 #endif /* def FEATURE_IMAGE_BLOCKING */
1135 #ifdef FEATURE_TRUST
1136 /*********************************************************************
1138 * Function : is_untrusted_url
1140 * Description : Should we "distrust" this URL (and block it)?
1142 * Yes if it matches a line in the trustfile, or if the
1143 * referrer matches a line starting with "+" in the
1148 * 1 : csp = Current client state (buffers, headers, etc...)
1150 * Returns : 0 => trusted, 1 => untrusted
1152 *********************************************************************/
1153 int is_untrusted_url(struct client_state *csp)
1155 struct file_list *fl;
1156 struct block_spec *b;
1157 struct url_spec **trusted_url;
1158 struct http_request rhttp[1];
1159 const char * referer;
1163 * If we don't have a trustlist, we trust everybody
1165 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1170 memset(rhttp, '\0', sizeof(*rhttp));
1173 * Do we trust the request URL itself?
1175 for (b = b->next; b ; b = b->next)
1177 if (url_match(b->url, csp->http))
1183 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1185 /* no referrer was supplied */
1191 * If not, do we maybe trust its referrer?
1193 err = parse_http_url(referer, rhttp, csp);
1199 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1201 if (url_match(*trusted_url, rhttp))
1203 /* if the URL's referrer is from a trusted referrer, then
1204 * add the target spec to the trustfile as an unblocked
1205 * domain and return NULL (which means it's OK).
1210 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1214 char * new_entry = strdup("~");
1216 string_append(&new_entry, csp->http->hostport);
1218 path = csp->http->path;
1219 if ( (path[0] == '/')
1221 && ((path_end = strchr(path + 2, '/')) != NULL))
1223 /* since this path points into a user's home space
1224 * be sure to include this spec in the trustfile.
1226 int path_len = path_end - path; /* save offset */
1227 path = strdup(path); /* Copy string */
1230 path_end = path + path_len; /* regenerate ptr to new buffer */
1231 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1233 string_join(&new_entry, path);
1236 if (new_entry != NULL)
1238 fprintf(fp, "%s\n", new_entry);
1243 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1244 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1254 #endif /* def FEATURE_TRUST */
1257 /*********************************************************************
1259 * Function : pcrs_filter_response
1261 * Description : Execute all text substitutions from all applying
1262 * +filter actions on the text buffer that's been accumulated
1263 * in csp->iob->buf. If this changes the contents, set
1264 * csp->content_length to the modified size and raise the
1265 * CSP_FLAG_MODIFIED flag.
1268 * 1 : csp = Current client state (buffers, headers, etc...)
1270 * Returns : a pointer to the (newly allocated) modified buffer.
1271 * or NULL if there were no hits or something went wrong
1273 *********************************************************************/
1274 char *pcrs_filter_response(struct client_state *csp)
1279 char *old = csp->iob->cur, *new = NULL;
1282 struct file_list *fl;
1283 struct re_filterfile_spec *b;
1284 struct list_entry *filtername;
1289 if (csp->iob->cur >= csp->iob->eod)
1293 size = csp->iob->eod - csp->iob->cur;
1295 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1297 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1302 * If the body has a "chunked" transfer-encoding,
1303 * get rid of it first, adjusting size and iob->eod
1305 if (csp->flags & CSP_FLAG_CHUNKED)
1307 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1308 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1312 csp->iob->eod = csp->iob->cur + size;
1313 csp->flags |= CSP_FLAG_MODIFIED;
1317 * For all applying +filter actions, look if a filter by that
1318 * name exists and if yes, execute it's pcrs_joblist on the
1321 for (b = fl->f; b; b = b->next)
1323 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1324 filtername ; filtername = filtername->next)
1326 if (strcmp(b->name, filtername->str) == 0)
1328 int current_hits = 0;
1330 if ( NULL == b->joblist )
1332 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1336 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1337 csp->http->hostport, csp->http->path, size, b->name);
1339 /* Apply all jobs from the joblist */
1340 for (job = b->joblist; NULL != job; job = job->next)
1342 current_hits += pcrs_execute(job, old, size, &new, &size);
1343 if (old != csp->iob->cur) free(old);
1347 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1348 hits += current_hits;
1354 * If there were no hits, destroy our copy and let
1355 * chat() use the original in csp->iob
1363 csp->flags |= CSP_FLAG_MODIFIED;
1364 csp->content_length = size;
1372 /*********************************************************************
1374 * Function : gif_deanimate_response
1376 * Description : Deanimate the GIF image that has been accumulated in
1377 * csp->iob->buf, set csp->content_length to the modified
1378 * size and raise the CSP_FLAG_MODIFIED flag.
1381 * 1 : csp = Current client state (buffers, headers, etc...)
1383 * Returns : a pointer to the (newly allocated) modified buffer.
1384 * or NULL in case something went wrong.
1386 *********************************************************************/
1387 char *gif_deanimate_response(struct client_state *csp)
1389 struct binbuffer *in, *out;
1391 size_t size = csp->iob->eod - csp->iob->cur;
1394 * If the body has a "chunked" transfer-encoding,
1395 * get rid of it first, adjusting size and iob->eod
1397 if (csp->flags & CSP_FLAG_CHUNKED)
1399 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1400 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1404 csp->iob->eod = csp->iob->cur + size;
1405 csp->flags |= CSP_FLAG_MODIFIED;
1408 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1409 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1411 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1415 in->buffer = csp->iob->cur;
1418 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1420 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1427 if ((int)size == out->offset)
1429 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1433 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1435 csp->content_length = out->offset;
1436 csp->flags |= CSP_FLAG_MODIFIED;
1446 /*********************************************************************
1448 * Function : remove_chunked_transfer_coding
1450 * Description : In-situ remove the "chunked" transfer coding as defined
1451 * in rfc2616 from a buffer.
1454 * 1 : buffer = Pointer to the text buffer
1455 * 2 : size = Number of bytes to be processed
1457 * Returns : The new size, i.e. the number of bytes from buffer which
1458 * are occupied by the stripped body, or 0 in case something
1461 *********************************************************************/
1462 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1465 unsigned int chunksize = 0;
1466 char *from_p, *to_p;
1469 from_p = to_p = buffer;
1471 if (sscanf(buffer, "%x", &chunksize) != 1)
1473 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1477 while (chunksize > 0)
1479 if (NULL == (from_p = strstr(from_p, "\r\n")))
1481 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1485 if ((newsize += chunksize) >= size)
1487 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
1492 memmove(to_p, from_p, (size_t) chunksize);
1493 to_p = buffer + newsize;
1494 from_p += chunksize + 2;
1496 if (sscanf(from_p, "%x", &chunksize) != 1)
1498 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1503 /* FIXME: Should this get its own loglevel? */
1504 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1510 /*********************************************************************
1512 * Function : url_actions
1514 * Description : Gets the actions for this URL.
1517 * 1 : http = http_request request for blocked URLs
1518 * 2 : csp = Current client state (buffers, headers, etc...)
1522 *********************************************************************/
1523 void url_actions(struct http_request *http,
1524 struct client_state *csp)
1526 struct file_list *fl;
1527 struct url_actions *b;
1530 init_current_action(csp->action);
1532 for (i = 0; i < MAX_ACTION_FILES; i++)
1534 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1539 apply_url_actions(csp->action, http, b);
1546 /*********************************************************************
1548 * Function : apply_url_actions
1550 * Description : Applies a list of URL actions.
1553 * 1 : action = Destination.
1554 * 2 : http = Current URL
1555 * 3 : b = list of URL actions to apply
1559 *********************************************************************/
1560 void apply_url_actions(struct current_action_spec *action,
1561 struct http_request *http,
1562 struct url_actions *b)
1566 /* Should never happen */
1570 for (b = b->next; NULL != b; b = b->next)
1572 if (url_match(b->url, http))
1574 merge_current_action(action, b->action);
1580 /*********************************************************************
1582 * Function : forward_url
1584 * Description : Should we forward this to another proxy?
1587 * 1 : http = http_request request for current URL
1588 * 2 : csp = Current client state (buffers, headers, etc...)
1590 * Returns : Pointer to forwarding information.
1592 *********************************************************************/
1593 const struct forward_spec * forward_url(struct http_request *http,
1594 struct client_state *csp)
1596 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1597 struct forward_spec *fwd = csp->config->forward;
1606 if (url_match(fwd->url, http))
1617 /*********************************************************************
1619 * Function : direct_response
1621 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
1622 * request and if so, return a HTTP 501 to the client.
1624 * FIXME: I have a stupid name and I should handle the
1625 * requests properly. Still, what we do here is rfc-
1626 * compliant, whereas ignoring or forwarding are not.
1629 * 1 : csp = Current client state (buffers, headers, etc...)
1631 * Returns : http_response if , NULL if nonmatch or handler fail
1633 *********************************************************************/
1634 struct http_response *direct_response(struct client_state *csp)
1636 struct http_response *rsp;
1637 struct list_entry *p;
1639 if ((0 == strcmpic(csp->http->gpc, "trace"))
1640 || (0 == strcmpic(csp->http->gpc, "options")))
1642 for (p = csp->headers->first; (p != NULL) ; p = p->next)
1644 if (!strncmp("Max-Forwards:", p->str, 13)
1645 && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
1647 /* FIXME: We could handle at least TRACE here,
1648 but that would require a verbatim copy of
1649 the request which we don't have anymore */
1651 log_error(LOG_LEVEL_HEADER, "Found Max-Forwards:0 in OPTIONS or TRACE request -- Returning 501");
1653 /* Get mem for response or fail*/
1654 if (NULL == (rsp = alloc_http_response()))
1656 return cgi_error_memory();
1659 if (NULL == (rsp->status = strdup("501 Not Implemented")))
1661 free_http_response(rsp);
1662 return cgi_error_memory();
1666 return(finish_http_response(rsp));