1 const char filters_rcs[] = "$Id: filters.c,v 1.58.2.5 2003/11/11 13:10:31 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 1.58.2.5 2003/11/11 13:10:31 oes
42 * Fixed bug #839859: "See why" link URL now gets url-encoded.
44 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
47 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
48 * Added basic support for OPTIONS and TRACE HTTP methods:
49 * New function direct_response which handles OPTIONS and
50 * TRACE requests whose Max-Forwards header field is zero.
52 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
53 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
55 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
56 * - Bugfix: Executing a filters without jobs no longer results in
57 * turing off *all* filters.
58 * - Security fix: Malicious web servers can't cause a seg fault
59 * through bogus chunk sizes anymore
61 * Revision 1.58 2002/04/24 02:11:17 oes
62 * Jon's multiple AF patch: url_actions now evaluates rules
65 * Revision 1.57 2002/04/08 20:38:34 swa
68 * Revision 1.56 2002/04/05 15:51:24 oes
69 * - bugfix: error-pages now get correct request protocol
70 * - fix for invalid HTML in trust info
72 * Revision 1.55 2002/04/02 16:13:51 oes
73 * Fix: No "Go there anyway" for SSL
75 * Revision 1.54 2002/04/02 14:55:56 oes
76 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
78 * Revision 1.53 2002/03/26 22:29:54 swa
79 * we have a new homepage!
81 * Revision 1.52 2002/03/24 16:35:57 jongfoster
84 * Revision 1.51 2002/03/24 15:23:33 jongfoster
87 * Revision 1.50 2002/03/24 13:25:43 swa
88 * name change related issues
90 * Revision 1.49 2002/03/16 20:29:14 oes
93 * Revision 1.48 2002/03/13 20:25:34 oes
94 * Better logging for content filters
96 * Revision 1.47 2002/03/13 00:30:52 jongfoster
98 * Added option of always sending redirect for imageblock,
99 * currently disabled with #if 0.
101 * Revision 1.46 2002/03/12 01:42:49 oes
102 * Introduced modular filters
104 * Revision 1.45 2002/03/08 16:47:50 oes
105 * Added choice beween GIF and PNG built-in images
107 * Revision 1.44 2002/03/07 03:49:31 oes
108 * - Fixed compiler warnings etc
109 * - Changed built-in images from GIF to PNG
110 * (with regard to Unisys patent issue)
111 * - Added a 4x4 pattern PNG which is less intrusive
112 * than the logo but also clearly marks the deleted banners
114 * Revision 1.43 2002/01/22 23:51:59 jongfoster
115 * Replacing strsav() with the safer string_append().
117 * Adding missing html_encode() to error message generators. Where encoded
118 * and unencoded versions of a string were provided, removing the unencoded
121 * Revision 1.42 2002/01/17 21:00:32 jongfoster
122 * Moving all our URL and URL pattern parsing code to urlmatch.c.
124 * Using a single, simple url_match(pattern,url) function - rather than
125 * the 3-line match routine which was repeated all over the place.
127 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
129 * Using parse_http_url() to parse URLs without faking a HTTP
130 * request line for parse_http_request().
132 * Revision 1.41 2001/11/13 00:14:07 jongfoster
133 * Fixing stupid bug now I've figured out what || means.
134 * (It always returns 0 or 1, not one of it's paramaters.)
136 * Revision 1.40 2001/10/26 17:37:55 oes
137 * - Re-enabled Netscape 200/404 bug workaround in block_url():
138 * - Removed OS/2 special case
139 * - Made block_url() independant from sed() having been run
140 * - Made trust_url independant from sed() having been run
141 * - Made is_imageurl independant from sed() having been run.
142 * It now checks User-Agent: and Accept: by itself.
145 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
146 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
147 * threads to call select() simultaneously. So, it's time to do a real, live,
148 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
149 * (native). Both versions will work, but using __OS2__ offers multi-threading.
151 * Revision 1.38 2001/10/23 21:32:33 jongfoster
152 * Adding error-checking to selected functions
154 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
155 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
156 * filters.c. Added a FIXME in front of the offending code. I'll gladly
157 * put in a better/more robust fix for all parties if one is presented...
158 * It seems that just returning 200 instead of 404 would pretty much fix
159 * it for everyone, but I don't know all the history of the problem.
161 * Revision 1.36 2001/10/10 16:44:16 oes
162 * Added match_portlist function
164 * Revision 1.35 2001/10/07 15:41:23 oes
165 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
167 * New function remove_chunked_transfer_coding that strips chunked
168 * transfer coding to plain and is called by pcrs_filter_response
169 * and gif_deanimate_response if neccessary
171 * Improved handling of zero-change re_filter runs
173 * pcrs_filter_response and gif_deanimate_response now remove
174 * chunked transfer codeing before processing the body.
176 * Revision 1.34 2001/09/20 15:49:36 steudten
178 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
179 * See cgi.c fill_template().
181 * Revision 1.33 2001/09/16 17:05:14 jongfoster
182 * Removing unused #include showarg.h
184 * Revision 1.32 2001/09/16 13:21:27 jongfoster
185 * Changes to use new list functions.
187 * Revision 1.31 2001/09/16 11:38:02 jongfoster
188 * Splitting fill_template() into 2 functions:
189 * template_load() loads the file
190 * template_fill() performs the PCRS regexps.
191 * This is because the CGI edit interface has a "table row"
192 * template which is used many times in the page - this
193 * change means it's only loaded from disk once.
195 * Revision 1.30 2001/09/16 11:00:10 jongfoster
196 * New function alloc_http_response, for symmetry with free_http_response
198 * Revision 1.29 2001/09/13 23:32:40 jongfoster
199 * Moving image data to cgi.c rather than cgi.h
200 * Fixing a GPF under Win32 (and any other OS that protects global
201 * constants from being written to).
203 * Revision 1.28 2001/09/10 10:18:51 oes
204 * Silenced compiler warnings
206 * Revision 1.27 2001/08/05 16:06:20 jongfoster
207 * Modifiying "struct map" so that there are now separate header and
208 * "map_entry" structures. This means that functions which modify a
209 * map no longer need to return a pointer to the modified map.
210 * Also, it no longer reverses the order of the entries (which may be
211 * important with some advanced template substitutions).
213 * Revision 1.26 2001/07/30 22:08:36 jongfoster
214 * Tidying up #defines:
215 * - All feature #defines are now of the form FEATURE_xxx
216 * - Permanently turned off WIN_GUI_EDIT
217 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
219 * Revision 1.25 2001/07/26 10:09:46 oes
220 * Made browser detection a little less naive
222 * Revision 1.24 2001/07/25 17:22:51 oes
223 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
225 * Revision 1.23 2001/07/23 13:40:12 oes
226 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
228 * Revision 1.22 2001/07/18 12:29:34 oes
229 * - Made gif_deanimate_response respect
230 * csp->action->string[ACTION_STRING_DEANIMATE]
231 * - Logging cosmetics
233 * Revision 1.21 2001/07/13 13:59:53 oes
234 * - Introduced gif_deanimate_response which shares the
235 * generic content modification interface of pcrs_filter_response
236 * and acts as a wrapper to deanimate.c:gif_deanimate()
237 * - Renamed re_process_buffer to pcrs_filter_response
238 * - pcrs_filter_response now returns NULL on failiure
239 * - Removed all #ifdef PCRS
241 * Revision 1.20 2001/07/01 17:01:04 oes
242 * Added comments and missing return statement in is_untrusted_url()
244 * Revision 1.19 2001/06/29 21:45:41 oes
245 * Indentation, CRLF->LF, Tab-> Space
247 * Revision 1.18 2001/06/29 13:27:38 oes
248 * - Cleaned up, renamed and reorderd functions
249 * and improved comments
252 * - Ported to CGI platform. Now delivers
253 * http_response or NULL
254 * - Unified HTML and GIF generation (moved image detection
255 * and GIF generation here from jcc.c:chat())
256 * - Fixed HTTP status to:
257 * - 403 (Forbidden) for the "blocked" HTML message
258 * - 200 (OK) for GIF answers
259 * - 302 (Redirect) for redirect to GIF
262 * - Ported to CGI platform. Now delivers
263 * http_response or NULL
264 * - Separated detection of untrusted URL into
265 * (bool)is_untrusted_url
266 * - Added enforcement of untrusted requests
268 * - Moved redirect_url() from cgi.c to here
269 * and ported it to the CGI platform
271 * - Removed logentry from cancelled commit
273 * Revision 1.17 2001/06/09 10:55:28 jongfoster
274 * Changing BUFSIZ ==> BUFFER_SIZE
276 * Revision 1.16 2001/06/07 23:10:26 jongfoster
277 * Allowing unanchored domain patterns to back off and retry
278 * if they partially match. Optimized right-anchored patterns.
279 * Moving ACL and forward files into config file.
280 * Replacing struct gateway with struct forward_spec
282 * Revision 1.15 2001/06/03 19:12:00 oes
283 * extracted-CGI relevant stuff
285 * Revision 1.14 2001/06/01 10:30:55 oes
286 * Added optional left-anchoring to domaincmp
288 * Revision 1.13 2001/05/31 21:21:30 jongfoster
289 * Permissionsfile / actions file changes:
290 * - Changed "permission" to "action" throughout
291 * - changes to file format to allow string parameters
292 * - Moved helper functions to actions.c
294 * Revision 1.12 2001/05/31 17:35:20 oes
296 * - Enhanced domain part globbing with infix and prefix asterisk
297 * matching and optional unanchored operation
299 * Revision 1.11 2001/05/29 11:53:23 oes
300 * "See why" link added to "blocked" page
302 * Revision 1.10 2001/05/29 09:50:24 jongfoster
303 * Unified blocklist/imagelist/permissionslist.
304 * File format is still under discussion, but the internal changes
307 * Also modified interceptor behaviour:
308 * - We now intercept all URLs beginning with one of the following
309 * prefixes (and *only* these prefixes):
311 * * http://ijbswa.sf.net/config/
312 * * http://ijbswa.sourceforge.net/config/
313 * - New interceptors "home page" - go to http://i.j.b/ to see it.
314 * - Internal changes so that intercepted and fast redirect pages
315 * are not replaced with an image.
316 * - Interceptors now have the option to send a binary page direct
317 * to the client. (i.e. ijb-send-banner uses this)
318 * - Implemented show-url-info interceptor. (Which is why I needed
319 * the above interceptors changes - a typical URL is
320 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
321 * The previous mechanism would not have intercepted that, and
322 * if it had been intercepted then it then it would have replaced
325 * Revision 1.9 2001/05/27 22:17:04 oes
327 * - re_process_buffer no longer writes the modified buffer
328 * to the client, which was very ugly. It now returns the
329 * buffer, which it is then written by chat.
331 * - content_length now adjusts the Content-Length: header
332 * for modified documents rather than crunch()ing it.
333 * (Length info in csp->content_length, which is 0 for
334 * unmodified documents)
336 * - For this to work, sed() is called twice when filtering.
338 * Revision 1.8 2001/05/26 17:13:28 jongfoster
339 * Filled in a function comment.
341 * Revision 1.7 2001/05/26 15:26:15 jongfoster
342 * ACL feature now provides more security by immediately dropping
343 * connections from untrusted hosts.
345 * Revision 1.6 2001/05/26 00:28:36 jongfoster
346 * Automatic reloading of config file.
347 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
348 * Most of the global variables have been moved to a new
349 * struct configuration_spec, accessed through csp->config->globalname
350 * Most of the globals remaining are used by the Win32 GUI.
352 * Revision 1.5 2001/05/25 22:34:30 jongfoster
355 * Revision 1.4 2001/05/22 18:46:04 oes
357 * - Enabled filtering banners by size rather than URL
358 * by adding patterns that replace all standard banner
359 * sizes with the "Junkbuster" gif to the re_filterfile
361 * - Enabled filtering WebBugs by providing a pattern
362 * which kills all 1x1 images
364 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
365 * which is selected by the (nonstandard and therefore
366 * capital) letter 'U' in the option string.
367 * It causes the quantifiers to be ungreedy by default.
368 * Appending a ? turns back to greedy (!).
370 * - Added a new interceptor ijb-send-banner, which
371 * sends back the "Junkbuster" gif. Without imagelist or
372 * MSIE detection support, or if tinygif = 1, or the
373 * URL isn't recognized as an imageurl, a lame HTML
374 * explanation is sent instead.
376 * - Added new feature, which permits blocking remote
377 * script redirects and firing back a local redirect
379 * The feature is conditionally compiled, i.e. it
380 * can be disabled with --disable-fast-redirects,
381 * plus it must be activated by a "fast-redirects"
382 * line in the config file, has its own log level
383 * and of course wants to be displayed by show-proxy-args
384 * Note: Boy, all the #ifdefs in 1001 locations and
385 * all the fumbling with configure.in and acconfig.h
386 * were *way* more work than the feature itself :-(
388 * - Because a generic redirect template was needed for
389 * this, tinygif = 3 now uses the same.
391 * - Moved GIFs, and other static HTTP response templates
396 * - Removed some >400 CRs again (Jon, you really worked
399 * Revision 1.3 2001/05/20 16:44:47 jongfoster
400 * Removing last hardcoded Junkbusters.com URLs.
402 * Revision 1.2 2001/05/20 01:21:20 jongfoster
403 * Version 2.9.4 checkin.
404 * - Merged popupfile and cookiefile, and added control over PCRS
405 * filtering, in new "permissionsfile".
406 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
407 * file error you now get a message box (in the Win32 GUI) rather
408 * than the program exiting with no explanation.
409 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
411 * - Removed tabs from "config"
412 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
413 * - Bumped up version number.
415 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
416 * Initial import of version 2.9.3 source tree
419 *********************************************************************/
425 #include <sys/types.h>
434 #endif /* ndef __OS2__ */
435 #include <netinet/in.h>
437 #include <winsock2.h>
438 #endif /* ndef _WIN32 */
442 #endif /* def __OS2__ */
450 #include "jbsockets.h"
451 #include "miscutil.h"
455 #include "deanimate.h"
456 #include "urlmatch.h"
462 const char filters_h_rcs[] = FILTERS_H_VERSION;
464 /* Fix a problem with Solaris. There should be no effect on other
466 * Solaris's isspace() is a macro which uses it's argument directly
467 * as an array index. Therefore we need to make sure that high-bit
468 * characters generate +ve values, and ideally we also want to make
469 * the argument match the declared parameter type of "int".
471 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
475 /*********************************************************************
477 * Function : block_acl
479 * Description : Block this request?
480 * Decide yes or no based on ACL file.
483 * 1 : dst = The proxy or gateway address this is going to.
484 * Or NULL to check all possible targets.
485 * 2 : csp = Current client state (buffers, headers, etc...)
486 * Also includes the client IP address.
488 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
490 *********************************************************************/
491 int block_acl(struct access_control_addr *dst, struct client_state *csp)
493 struct access_control_list *acl = csp->config->acl;
495 /* if not using an access control list, then permit the connection */
501 /* search the list */
504 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
508 /* Just want to check if they have any access */
509 if (acl->action == ACL_PERMIT)
514 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
515 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
517 if (acl->action == ACL_PERMIT)
535 /*********************************************************************
537 * Function : acl_addr
539 * Description : Called from `load_config' to parse an ACL address.
542 * 1 : aspec = String specifying ACL address.
543 * 2 : aca = struct access_control_addr to fill in.
545 * Returns : 0 => Ok, everything else is an error.
547 *********************************************************************/
548 int acl_addr(char *aspec, struct access_control_addr *aca)
550 int i, masklength, port;
556 if ((p = strchr(aspec, '/')) != NULL)
560 if (ijb_isdigit(*p) == 0)
564 masklength = atoi(p);
567 if ((masklength < 0) || (masklength > 32))
572 if ((p = strchr(aspec, ':')) != NULL)
576 if (ijb_isdigit(*p) == 0)
585 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
587 if (aca->addr == INADDR_NONE)
592 /* build the netmask */
594 for (i=1; i <= masklength ; i++)
596 aca->mask |= (1 << (32 - i));
599 /* now mask off the host portion of the ip address
600 * (i.e. save on the network portion of the address).
602 aca->addr = aca->addr & aca->mask;
607 #endif /* def FEATURE_ACL */
610 /*********************************************************************
612 * Function : match_portlist
614 * Description : Check if a given number is covered by a comma
615 * separated list of numbers and ranges (a,b-c,d,..)
618 * 1 : portlist = String with list
619 * 2 : port = port to check
621 * Returns : 0 => no match
624 *********************************************************************/
625 int match_portlist(const char *portlist, int port)
627 char *min, *max, *next, *portlist_copy;
629 min = next = portlist_copy = strdup(portlist);
632 * Zero-terminate first item and remember offset for next
634 if (NULL != (next = strchr(portlist_copy, (int) ',')))
640 * Loop through all items, checking for match
644 if (NULL == (max = strchr(min, (int) '-')))
647 * No dash, check for equality
649 if (port == atoi(min))
658 * This is a range, so check if between min and max,
659 * or, if max was omitted, between min and 65K
662 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
676 * Zero-terminate next item and remember offset for n+1
678 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
690 /*********************************************************************
692 * Function : block_url
694 * Description : Called from `chat'. Check to see if we need to block this.
697 * 1 : csp = Current client state (buffers, headers, etc...)
699 * Returns : NULL => unblocked, else HTTP block response
701 *********************************************************************/
702 struct http_response *block_url(struct client_state *csp)
704 struct http_response *rsp;
707 * If it's not blocked, don't block it ;-)
709 if ((csp->action->flags & ACTION_BLOCK) == 0)
715 * Else, prepare a response
717 if (NULL == (rsp = alloc_http_response()))
719 return cgi_error_memory();
723 * If it's an image-url, send back an image or redirect
724 * as specified by the relevant +image action
726 #ifdef FEATURE_IMAGE_BLOCKING
727 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
731 /* determine HOW images should be blocked */
732 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
734 #if 1 /* Two alternative strategies, use this one for now: */
736 /* and handle accordingly: */
737 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
739 rsp->body = bindup(image_pattern_data, image_pattern_length);
740 if (rsp->body == NULL)
742 free_http_response(rsp);
743 return cgi_error_memory();
745 rsp->content_length = image_pattern_length;
747 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
749 free_http_response(rsp);
750 return cgi_error_memory();
754 else if (0 == strcmpic(p, "blank"))
756 rsp->body = bindup(image_blank_data, image_blank_length);
757 if (rsp->body == NULL)
759 free_http_response(rsp);
760 return cgi_error_memory();
762 rsp->content_length = image_blank_length;
764 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
766 free_http_response(rsp);
767 return cgi_error_memory();
773 rsp->status = strdup("302 Local Redirect from Privoxy");
774 if (rsp->status == NULL)
776 free_http_response(rsp);
777 return cgi_error_memory();
780 if (enlist_unique_header(rsp->headers, "Location", p))
782 free_http_response(rsp);
783 return cgi_error_memory();
787 #else /* Following code is disabled for now */
789 /* and handle accordingly: */
790 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
792 p = CGI_PREFIX "send-banner?type=pattern";
794 else if (0 == strcmpic(p, "blank"))
796 p = CGI_PREFIX "send-banner?type=blank";
798 rsp->status = strdup("302 Local Redirect from Privoxy");
799 if (rsp->status == NULL)
801 free_http_response(rsp);
802 return cgi_error_memory();
805 if (enlist_unique_header(rsp->headers, "Location", p))
807 free_http_response(rsp);
808 return cgi_error_memory();
810 #endif /* Preceeding code is disabled for now */
813 #endif /* def FEATURE_IMAGE_BLOCKING */
816 * Else, generate an HTML "blocked" message:
820 struct map * exports;
824 * Workaround for stupid Netscape bug which prevents
825 * pages from being displayed if loading a referenced
826 * JavaScript or style sheet fails. So make it appear
827 * as if it succeeded.
829 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
830 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
831 && !strstr(p, "Gecko") /* save Mozilla, */
832 && !strstr(p, "compatible") /* MSIE */
833 && !strstr(p, "Opera")) /* and Opera. */
835 rsp->status = strdup("200 Request for blocked URL");
839 rsp->status = strdup("404 Request for blocked URL");
842 if (rsp->status == NULL)
844 free_http_response(rsp);
845 return cgi_error_memory();
848 exports = default_exports(csp, NULL);
851 free_http_response(rsp);
852 return cgi_error_memory();
855 #ifdef FEATURE_FORCE_LOAD
856 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
857 if (csp->http->ssl != 0)
858 #endif /* ndef FEATURE_FORCE_LOAD */
860 err = map_block_killer(exports, "force-support");
863 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
864 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
865 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
866 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
871 free_http_response(rsp);
872 return cgi_error_memory();
875 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
878 free_http_response(rsp);
879 return cgi_error_memory();
883 return finish_http_response(rsp);
889 /*********************************************************************
891 * Function : trust_url FIXME: I should be called distrust_url
893 * Description : Calls is_untrusted_url to determine if the URL is trusted
894 * and if not, returns a HTTP 304 response with a reject message.
897 * 1 : csp = Current client state (buffers, headers, etc...)
899 * Returns : NULL => trusted, else http_response.
901 *********************************************************************/
902 struct http_response *trust_url(struct client_state *csp)
904 struct http_response *rsp;
905 struct map * exports;
906 char buf[BUFFER_SIZE];
908 struct url_spec **tl;
913 * Don't bother to work on trusted URLs
915 if (!is_untrusted_url(csp))
921 * Else, prepare a response:
923 if (NULL == (rsp = alloc_http_response()))
925 return cgi_error_memory();
928 exports = default_exports(csp, NULL);
931 free_http_response(rsp);
932 return cgi_error_memory();
936 * Export the protocol, host, port, and referrer information
938 err = map(exports, "hostport", 1, csp->http->hostport, 1);
939 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
940 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
942 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
944 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
948 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
954 free_http_response(rsp);
955 return cgi_error_memory();
959 * Export the trust list
962 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
964 sprintf(buf, "<li>%s</li>\n", t->spec);
965 string_append(&p, buf);
967 err = map(exports, "trusted-referrers", 1, p, 0);
972 free_http_response(rsp);
973 return cgi_error_memory();
977 * Export the trust info, if available
979 if (csp->config->trust_info->first)
981 struct list_entry *l;
984 for (l = csp->config->trust_info->first; l ; l = l->next)
986 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
987 string_append(&p, buf);
989 err = map(exports, "trust-info", 1, p, 0);
993 err = map_block_killer(exports, "have-trust-info");
999 free_http_response(rsp);
1000 return cgi_error_memory();
1004 * Export the force prefix or the force conditional block killer
1006 #ifdef FEATURE_FORCE_LOAD
1007 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1008 #else /* ifndef FEATURE_FORCE_LOAD */
1009 err = map_block_killer(exports, "force-support");
1010 #endif /* ndef FEATURE_FORCE_LOAD */
1015 free_http_response(rsp);
1016 return cgi_error_memory();
1020 * Build the response
1022 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1025 free_http_response(rsp);
1026 return cgi_error_memory();
1029 return finish_http_response(rsp);
1031 #endif /* def FEATURE_TRUST */
1034 #ifdef FEATURE_FAST_REDIRECTS
1035 /*********************************************************************
1037 * Function : redirect_url
1039 * Description : Checks for redirection URLs and returns a HTTP redirect
1040 * to the destination URL, if necessary
1043 * 1 : csp = Current client state (buffers, headers, etc...)
1045 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1047 *********************************************************************/
1048 struct http_response *redirect_url(struct client_state *csp)
1051 struct http_response *rsp;
1053 p = q = csp->http->path;
1054 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1057 * find the last URL encoded in the request
1059 while ((p = strstr(p, "http://")) != NULL)
1065 * if there was any, generate and return a HTTP redirect
1067 if (q != csp->http->path)
1069 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1071 if (NULL == (rsp = alloc_http_response()))
1073 return cgi_error_memory();
1076 if ( enlist_unique_header(rsp->headers, "Location", q)
1077 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1079 free_http_response(rsp);
1080 return cgi_error_memory();
1083 return finish_http_response(rsp);
1091 #endif /* def FEATURE_FAST_REDIRECTS */
1094 #ifdef FEATURE_IMAGE_BLOCKING
1095 /*********************************************************************
1097 * Function : is_imageurl
1099 * Description : Given a URL, decide whether it is an image or not,
1100 * using either the info from a previous +image action
1101 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1102 * is MSIE and not on a Mac, tell from the browser's accept
1106 * 1 : csp = Current client state (buffers, headers, etc...)
1108 * Returns : True (nonzero) if URL is an image, false (0)
1111 *********************************************************************/
1112 int is_imageurl(struct client_state *csp)
1114 #ifdef FEATURE_IMAGE_DETECT_MSIE
1117 tmp = get_header_value(csp->headers, "User-Agent:");
1118 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1120 tmp = get_header_value(csp->headers, "Accept:");
1121 if (tmp && strstr(tmp, "image/gif"))
1123 /* Client will accept HTML. If this seems counterintuitive,
1133 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1135 return ((csp->action->flags & ACTION_IMAGE) != 0);
1138 #endif /* def FEATURE_IMAGE_BLOCKING */
1141 #ifdef FEATURE_TRUST
1142 /*********************************************************************
1144 * Function : is_untrusted_url
1146 * Description : Should we "distrust" this URL (and block it)?
1148 * Yes if it matches a line in the trustfile, or if the
1149 * referrer matches a line starting with "+" in the
1154 * 1 : csp = Current client state (buffers, headers, etc...)
1156 * Returns : 0 => trusted, 1 => untrusted
1158 *********************************************************************/
1159 int is_untrusted_url(struct client_state *csp)
1161 struct file_list *fl;
1162 struct block_spec *b;
1163 struct url_spec **trusted_url;
1164 struct http_request rhttp[1];
1165 const char * referer;
1169 * If we don't have a trustlist, we trust everybody
1171 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1176 memset(rhttp, '\0', sizeof(*rhttp));
1179 * Do we trust the request URL itself?
1181 for (b = b->next; b ; b = b->next)
1183 if (url_match(b->url, csp->http))
1189 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1191 /* no referrer was supplied */
1197 * If not, do we maybe trust its referrer?
1199 err = parse_http_url(referer, rhttp, csp);
1205 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1207 if (url_match(*trusted_url, rhttp))
1209 /* if the URL's referrer is from a trusted referrer, then
1210 * add the target spec to the trustfile as an unblocked
1211 * domain and return NULL (which means it's OK).
1216 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1220 char * new_entry = strdup("~");
1222 string_append(&new_entry, csp->http->hostport);
1224 path = csp->http->path;
1225 if ( (path[0] == '/')
1227 && ((path_end = strchr(path + 2, '/')) != NULL))
1229 /* since this path points into a user's home space
1230 * be sure to include this spec in the trustfile.
1232 int path_len = path_end - path; /* save offset */
1233 path = strdup(path); /* Copy string */
1236 path_end = path + path_len; /* regenerate ptr to new buffer */
1237 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1239 string_join(&new_entry, path);
1242 if (new_entry != NULL)
1244 fprintf(fp, "%s\n", new_entry);
1249 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1250 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1260 #endif /* def FEATURE_TRUST */
1263 /*********************************************************************
1265 * Function : pcrs_filter_response
1267 * Description : Execute all text substitutions from all applying
1268 * +filter actions on the text buffer that's been accumulated
1269 * in csp->iob->buf. If this changes the contents, set
1270 * csp->content_length to the modified size and raise the
1271 * CSP_FLAG_MODIFIED flag.
1274 * 1 : csp = Current client state (buffers, headers, etc...)
1276 * Returns : a pointer to the (newly allocated) modified buffer.
1277 * or NULL if there were no hits or something went wrong
1279 *********************************************************************/
1280 char *pcrs_filter_response(struct client_state *csp)
1285 char *old = csp->iob->cur, *new = NULL;
1288 struct file_list *fl;
1289 struct re_filterfile_spec *b;
1290 struct list_entry *filtername;
1295 if (csp->iob->cur >= csp->iob->eod)
1299 size = csp->iob->eod - csp->iob->cur;
1301 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1303 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1308 * If the body has a "chunked" transfer-encoding,
1309 * get rid of it first, adjusting size and iob->eod
1311 if (csp->flags & CSP_FLAG_CHUNKED)
1313 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1314 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1318 csp->iob->eod = csp->iob->cur + size;
1319 csp->flags |= CSP_FLAG_MODIFIED;
1323 * For all applying +filter actions, look if a filter by that
1324 * name exists and if yes, execute it's pcrs_joblist on the
1327 for (b = fl->f; b; b = b->next)
1329 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1330 filtername ; filtername = filtername->next)
1332 if (strcmp(b->name, filtername->str) == 0)
1334 int current_hits = 0;
1336 if ( NULL == b->joblist )
1338 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1342 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1343 csp->http->hostport, csp->http->path, size, b->name);
1345 /* Apply all jobs from the joblist */
1346 for (job = b->joblist; NULL != job; job = job->next)
1348 current_hits += pcrs_execute(job, old, size, &new, &size);
1349 if (old != csp->iob->cur) free(old);
1353 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1354 hits += current_hits;
1360 * If there were no hits, destroy our copy and let
1361 * chat() use the original in csp->iob
1369 csp->flags |= CSP_FLAG_MODIFIED;
1370 csp->content_length = size;
1378 /*********************************************************************
1380 * Function : gif_deanimate_response
1382 * Description : Deanimate the GIF image that has been accumulated in
1383 * csp->iob->buf, set csp->content_length to the modified
1384 * size and raise the CSP_FLAG_MODIFIED flag.
1387 * 1 : csp = Current client state (buffers, headers, etc...)
1389 * Returns : a pointer to the (newly allocated) modified buffer.
1390 * or NULL in case something went wrong.
1392 *********************************************************************/
1393 char *gif_deanimate_response(struct client_state *csp)
1395 struct binbuffer *in, *out;
1397 size_t size = csp->iob->eod - csp->iob->cur;
1400 * If the body has a "chunked" transfer-encoding,
1401 * get rid of it first, adjusting size and iob->eod
1403 if (csp->flags & CSP_FLAG_CHUNKED)
1405 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1406 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1410 csp->iob->eod = csp->iob->cur + size;
1411 csp->flags |= CSP_FLAG_MODIFIED;
1414 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1415 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1417 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1421 in->buffer = csp->iob->cur;
1424 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1426 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1433 if ((int)size == out->offset)
1435 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1439 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1441 csp->content_length = out->offset;
1442 csp->flags |= CSP_FLAG_MODIFIED;
1452 /*********************************************************************
1454 * Function : remove_chunked_transfer_coding
1456 * Description : In-situ remove the "chunked" transfer coding as defined
1457 * in rfc2616 from a buffer.
1460 * 1 : buffer = Pointer to the text buffer
1461 * 2 : size = Number of bytes to be processed
1463 * Returns : The new size, i.e. the number of bytes from buffer which
1464 * are occupied by the stripped body, or 0 in case something
1467 *********************************************************************/
1468 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1471 unsigned int chunksize = 0;
1472 char *from_p, *to_p;
1475 from_p = to_p = buffer;
1477 if (sscanf(buffer, "%x", &chunksize) != 1)
1479 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1483 while (chunksize > 0)
1485 if (NULL == (from_p = strstr(from_p, "\r\n")))
1487 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1491 if ((newsize += chunksize) >= size)
1493 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
1498 memmove(to_p, from_p, (size_t) chunksize);
1499 to_p = buffer + newsize;
1500 from_p += chunksize + 2;
1502 if (sscanf(from_p, "%x", &chunksize) != 1)
1504 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1509 /* FIXME: Should this get its own loglevel? */
1510 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1516 /*********************************************************************
1518 * Function : url_actions
1520 * Description : Gets the actions for this URL.
1523 * 1 : http = http_request request for blocked URLs
1524 * 2 : csp = Current client state (buffers, headers, etc...)
1528 *********************************************************************/
1529 void url_actions(struct http_request *http,
1530 struct client_state *csp)
1532 struct file_list *fl;
1533 struct url_actions *b;
1536 init_current_action(csp->action);
1538 for (i = 0; i < MAX_ACTION_FILES; i++)
1540 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1545 apply_url_actions(csp->action, http, b);
1552 /*********************************************************************
1554 * Function : apply_url_actions
1556 * Description : Applies a list of URL actions.
1559 * 1 : action = Destination.
1560 * 2 : http = Current URL
1561 * 3 : b = list of URL actions to apply
1565 *********************************************************************/
1566 void apply_url_actions(struct current_action_spec *action,
1567 struct http_request *http,
1568 struct url_actions *b)
1572 /* Should never happen */
1576 for (b = b->next; NULL != b; b = b->next)
1578 if (url_match(b->url, http))
1580 merge_current_action(action, b->action);
1586 /*********************************************************************
1588 * Function : forward_url
1590 * Description : Should we forward this to another proxy?
1593 * 1 : http = http_request request for current URL
1594 * 2 : csp = Current client state (buffers, headers, etc...)
1596 * Returns : Pointer to forwarding information.
1598 *********************************************************************/
1599 const struct forward_spec * forward_url(struct http_request *http,
1600 struct client_state *csp)
1602 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1603 struct forward_spec *fwd = csp->config->forward;
1612 if (url_match(fwd->url, http))
1623 /*********************************************************************
1625 * Function : direct_response
1627 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
1628 * request and if so, return a HTTP 501 to the client.
1630 * FIXME: I have a stupid name and I should handle the
1631 * requests properly. Still, what we do here is rfc-
1632 * compliant, whereas ignoring or forwarding are not.
1635 * 1 : csp = Current client state (buffers, headers, etc...)
1637 * Returns : http_response if , NULL if nonmatch or handler fail
1639 *********************************************************************/
1640 struct http_response *direct_response(struct client_state *csp)
1642 struct http_response *rsp;
1643 struct list_entry *p;
1645 if ((0 == strcmpic(csp->http->gpc, "trace"))
1646 || (0 == strcmpic(csp->http->gpc, "options")))
1648 for (p = csp->headers->first; (p != NULL) ; p = p->next)
1650 if (!strncmp("Max-Forwards:", p->str, 13)
1651 && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
1653 /* FIXME: We could handle at least TRACE here,
1654 but that would require a verbatim copy of
1655 the request which we don't have anymore */
1657 log_error(LOG_LEVEL_HEADER, "Found Max-Forwards:0 in OPTIONS or TRACE request -- Returning 501");
1659 /* Get mem for response or fail*/
1660 if (NULL == (rsp = alloc_http_response()))
1662 return cgi_error_memory();
1665 if (NULL == (rsp->status = strdup("501 Not Implemented")))
1667 free_http_response(rsp);
1668 return cgi_error_memory();
1672 return(finish_http_response(rsp));