1 const char filters_rcs[] = "$Id: filters.c,v 1.58.2.1 2002/07/26 15:18:53 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 1.58.2.1 2002/07/26 15:18:53 oes
42 * - Bugfix: Executing a filters without jobs no longer results in
43 * turing off *all* filters.
44 * - Security fix: Malicious web servers can't cause a seg fault
45 * through bogus chunk sizes anymore
47 * Revision 1.58 2002/04/24 02:11:17 oes
48 * Jon's multiple AF patch: url_actions now evaluates rules
51 * Revision 1.57 2002/04/08 20:38:34 swa
54 * Revision 1.56 2002/04/05 15:51:24 oes
55 * - bugfix: error-pages now get correct request protocol
56 * - fix for invalid HTML in trust info
58 * Revision 1.55 2002/04/02 16:13:51 oes
59 * Fix: No "Go there anyway" for SSL
61 * Revision 1.54 2002/04/02 14:55:56 oes
62 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
64 * Revision 1.53 2002/03/26 22:29:54 swa
65 * we have a new homepage!
67 * Revision 1.52 2002/03/24 16:35:57 jongfoster
70 * Revision 1.51 2002/03/24 15:23:33 jongfoster
73 * Revision 1.50 2002/03/24 13:25:43 swa
74 * name change related issues
76 * Revision 1.49 2002/03/16 20:29:14 oes
79 * Revision 1.48 2002/03/13 20:25:34 oes
80 * Better logging for content filters
82 * Revision 1.47 2002/03/13 00:30:52 jongfoster
84 * Added option of always sending redirect for imageblock,
85 * currently disabled with #if 0.
87 * Revision 1.46 2002/03/12 01:42:49 oes
88 * Introduced modular filters
90 * Revision 1.45 2002/03/08 16:47:50 oes
91 * Added choice beween GIF and PNG built-in images
93 * Revision 1.44 2002/03/07 03:49:31 oes
94 * - Fixed compiler warnings etc
95 * - Changed built-in images from GIF to PNG
96 * (with regard to Unisys patent issue)
97 * - Added a 4x4 pattern PNG which is less intrusive
98 * than the logo but also clearly marks the deleted banners
100 * Revision 1.43 2002/01/22 23:51:59 jongfoster
101 * Replacing strsav() with the safer string_append().
103 * Adding missing html_encode() to error message generators. Where encoded
104 * and unencoded versions of a string were provided, removing the unencoded
107 * Revision 1.42 2002/01/17 21:00:32 jongfoster
108 * Moving all our URL and URL pattern parsing code to urlmatch.c.
110 * Using a single, simple url_match(pattern,url) function - rather than
111 * the 3-line match routine which was repeated all over the place.
113 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
115 * Using parse_http_url() to parse URLs without faking a HTTP
116 * request line for parse_http_request().
118 * Revision 1.41 2001/11/13 00:14:07 jongfoster
119 * Fixing stupid bug now I've figured out what || means.
120 * (It always returns 0 or 1, not one of it's paramaters.)
122 * Revision 1.40 2001/10/26 17:37:55 oes
123 * - Re-enabled Netscape 200/404 bug workaround in block_url():
124 * - Removed OS/2 special case
125 * - Made block_url() independant from sed() having been run
126 * - Made trust_url independant from sed() having been run
127 * - Made is_imageurl independant from sed() having been run.
128 * It now checks User-Agent: and Accept: by itself.
131 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
132 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
133 * threads to call select() simultaneously. So, it's time to do a real, live,
134 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
135 * (native). Both versions will work, but using __OS2__ offers multi-threading.
137 * Revision 1.38 2001/10/23 21:32:33 jongfoster
138 * Adding error-checking to selected functions
140 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
141 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
142 * filters.c. Added a FIXME in front of the offending code. I'll gladly
143 * put in a better/more robust fix for all parties if one is presented...
144 * It seems that just returning 200 instead of 404 would pretty much fix
145 * it for everyone, but I don't know all the history of the problem.
147 * Revision 1.36 2001/10/10 16:44:16 oes
148 * Added match_portlist function
150 * Revision 1.35 2001/10/07 15:41:23 oes
151 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
153 * New function remove_chunked_transfer_coding that strips chunked
154 * transfer coding to plain and is called by pcrs_filter_response
155 * and gif_deanimate_response if neccessary
157 * Improved handling of zero-change re_filter runs
159 * pcrs_filter_response and gif_deanimate_response now remove
160 * chunked transfer codeing before processing the body.
162 * Revision 1.34 2001/09/20 15:49:36 steudten
164 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
165 * See cgi.c fill_template().
167 * Revision 1.33 2001/09/16 17:05:14 jongfoster
168 * Removing unused #include showarg.h
170 * Revision 1.32 2001/09/16 13:21:27 jongfoster
171 * Changes to use new list functions.
173 * Revision 1.31 2001/09/16 11:38:02 jongfoster
174 * Splitting fill_template() into 2 functions:
175 * template_load() loads the file
176 * template_fill() performs the PCRS regexps.
177 * This is because the CGI edit interface has a "table row"
178 * template which is used many times in the page - this
179 * change means it's only loaded from disk once.
181 * Revision 1.30 2001/09/16 11:00:10 jongfoster
182 * New function alloc_http_response, for symmetry with free_http_response
184 * Revision 1.29 2001/09/13 23:32:40 jongfoster
185 * Moving image data to cgi.c rather than cgi.h
186 * Fixing a GPF under Win32 (and any other OS that protects global
187 * constants from being written to).
189 * Revision 1.28 2001/09/10 10:18:51 oes
190 * Silenced compiler warnings
192 * Revision 1.27 2001/08/05 16:06:20 jongfoster
193 * Modifiying "struct map" so that there are now separate header and
194 * "map_entry" structures. This means that functions which modify a
195 * map no longer need to return a pointer to the modified map.
196 * Also, it no longer reverses the order of the entries (which may be
197 * important with some advanced template substitutions).
199 * Revision 1.26 2001/07/30 22:08:36 jongfoster
200 * Tidying up #defines:
201 * - All feature #defines are now of the form FEATURE_xxx
202 * - Permanently turned off WIN_GUI_EDIT
203 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
205 * Revision 1.25 2001/07/26 10:09:46 oes
206 * Made browser detection a little less naive
208 * Revision 1.24 2001/07/25 17:22:51 oes
209 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
211 * Revision 1.23 2001/07/23 13:40:12 oes
212 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
214 * Revision 1.22 2001/07/18 12:29:34 oes
215 * - Made gif_deanimate_response respect
216 * csp->action->string[ACTION_STRING_DEANIMATE]
217 * - Logging cosmetics
219 * Revision 1.21 2001/07/13 13:59:53 oes
220 * - Introduced gif_deanimate_response which shares the
221 * generic content modification interface of pcrs_filter_response
222 * and acts as a wrapper to deanimate.c:gif_deanimate()
223 * - Renamed re_process_buffer to pcrs_filter_response
224 * - pcrs_filter_response now returns NULL on failiure
225 * - Removed all #ifdef PCRS
227 * Revision 1.20 2001/07/01 17:01:04 oes
228 * Added comments and missing return statement in is_untrusted_url()
230 * Revision 1.19 2001/06/29 21:45:41 oes
231 * Indentation, CRLF->LF, Tab-> Space
233 * Revision 1.18 2001/06/29 13:27:38 oes
234 * - Cleaned up, renamed and reorderd functions
235 * and improved comments
238 * - Ported to CGI platform. Now delivers
239 * http_response or NULL
240 * - Unified HTML and GIF generation (moved image detection
241 * and GIF generation here from jcc.c:chat())
242 * - Fixed HTTP status to:
243 * - 403 (Forbidden) for the "blocked" HTML message
244 * - 200 (OK) for GIF answers
245 * - 302 (Redirect) for redirect to GIF
248 * - Ported to CGI platform. Now delivers
249 * http_response or NULL
250 * - Separated detection of untrusted URL into
251 * (bool)is_untrusted_url
252 * - Added enforcement of untrusted requests
254 * - Moved redirect_url() from cgi.c to here
255 * and ported it to the CGI platform
257 * - Removed logentry from cancelled commit
259 * Revision 1.17 2001/06/09 10:55:28 jongfoster
260 * Changing BUFSIZ ==> BUFFER_SIZE
262 * Revision 1.16 2001/06/07 23:10:26 jongfoster
263 * Allowing unanchored domain patterns to back off and retry
264 * if they partially match. Optimized right-anchored patterns.
265 * Moving ACL and forward files into config file.
266 * Replacing struct gateway with struct forward_spec
268 * Revision 1.15 2001/06/03 19:12:00 oes
269 * extracted-CGI relevant stuff
271 * Revision 1.14 2001/06/01 10:30:55 oes
272 * Added optional left-anchoring to domaincmp
274 * Revision 1.13 2001/05/31 21:21:30 jongfoster
275 * Permissionsfile / actions file changes:
276 * - Changed "permission" to "action" throughout
277 * - changes to file format to allow string parameters
278 * - Moved helper functions to actions.c
280 * Revision 1.12 2001/05/31 17:35:20 oes
282 * - Enhanced domain part globbing with infix and prefix asterisk
283 * matching and optional unanchored operation
285 * Revision 1.11 2001/05/29 11:53:23 oes
286 * "See why" link added to "blocked" page
288 * Revision 1.10 2001/05/29 09:50:24 jongfoster
289 * Unified blocklist/imagelist/permissionslist.
290 * File format is still under discussion, but the internal changes
293 * Also modified interceptor behaviour:
294 * - We now intercept all URLs beginning with one of the following
295 * prefixes (and *only* these prefixes):
297 * * http://ijbswa.sf.net/config/
298 * * http://ijbswa.sourceforge.net/config/
299 * - New interceptors "home page" - go to http://i.j.b/ to see it.
300 * - Internal changes so that intercepted and fast redirect pages
301 * are not replaced with an image.
302 * - Interceptors now have the option to send a binary page direct
303 * to the client. (i.e. ijb-send-banner uses this)
304 * - Implemented show-url-info interceptor. (Which is why I needed
305 * the above interceptors changes - a typical URL is
306 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
307 * The previous mechanism would not have intercepted that, and
308 * if it had been intercepted then it then it would have replaced
311 * Revision 1.9 2001/05/27 22:17:04 oes
313 * - re_process_buffer no longer writes the modified buffer
314 * to the client, which was very ugly. It now returns the
315 * buffer, which it is then written by chat.
317 * - content_length now adjusts the Content-Length: header
318 * for modified documents rather than crunch()ing it.
319 * (Length info in csp->content_length, which is 0 for
320 * unmodified documents)
322 * - For this to work, sed() is called twice when filtering.
324 * Revision 1.8 2001/05/26 17:13:28 jongfoster
325 * Filled in a function comment.
327 * Revision 1.7 2001/05/26 15:26:15 jongfoster
328 * ACL feature now provides more security by immediately dropping
329 * connections from untrusted hosts.
331 * Revision 1.6 2001/05/26 00:28:36 jongfoster
332 * Automatic reloading of config file.
333 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
334 * Most of the global variables have been moved to a new
335 * struct configuration_spec, accessed through csp->config->globalname
336 * Most of the globals remaining are used by the Win32 GUI.
338 * Revision 1.5 2001/05/25 22:34:30 jongfoster
341 * Revision 1.4 2001/05/22 18:46:04 oes
343 * - Enabled filtering banners by size rather than URL
344 * by adding patterns that replace all standard banner
345 * sizes with the "Junkbuster" gif to the re_filterfile
347 * - Enabled filtering WebBugs by providing a pattern
348 * which kills all 1x1 images
350 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
351 * which is selected by the (nonstandard and therefore
352 * capital) letter 'U' in the option string.
353 * It causes the quantifiers to be ungreedy by default.
354 * Appending a ? turns back to greedy (!).
356 * - Added a new interceptor ijb-send-banner, which
357 * sends back the "Junkbuster" gif. Without imagelist or
358 * MSIE detection support, or if tinygif = 1, or the
359 * URL isn't recognized as an imageurl, a lame HTML
360 * explanation is sent instead.
362 * - Added new feature, which permits blocking remote
363 * script redirects and firing back a local redirect
365 * The feature is conditionally compiled, i.e. it
366 * can be disabled with --disable-fast-redirects,
367 * plus it must be activated by a "fast-redirects"
368 * line in the config file, has its own log level
369 * and of course wants to be displayed by show-proxy-args
370 * Note: Boy, all the #ifdefs in 1001 locations and
371 * all the fumbling with configure.in and acconfig.h
372 * were *way* more work than the feature itself :-(
374 * - Because a generic redirect template was needed for
375 * this, tinygif = 3 now uses the same.
377 * - Moved GIFs, and other static HTTP response templates
382 * - Removed some >400 CRs again (Jon, you really worked
385 * Revision 1.3 2001/05/20 16:44:47 jongfoster
386 * Removing last hardcoded Junkbusters.com URLs.
388 * Revision 1.2 2001/05/20 01:21:20 jongfoster
389 * Version 2.9.4 checkin.
390 * - Merged popupfile and cookiefile, and added control over PCRS
391 * filtering, in new "permissionsfile".
392 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
393 * file error you now get a message box (in the Win32 GUI) rather
394 * than the program exiting with no explanation.
395 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
397 * - Removed tabs from "config"
398 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
399 * - Bumped up version number.
401 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
402 * Initial import of version 2.9.3 source tree
405 *********************************************************************/
411 #include <sys/types.h>
420 #endif /* ndef __OS2__ */
421 #include <netinet/in.h>
423 #include <winsock2.h>
424 #endif /* ndef _WIN32 */
428 #endif /* def __OS2__ */
436 #include "jbsockets.h"
437 #include "miscutil.h"
441 #include "deanimate.h"
442 #include "urlmatch.h"
448 const char filters_h_rcs[] = FILTERS_H_VERSION;
450 /* Fix a problem with Solaris. There should be no effect on other
452 * Solaris's isspace() is a macro which uses it's argument directly
453 * as an array index. Therefore we need to make sure that high-bit
454 * characters generate +ve values, and ideally we also want to make
455 * the argument match the declared parameter type of "int".
457 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
461 /*********************************************************************
463 * Function : block_acl
465 * Description : Block this request?
466 * Decide yes or no based on ACL file.
469 * 1 : dst = The proxy or gateway address this is going to.
470 * Or NULL to check all possible targets.
471 * 2 : csp = Current client state (buffers, headers, etc...)
472 * Also includes the client IP address.
474 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
476 *********************************************************************/
477 int block_acl(struct access_control_addr *dst, struct client_state *csp)
479 struct access_control_list *acl = csp->config->acl;
481 /* if not using an access control list, then permit the connection */
487 /* search the list */
490 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
494 /* Just want to check if they have any access */
495 if (acl->action == ACL_PERMIT)
500 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
501 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
503 if (acl->action == ACL_PERMIT)
521 /*********************************************************************
523 * Function : acl_addr
525 * Description : Called from `load_config' to parse an ACL address.
528 * 1 : aspec = String specifying ACL address.
529 * 2 : aca = struct access_control_addr to fill in.
531 * Returns : 0 => Ok, everything else is an error.
533 *********************************************************************/
534 int acl_addr(char *aspec, struct access_control_addr *aca)
536 int i, masklength, port;
542 if ((p = strchr(aspec, '/')) != NULL)
546 if (ijb_isdigit(*p) == 0)
550 masklength = atoi(p);
553 if ((masklength < 0) || (masklength > 32))
558 if ((p = strchr(aspec, ':')) != NULL)
562 if (ijb_isdigit(*p) == 0)
571 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
573 if (aca->addr == INADDR_NONE)
578 /* build the netmask */
580 for (i=1; i <= masklength ; i++)
582 aca->mask |= (1 << (32 - i));
585 /* now mask off the host portion of the ip address
586 * (i.e. save on the network portion of the address).
588 aca->addr = aca->addr & aca->mask;
593 #endif /* def FEATURE_ACL */
596 /*********************************************************************
598 * Function : match_portlist
600 * Description : Check if a given number is covered by a comma
601 * separated list of numbers and ranges (a,b-c,d,..)
604 * 1 : portlist = String with list
605 * 2 : port = port to check
607 * Returns : 0 => no match
610 *********************************************************************/
611 int match_portlist(const char *portlist, int port)
613 char *min, *max, *next, *portlist_copy;
615 min = next = portlist_copy = strdup(portlist);
618 * Zero-terminate first item and remember offset for next
620 if (NULL != (next = strchr(portlist_copy, (int) ',')))
626 * Loop through all items, checking for match
630 if (NULL == (max = strchr(min, (int) '-')))
633 * No dash, check for equality
635 if (port == atoi(min))
644 * This is a range, so check if between min and max,
645 * or, if max was omitted, between min and 65K
648 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
662 * Zero-terminate next item and remember offset for n+1
664 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
676 /*********************************************************************
678 * Function : block_url
680 * Description : Called from `chat'. Check to see if we need to block this.
683 * 1 : csp = Current client state (buffers, headers, etc...)
685 * Returns : NULL => unblocked, else HTTP block response
687 *********************************************************************/
688 struct http_response *block_url(struct client_state *csp)
690 #ifdef FEATURE_IMAGE_BLOCKING
692 #endif /* def FEATURE_IMAGE_BLOCKING */
693 struct http_response *rsp;
696 * If it's not blocked, don't block it ;-)
698 if ((csp->action->flags & ACTION_BLOCK) == 0)
704 * Else, prepare a response
706 if (NULL == (rsp = alloc_http_response()))
708 return cgi_error_memory();
712 * If it's an image-url, send back an image or redirect
713 * as specified by the relevant +image action
715 #ifdef FEATURE_IMAGE_BLOCKING
716 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
719 /* determine HOW images should be blocked */
720 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
722 #if 1 /* Two alternative strategies, use this one for now: */
724 /* and handle accordingly: */
725 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
727 rsp->body = bindup(image_pattern_data, image_pattern_length);
728 if (rsp->body == NULL)
730 free_http_response(rsp);
731 return cgi_error_memory();
733 rsp->content_length = image_pattern_length;
735 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
737 free_http_response(rsp);
738 return cgi_error_memory();
742 else if (0 == strcmpic(p, "blank"))
744 rsp->body = bindup(image_blank_data, image_blank_length);
745 if (rsp->body == NULL)
747 free_http_response(rsp);
748 return cgi_error_memory();
750 rsp->content_length = image_blank_length;
752 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
754 free_http_response(rsp);
755 return cgi_error_memory();
761 rsp->status = strdup("302 Local Redirect from Privoxy");
762 if (rsp->status == NULL)
764 free_http_response(rsp);
765 return cgi_error_memory();
768 if (enlist_unique_header(rsp->headers, "Location", p))
770 free_http_response(rsp);
771 return cgi_error_memory();
775 #else /* Following code is disabled for now */
777 /* and handle accordingly: */
778 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
780 p = CGI_PREFIX "send-banner?type=pattern";
782 else if (0 == strcmpic(p, "blank"))
784 p = CGI_PREFIX "send-banner?type=blank";
786 rsp->status = strdup("302 Local Redirect from Privoxy");
787 if (rsp->status == NULL)
789 free_http_response(rsp);
790 return cgi_error_memory();
793 if (enlist_unique_header(rsp->headers, "Location", p))
795 free_http_response(rsp);
796 return cgi_error_memory();
798 #endif /* Preceeding code is disabled for now */
801 #endif /* def FEATURE_IMAGE_BLOCKING */
804 * Else, generate an HTML "blocked" message:
808 struct map * exports;
811 * Workaround for stupid Netscape bug which prevents
812 * pages from being displayed if loading a referenced
813 * JavaScript or style sheet fails. So make it appear
814 * as if it succeeded.
816 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
817 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
818 && !strstr(p, "Gecko") /* save Mozilla, */
819 && !strstr(p, "compatible") /* MSIE */
820 && !strstr(p, "Opera")) /* and Opera. */
822 rsp->status = strdup("200 Request for blocked URL");
826 rsp->status = strdup("404 Request for blocked URL");
829 if (rsp->status == NULL)
831 free_http_response(rsp);
832 return cgi_error_memory();
835 exports = default_exports(csp, NULL);
838 free_http_response(rsp);
839 return cgi_error_memory();
842 #ifdef FEATURE_FORCE_LOAD
843 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
844 if (csp->http->ssl != 0)
845 #endif /* ndef FEATURE_FORCE_LOAD */
847 err = map_block_killer(exports, "force-support");
850 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
851 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
852 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
857 free_http_response(rsp);
858 return cgi_error_memory();
861 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
864 free_http_response(rsp);
865 return cgi_error_memory();
869 return finish_http_response(rsp);
875 /*********************************************************************
877 * Function : trust_url FIXME: I should be called distrust_url
879 * Description : Calls is_untrusted_url to determine if the URL is trusted
880 * and if not, returns a HTTP 304 response with a reject message.
883 * 1 : csp = Current client state (buffers, headers, etc...)
885 * Returns : NULL => trusted, else http_response.
887 *********************************************************************/
888 struct http_response *trust_url(struct client_state *csp)
890 struct http_response *rsp;
891 struct map * exports;
892 char buf[BUFFER_SIZE];
894 struct url_spec **tl;
899 * Don't bother to work on trusted URLs
901 if (!is_untrusted_url(csp))
907 * Else, prepare a response:
909 if (NULL == (rsp = alloc_http_response()))
911 return cgi_error_memory();
914 exports = default_exports(csp, NULL);
917 free_http_response(rsp);
918 return cgi_error_memory();
922 * Export the protocol, host, port, and referrer information
924 err = map(exports, "hostport", 1, csp->http->hostport, 1);
925 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
926 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
928 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
930 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
934 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
940 free_http_response(rsp);
941 return cgi_error_memory();
945 * Export the trust list
948 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
950 sprintf(buf, "<li>%s</li>\n", t->spec);
951 string_append(&p, buf);
953 err = map(exports, "trusted-referrers", 1, p, 0);
958 free_http_response(rsp);
959 return cgi_error_memory();
963 * Export the trust info, if available
965 if (csp->config->trust_info->first)
967 struct list_entry *l;
970 for (l = csp->config->trust_info->first; l ; l = l->next)
972 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
973 string_append(&p, buf);
975 err = map(exports, "trust-info", 1, p, 0);
979 err = map_block_killer(exports, "have-trust-info");
985 free_http_response(rsp);
986 return cgi_error_memory();
990 * Export the force prefix or the force conditional block killer
992 #ifdef FEATURE_FORCE_LOAD
993 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
994 #else /* ifndef FEATURE_FORCE_LOAD */
995 err = map_block_killer(exports, "force-support");
996 #endif /* ndef FEATURE_FORCE_LOAD */
1001 free_http_response(rsp);
1002 return cgi_error_memory();
1006 * Build the response
1008 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1011 free_http_response(rsp);
1012 return cgi_error_memory();
1015 return finish_http_response(rsp);
1017 #endif /* def FEATURE_TRUST */
1020 #ifdef FEATURE_FAST_REDIRECTS
1021 /*********************************************************************
1023 * Function : redirect_url
1025 * Description : Checks for redirection URLs and returns a HTTP redirect
1026 * to the destination URL, if necessary
1029 * 1 : csp = Current client state (buffers, headers, etc...)
1031 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1033 *********************************************************************/
1034 struct http_response *redirect_url(struct client_state *csp)
1037 struct http_response *rsp;
1039 p = q = csp->http->path;
1040 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1043 * find the last URL encoded in the request
1045 while ((p = strstr(p, "http://")) != NULL)
1051 * if there was any, generate and return a HTTP redirect
1053 if (q != csp->http->path)
1055 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1057 if (NULL == (rsp = alloc_http_response()))
1059 return cgi_error_memory();
1062 if ( enlist_unique_header(rsp->headers, "Location", q)
1063 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1065 free_http_response(rsp);
1066 return cgi_error_memory();
1069 return finish_http_response(rsp);
1077 #endif /* def FEATURE_FAST_REDIRECTS */
1080 #ifdef FEATURE_IMAGE_BLOCKING
1081 /*********************************************************************
1083 * Function : is_imageurl
1085 * Description : Given a URL, decide whether it is an image or not,
1086 * using either the info from a previous +image action
1087 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1088 * is MSIE and not on a Mac, tell from the browser's accept
1092 * 1 : csp = Current client state (buffers, headers, etc...)
1094 * Returns : True (nonzero) if URL is an image, false (0)
1097 *********************************************************************/
1098 int is_imageurl(struct client_state *csp)
1100 #ifdef FEATURE_IMAGE_DETECT_MSIE
1103 tmp = get_header_value(csp->headers, "User-Agent:");
1104 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1106 tmp = get_header_value(csp->headers, "Accept:");
1107 if (tmp && strstr(tmp, "image/gif"))
1109 /* Client will accept HTML. If this seems counterintuitive,
1119 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1121 return ((csp->action->flags & ACTION_IMAGE) != 0);
1124 #endif /* def FEATURE_IMAGE_BLOCKING */
1127 #ifdef FEATURE_TRUST
1128 /*********************************************************************
1130 * Function : is_untrusted_url
1132 * Description : Should we "distrust" this URL (and block it)?
1134 * Yes if it matches a line in the trustfile, or if the
1135 * referrer matches a line starting with "+" in the
1140 * 1 : csp = Current client state (buffers, headers, etc...)
1142 * Returns : 0 => trusted, 1 => untrusted
1144 *********************************************************************/
1145 int is_untrusted_url(struct client_state *csp)
1147 struct file_list *fl;
1148 struct block_spec *b;
1149 struct url_spec **trusted_url;
1150 struct http_request rhttp[1];
1151 const char * referer;
1155 * If we don't have a trustlist, we trust everybody
1157 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1162 memset(rhttp, '\0', sizeof(*rhttp));
1165 * Do we trust the request URL itself?
1167 for (b = b->next; b ; b = b->next)
1169 if (url_match(b->url, csp->http))
1175 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1177 /* no referrer was supplied */
1183 * If not, do we maybe trust its referrer?
1185 err = parse_http_url(referer, rhttp, csp);
1191 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1193 if (url_match(*trusted_url, rhttp))
1195 /* if the URL's referrer is from a trusted referrer, then
1196 * add the target spec to the trustfile as an unblocked
1197 * domain and return NULL (which means it's OK).
1202 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1206 char * new_entry = strdup("~");
1208 string_append(&new_entry, csp->http->hostport);
1210 path = csp->http->path;
1211 if ( (path[0] == '/')
1213 && ((path_end = strchr(path + 2, '/')) != NULL))
1215 /* since this path points into a user's home space
1216 * be sure to include this spec in the trustfile.
1218 int path_len = path_end - path; /* save offset */
1219 path = strdup(path); /* Copy string */
1222 path_end = path + path_len; /* regenerate ptr to new buffer */
1223 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1225 string_join(&new_entry, path);
1228 if (new_entry != NULL)
1230 fprintf(fp, "%s\n", new_entry);
1235 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1236 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1246 #endif /* def FEATURE_TRUST */
1249 /*********************************************************************
1251 * Function : pcrs_filter_response
1253 * Description : Ecexute all text substitutions from all applying
1254 * +filter actions on the text buffer that's been accumulated
1255 * in csp->iob->buf. If this changes the contents, set
1256 * csp->content_length to the modified size and raise the
1257 * CSP_FLAG_MODIFIED flag.
1260 * 1 : csp = Current client state (buffers, headers, etc...)
1262 * Returns : a pointer to the (newly allocated) modified buffer.
1263 * or NULL if there were no hits or something went wrong
1265 *********************************************************************/
1266 char *pcrs_filter_response(struct client_state *csp)
1271 char *old = csp->iob->cur, *new = NULL;
1274 struct file_list *fl;
1275 struct re_filterfile_spec *b;
1276 struct list_entry *filtername;
1281 if (csp->iob->cur >= csp->iob->eod)
1285 size = csp->iob->eod - csp->iob->cur;
1287 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1289 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1294 * If the body has a "chunked" transfer-encoding,
1295 * get rid of it first, adjusting size and iob->eod
1297 if (csp->flags & CSP_FLAG_CHUNKED)
1299 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1300 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1304 csp->iob->eod = csp->iob->cur + size;
1305 csp->flags |= CSP_FLAG_MODIFIED;
1309 * For all applying +filter actions, look if a filter by that
1310 * name exists and if yes, execute it's pcrs_joblist on the
1313 for (b = fl->f; b; b = b->next)
1315 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1316 filtername ; filtername = filtername->next)
1318 if (strcmp(b->name, filtername->str) == 0)
1320 int current_hits = 0;
1322 if ( NULL == b->joblist )
1324 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1328 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1329 csp->http->hostport, csp->http->path, size, b->name);
1331 /* Apply all jobs from the joblist */
1332 for (job = b->joblist; NULL != job; job = job->next)
1334 current_hits += pcrs_execute(job, old, size, &new, &size);
1335 if (old != csp->iob->cur) free(old);
1339 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1340 hits += current_hits;
1346 * If there were no hits, destroy our copy and let
1347 * chat() use the original in csp->iob
1355 csp->flags |= CSP_FLAG_MODIFIED;
1356 csp->content_length = size;
1364 /*********************************************************************
1366 * Function : gif_deanimate_response
1368 * Description : Deanimate the GIF image that has been accumulated in
1369 * csp->iob->buf, set csp->content_length to the modified
1370 * size and raise the CSP_FLAG_MODIFIED flag.
1373 * 1 : csp = Current client state (buffers, headers, etc...)
1375 * Returns : a pointer to the (newly allocated) modified buffer.
1376 * or NULL in case something went wrong.
1378 *********************************************************************/
1379 char *gif_deanimate_response(struct client_state *csp)
1381 struct binbuffer *in, *out;
1383 size_t size = csp->iob->eod - csp->iob->cur;
1386 * If the body has a "chunked" transfer-encoding,
1387 * get rid of it first, adjusting size and iob->eod
1389 if (csp->flags & CSP_FLAG_CHUNKED)
1391 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1392 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1396 csp->iob->eod = csp->iob->cur + size;
1397 csp->flags |= CSP_FLAG_MODIFIED;
1400 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1401 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1403 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1407 in->buffer = csp->iob->cur;
1410 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1412 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1419 if ((int)size == out->offset)
1421 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1425 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1427 csp->content_length = out->offset;
1428 csp->flags |= CSP_FLAG_MODIFIED;
1438 /*********************************************************************
1440 * Function : remove_chunked_transfer_coding
1442 * Description : In-situ remove the "chunked" transfer coding as defined
1443 * in rfc2616 from a buffer.
1446 * 1 : buffer = Pointer to the text buffer
1447 * 2 : size = Number of bytes to be processed
1449 * Returns : The new size, i.e. the number of bytes from buffer which
1450 * are occupied by the stripped body, or 0 in case something
1453 *********************************************************************/
1454 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1457 unsigned int chunksize = 0;
1458 char *from_p, *to_p;
1461 from_p = to_p = buffer;
1463 if (sscanf(buffer, "%x", &chunksize) != 1)
1465 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1469 while (chunksize > 0)
1471 if (NULL == (from_p = strstr(from_p, "\r\n")))
1473 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1477 if ((newsize += chunksize) >= size)
1479 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
1484 memmove(to_p, from_p, (size_t) chunksize);
1485 to_p = buffer + newsize;
1486 from_p += chunksize + 2;
1488 if (sscanf(from_p, "%x", &chunksize) != 1)
1490 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1495 /* FIXME: Should this get its own loglevel? */
1496 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1502 /*********************************************************************
1504 * Function : url_actions
1506 * Description : Gets the actions for this URL.
1509 * 1 : http = http_request request for blocked URLs
1510 * 2 : csp = Current client state (buffers, headers, etc...)
1514 *********************************************************************/
1515 void url_actions(struct http_request *http,
1516 struct client_state *csp)
1518 struct file_list *fl;
1519 struct url_actions *b;
1522 init_current_action(csp->action);
1524 for (i = 0; i < MAX_ACTION_FILES; i++)
1526 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
1531 apply_url_actions(csp->action, http, b);
1538 /*********************************************************************
1540 * Function : apply_url_actions
1542 * Description : Applies a list of URL actions.
1545 * 1 : action = Destination.
1546 * 2 : http = Current URL
1547 * 3 : b = list of URL actions to apply
1551 *********************************************************************/
1552 void apply_url_actions(struct current_action_spec *action,
1553 struct http_request *http,
1554 struct url_actions *b)
1558 /* Should never happen */
1562 for (b = b->next; NULL != b; b = b->next)
1564 if (url_match(b->url, http))
1566 merge_current_action(action, b->action);
1572 /*********************************************************************
1574 * Function : forward_url
1576 * Description : Should we forward this to another proxy?
1579 * 1 : http = http_request request for current URL
1580 * 2 : csp = Current client state (buffers, headers, etc...)
1582 * Returns : Pointer to forwarding information.
1584 *********************************************************************/
1585 const struct forward_spec * forward_url(struct http_request *http,
1586 struct client_state *csp)
1588 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1589 struct forward_spec *fwd = csp->config->forward;
1598 if (url_match(fwd->url, http))