1 const char filters_rcs[] = "$Id: filters.c,v 1.56 2002/04/05 15:51:24 oes Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/filters.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
7 * Functions declared include:
8 * `acl_addr', `add_stats', `block_acl', `block_imageurl',
9 * `block_url', `url_actions', `domain_split',
10 * `filter_popups', `forward_url', 'redirect_url',
11 * `ij_untrusted_url', `intercept_url', `pcrs_filter_respose',
12 * 'ijb_send_banner', 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 1.56 2002/04/05 15:51:24 oes
42 * - bugfix: error-pages now get correct request protocol
43 * - fix for invalid HTML in trust info
45 * Revision 1.55 2002/04/02 16:13:51 oes
46 * Fix: No "Go there anyway" for SSL
48 * Revision 1.54 2002/04/02 14:55:56 oes
49 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
51 * Revision 1.53 2002/03/26 22:29:54 swa
52 * we have a new homepage!
54 * Revision 1.52 2002/03/24 16:35:57 jongfoster
57 * Revision 1.51 2002/03/24 15:23:33 jongfoster
60 * Revision 1.50 2002/03/24 13:25:43 swa
61 * name change related issues
63 * Revision 1.49 2002/03/16 20:29:14 oes
66 * Revision 1.48 2002/03/13 20:25:34 oes
67 * Better logging for content filters
69 * Revision 1.47 2002/03/13 00:30:52 jongfoster
71 * Added option of always sending redirect for imageblock,
72 * currently disabled with #if 0.
74 * Revision 1.46 2002/03/12 01:42:49 oes
75 * Introduced modular filters
77 * Revision 1.45 2002/03/08 16:47:50 oes
78 * Added choice beween GIF and PNG built-in images
80 * Revision 1.44 2002/03/07 03:49:31 oes
81 * - Fixed compiler warnings etc
82 * - Changed built-in images from GIF to PNG
83 * (with regard to Unisys patent issue)
84 * - Added a 4x4 pattern PNG which is less intrusive
85 * than the logo but also clearly marks the deleted banners
87 * Revision 1.43 2002/01/22 23:51:59 jongfoster
88 * Replacing strsav() with the safer string_append().
90 * Adding missing html_encode() to error message generators. Where encoded
91 * and unencoded versions of a string were provided, removing the unencoded
94 * Revision 1.42 2002/01/17 21:00:32 jongfoster
95 * Moving all our URL and URL pattern parsing code to urlmatch.c.
97 * Using a single, simple url_match(pattern,url) function - rather than
98 * the 3-line match routine which was repeated all over the place.
100 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
102 * Using parse_http_url() to parse URLs without faking a HTTP
103 * request line for parse_http_request().
105 * Revision 1.41 2001/11/13 00:14:07 jongfoster
106 * Fixing stupid bug now I've figured out what || means.
107 * (It always returns 0 or 1, not one of it's paramaters.)
109 * Revision 1.40 2001/10/26 17:37:55 oes
110 * - Re-enabled Netscape 200/404 bug workaround in block_url():
111 * - Removed OS/2 special case
112 * - Made block_url() independant from sed() having been run
113 * - Made trust_url independant from sed() having been run
114 * - Made is_imageurl independant from sed() having been run.
115 * It now checks User-Agent: and Accept: by itself.
118 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
119 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
120 * threads to call select() simultaneously. So, it's time to do a real, live,
121 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
122 * (native). Both versions will work, but using __OS2__ offers multi-threading.
124 * Revision 1.38 2001/10/23 21:32:33 jongfoster
125 * Adding error-checking to selected functions
127 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
128 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
129 * filters.c. Added a FIXME in front of the offending code. I'll gladly
130 * put in a better/more robust fix for all parties if one is presented...
131 * It seems that just returning 200 instead of 404 would pretty much fix
132 * it for everyone, but I don't know all the history of the problem.
134 * Revision 1.36 2001/10/10 16:44:16 oes
135 * Added match_portlist function
137 * Revision 1.35 2001/10/07 15:41:23 oes
138 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
140 * New function remove_chunked_transfer_coding that strips chunked
141 * transfer coding to plain and is called by pcrs_filter_response
142 * and gif_deanimate_response if neccessary
144 * Improved handling of zero-change re_filter runs
146 * pcrs_filter_response and gif_deanimate_response now remove
147 * chunked transfer codeing before processing the body.
149 * Revision 1.34 2001/09/20 15:49:36 steudten
151 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
152 * See cgi.c fill_template().
154 * Revision 1.33 2001/09/16 17:05:14 jongfoster
155 * Removing unused #include showarg.h
157 * Revision 1.32 2001/09/16 13:21:27 jongfoster
158 * Changes to use new list functions.
160 * Revision 1.31 2001/09/16 11:38:02 jongfoster
161 * Splitting fill_template() into 2 functions:
162 * template_load() loads the file
163 * template_fill() performs the PCRS regexps.
164 * This is because the CGI edit interface has a "table row"
165 * template which is used many times in the page - this
166 * change means it's only loaded from disk once.
168 * Revision 1.30 2001/09/16 11:00:10 jongfoster
169 * New function alloc_http_response, for symmetry with free_http_response
171 * Revision 1.29 2001/09/13 23:32:40 jongfoster
172 * Moving image data to cgi.c rather than cgi.h
173 * Fixing a GPF under Win32 (and any other OS that protects global
174 * constants from being written to).
176 * Revision 1.28 2001/09/10 10:18:51 oes
177 * Silenced compiler warnings
179 * Revision 1.27 2001/08/05 16:06:20 jongfoster
180 * Modifiying "struct map" so that there are now separate header and
181 * "map_entry" structures. This means that functions which modify a
182 * map no longer need to return a pointer to the modified map.
183 * Also, it no longer reverses the order of the entries (which may be
184 * important with some advanced template substitutions).
186 * Revision 1.26 2001/07/30 22:08:36 jongfoster
187 * Tidying up #defines:
188 * - All feature #defines are now of the form FEATURE_xxx
189 * - Permanently turned off WIN_GUI_EDIT
190 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
192 * Revision 1.25 2001/07/26 10:09:46 oes
193 * Made browser detection a little less naive
195 * Revision 1.24 2001/07/25 17:22:51 oes
196 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
198 * Revision 1.23 2001/07/23 13:40:12 oes
199 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
201 * Revision 1.22 2001/07/18 12:29:34 oes
202 * - Made gif_deanimate_response respect
203 * csp->action->string[ACTION_STRING_DEANIMATE]
204 * - Logging cosmetics
206 * Revision 1.21 2001/07/13 13:59:53 oes
207 * - Introduced gif_deanimate_response which shares the
208 * generic content modification interface of pcrs_filter_response
209 * and acts as a wrapper to deanimate.c:gif_deanimate()
210 * - Renamed re_process_buffer to pcrs_filter_response
211 * - pcrs_filter_response now returns NULL on failiure
212 * - Removed all #ifdef PCRS
214 * Revision 1.20 2001/07/01 17:01:04 oes
215 * Added comments and missing return statement in is_untrusted_url()
217 * Revision 1.19 2001/06/29 21:45:41 oes
218 * Indentation, CRLF->LF, Tab-> Space
220 * Revision 1.18 2001/06/29 13:27:38 oes
221 * - Cleaned up, renamed and reorderd functions
222 * and improved comments
225 * - Ported to CGI platform. Now delivers
226 * http_response or NULL
227 * - Unified HTML and GIF generation (moved image detection
228 * and GIF generation here from jcc.c:chat())
229 * - Fixed HTTP status to:
230 * - 403 (Forbidden) for the "blocked" HTML message
231 * - 200 (OK) for GIF answers
232 * - 302 (Redirect) for redirect to GIF
235 * - Ported to CGI platform. Now delivers
236 * http_response or NULL
237 * - Separated detection of untrusted URL into
238 * (bool)is_untrusted_url
239 * - Added enforcement of untrusted requests
241 * - Moved redirect_url() from cgi.c to here
242 * and ported it to the CGI platform
244 * - Removed logentry from cancelled commit
246 * Revision 1.17 2001/06/09 10:55:28 jongfoster
247 * Changing BUFSIZ ==> BUFFER_SIZE
249 * Revision 1.16 2001/06/07 23:10:26 jongfoster
250 * Allowing unanchored domain patterns to back off and retry
251 * if they partially match. Optimized right-anchored patterns.
252 * Moving ACL and forward files into config file.
253 * Replacing struct gateway with struct forward_spec
255 * Revision 1.15 2001/06/03 19:12:00 oes
256 * extracted-CGI relevant stuff
258 * Revision 1.14 2001/06/01 10:30:55 oes
259 * Added optional left-anchoring to domaincmp
261 * Revision 1.13 2001/05/31 21:21:30 jongfoster
262 * Permissionsfile / actions file changes:
263 * - Changed "permission" to "action" throughout
264 * - changes to file format to allow string parameters
265 * - Moved helper functions to actions.c
267 * Revision 1.12 2001/05/31 17:35:20 oes
269 * - Enhanced domain part globbing with infix and prefix asterisk
270 * matching and optional unanchored operation
272 * Revision 1.11 2001/05/29 11:53:23 oes
273 * "See why" link added to "blocked" page
275 * Revision 1.10 2001/05/29 09:50:24 jongfoster
276 * Unified blocklist/imagelist/permissionslist.
277 * File format is still under discussion, but the internal changes
280 * Also modified interceptor behaviour:
281 * - We now intercept all URLs beginning with one of the following
282 * prefixes (and *only* these prefixes):
284 * * http://ijbswa.sf.net/config/
285 * * http://ijbswa.sourceforge.net/config/
286 * - New interceptors "home page" - go to http://i.j.b/ to see it.
287 * - Internal changes so that intercepted and fast redirect pages
288 * are not replaced with an image.
289 * - Interceptors now have the option to send a binary page direct
290 * to the client. (i.e. ijb-send-banner uses this)
291 * - Implemented show-url-info interceptor. (Which is why I needed
292 * the above interceptors changes - a typical URL is
293 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
294 * The previous mechanism would not have intercepted that, and
295 * if it had been intercepted then it then it would have replaced
298 * Revision 1.9 2001/05/27 22:17:04 oes
300 * - re_process_buffer no longer writes the modified buffer
301 * to the client, which was very ugly. It now returns the
302 * buffer, which it is then written by chat.
304 * - content_length now adjusts the Content-Length: header
305 * for modified documents rather than crunch()ing it.
306 * (Length info in csp->content_length, which is 0 for
307 * unmodified documents)
309 * - For this to work, sed() is called twice when filtering.
311 * Revision 1.8 2001/05/26 17:13:28 jongfoster
312 * Filled in a function comment.
314 * Revision 1.7 2001/05/26 15:26:15 jongfoster
315 * ACL feature now provides more security by immediately dropping
316 * connections from untrusted hosts.
318 * Revision 1.6 2001/05/26 00:28:36 jongfoster
319 * Automatic reloading of config file.
320 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
321 * Most of the global variables have been moved to a new
322 * struct configuration_spec, accessed through csp->config->globalname
323 * Most of the globals remaining are used by the Win32 GUI.
325 * Revision 1.5 2001/05/25 22:34:30 jongfoster
328 * Revision 1.4 2001/05/22 18:46:04 oes
330 * - Enabled filtering banners by size rather than URL
331 * by adding patterns that replace all standard banner
332 * sizes with the "Junkbuster" gif to the re_filterfile
334 * - Enabled filtering WebBugs by providing a pattern
335 * which kills all 1x1 images
337 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
338 * which is selected by the (nonstandard and therefore
339 * capital) letter 'U' in the option string.
340 * It causes the quantifiers to be ungreedy by default.
341 * Appending a ? turns back to greedy (!).
343 * - Added a new interceptor ijb-send-banner, which
344 * sends back the "Junkbuster" gif. Without imagelist or
345 * MSIE detection support, or if tinygif = 1, or the
346 * URL isn't recognized as an imageurl, a lame HTML
347 * explanation is sent instead.
349 * - Added new feature, which permits blocking remote
350 * script redirects and firing back a local redirect
352 * The feature is conditionally compiled, i.e. it
353 * can be disabled with --disable-fast-redirects,
354 * plus it must be activated by a "fast-redirects"
355 * line in the config file, has its own log level
356 * and of course wants to be displayed by show-proxy-args
357 * Note: Boy, all the #ifdefs in 1001 locations and
358 * all the fumbling with configure.in and acconfig.h
359 * were *way* more work than the feature itself :-(
361 * - Because a generic redirect template was needed for
362 * this, tinygif = 3 now uses the same.
364 * - Moved GIFs, and other static HTTP response templates
369 * - Removed some >400 CRs again (Jon, you really worked
372 * Revision 1.3 2001/05/20 16:44:47 jongfoster
373 * Removing last hardcoded Junkbusters.com URLs.
375 * Revision 1.2 2001/05/20 01:21:20 jongfoster
376 * Version 2.9.4 checkin.
377 * - Merged popupfile and cookiefile, and added control over PCRS
378 * filtering, in new "permissionsfile".
379 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
380 * file error you now get a message box (in the Win32 GUI) rather
381 * than the program exiting with no explanation.
382 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
384 * - Removed tabs from "config"
385 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
386 * - Bumped up version number.
388 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
389 * Initial import of version 2.9.3 source tree
392 *********************************************************************/
398 #include <sys/types.h>
407 #endif /* ndef __OS2__ */
408 #include <netinet/in.h>
410 #include <winsock2.h>
411 #endif /* ndef _WIN32 */
415 #endif /* def __OS2__ */
423 #include "jbsockets.h"
424 #include "miscutil.h"
428 #include "deanimate.h"
429 #include "urlmatch.h"
435 const char filters_h_rcs[] = FILTERS_H_VERSION;
437 /* Fix a problem with Solaris. There should be no effect on other
439 * Solaris's isspace() is a macro which uses it's argument directly
440 * as an array index. Therefore we need to make sure that high-bit
441 * characters generate +ve values, and ideally we also want to make
442 * the argument match the declared parameter type of "int".
444 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
448 /*********************************************************************
450 * Function : block_acl
452 * Description : Block this request?
453 * Decide yes or no based on ACL file.
456 * 1 : dst = The proxy or gateway address this is going to.
457 * Or NULL to check all possible targets.
458 * 2 : csp = Current client state (buffers, headers, etc...)
459 * Also includes the client IP address.
461 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
463 *********************************************************************/
464 int block_acl(struct access_control_addr *dst, struct client_state *csp)
466 struct access_control_list *acl = csp->config->acl;
468 /* if not using an access control list, then permit the connection */
474 /* search the list */
477 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
481 /* Just want to check if they have any access */
482 if (acl->action == ACL_PERMIT)
487 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
488 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
490 if (acl->action == ACL_PERMIT)
508 /*********************************************************************
510 * Function : acl_addr
512 * Description : Called from `load_config' to parse an ACL address.
515 * 1 : aspec = String specifying ACL address.
516 * 2 : aca = struct access_control_addr to fill in.
518 * Returns : 0 => Ok, everything else is an error.
520 *********************************************************************/
521 int acl_addr(char *aspec, struct access_control_addr *aca)
523 int i, masklength, port;
529 if ((p = strchr(aspec, '/')) != NULL)
533 if (ijb_isdigit(*p) == 0)
537 masklength = atoi(p);
540 if ((masklength < 0) || (masklength > 32))
545 if ((p = strchr(aspec, ':')) != NULL)
549 if (ijb_isdigit(*p) == 0)
558 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
560 if (aca->addr == INADDR_NONE)
565 /* build the netmask */
567 for (i=1; i <= masklength ; i++)
569 aca->mask |= (1 << (32 - i));
572 /* now mask off the host portion of the ip address
573 * (i.e. save on the network portion of the address).
575 aca->addr = aca->addr & aca->mask;
580 #endif /* def FEATURE_ACL */
583 /*********************************************************************
585 * Function : match_portlist
587 * Description : Check if a given number is covered by a comma
588 * separated list of numbers and ranges (a,b-c,d,..)
591 * 1 : portlist = String with list
592 * 2 : port = port to check
594 * Returns : 0 => no match
597 *********************************************************************/
598 int match_portlist(const char *portlist, int port)
600 char *min, *max, *next, *portlist_copy;
602 min = next = portlist_copy = strdup(portlist);
605 * Zero-terminate first item and remember offset for next
607 if (NULL != (next = strchr(portlist_copy, (int) ',')))
613 * Loop through all items, checking for match
617 if (NULL == (max = strchr(min, (int) '-')))
620 * No dash, check for equality
622 if (port == atoi(min))
631 * This is a range, so check if between min and max,
632 * or, if max was omitted, between min and 65K
635 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
649 * Zero-terminate next item and remember offset for n+1
651 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
663 /*********************************************************************
665 * Function : block_url
667 * Description : Called from `chat'. Check to see if we need to block this.
670 * 1 : csp = Current client state (buffers, headers, etc...)
672 * Returns : NULL => unblocked, else HTTP block response
674 *********************************************************************/
675 struct http_response *block_url(struct client_state *csp)
677 #ifdef FEATURE_IMAGE_BLOCKING
679 #endif /* def FEATURE_IMAGE_BLOCKING */
680 struct http_response *rsp;
683 * If it's not blocked, don't block it ;-)
685 if ((csp->action->flags & ACTION_BLOCK) == 0)
691 * Else, prepare a response
693 if (NULL == (rsp = alloc_http_response()))
695 return cgi_error_memory();
699 * If it's an image-url, send back an image or redirect
700 * as specified by the relevant +image action
702 #ifdef FEATURE_IMAGE_BLOCKING
703 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
706 /* determine HOW images should be blocked */
707 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
709 #if 1 /* Two alternative strategies, use this one for now: */
711 /* and handle accordingly: */
712 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
714 rsp->body = bindup(image_pattern_data, image_pattern_length);
715 if (rsp->body == NULL)
717 free_http_response(rsp);
718 return cgi_error_memory();
720 rsp->content_length = image_pattern_length;
722 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
724 free_http_response(rsp);
725 return cgi_error_memory();
729 else if (0 == strcmpic(p, "blank"))
731 rsp->body = bindup(image_blank_data, image_blank_length);
732 if (rsp->body == NULL)
734 free_http_response(rsp);
735 return cgi_error_memory();
737 rsp->content_length = image_blank_length;
739 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
741 free_http_response(rsp);
742 return cgi_error_memory();
748 rsp->status = strdup("302 Local Redirect from Privoxy");
749 if (rsp->status == NULL)
751 free_http_response(rsp);
752 return cgi_error_memory();
755 if (enlist_unique_header(rsp->headers, "Location", p))
757 free_http_response(rsp);
758 return cgi_error_memory();
762 #else /* Following code is disabled for now */
764 /* and handle accordingly: */
765 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
767 p = CGI_PREFIX "send-banner?type=pattern";
769 else if (0 == strcmpic(p, "blank"))
771 p = CGI_PREFIX "send-banner?type=blank";
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();
785 #endif /* Preceeding code is disabled for now */
788 #endif /* def FEATURE_IMAGE_BLOCKING */
791 * Else, generate an HTML "blocked" message:
795 struct map * exports;
798 * Workaround for stupid Netscape bug which prevents
799 * pages from being displayed if loading a referenced
800 * JavaScript or style sheet fails. So make it appear
801 * as if it succeeded.
803 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
804 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
805 && !strstr(p, "Gecko") /* save Mozilla, */
806 && !strstr(p, "compatible") /* MSIE */
807 && !strstr(p, "Opera")) /* and Opera. */
809 rsp->status = strdup("200 Request for blocked URL");
813 rsp->status = strdup("404 Request for blocked URL");
816 if (rsp->status == NULL)
818 free_http_response(rsp);
819 return cgi_error_memory();
822 exports = default_exports(csp, NULL);
825 free_http_response(rsp);
826 return cgi_error_memory();
829 #ifdef FEATURE_FORCE_LOAD
830 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
831 if (csp->http->ssl != 0)
832 #endif /* ndef FEATURE_FORCE_LOAD */
834 err = map_block_killer(exports, "force-support");
837 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
838 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
839 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
844 free_http_response(rsp);
845 return cgi_error_memory();
848 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
851 free_http_response(rsp);
852 return cgi_error_memory();
856 return finish_http_response(rsp);
862 /*********************************************************************
864 * Function : trust_url FIXME: I should be called distrust_url
866 * Description : Calls is_untrusted_url to determine if the URL is trusted
867 * and if not, returns a HTTP 304 response with a reject message.
870 * 1 : csp = Current client state (buffers, headers, etc...)
872 * Returns : NULL => trusted, else http_response.
874 *********************************************************************/
875 struct http_response *trust_url(struct client_state *csp)
877 struct http_response *rsp;
878 struct map * exports;
879 char buf[BUFFER_SIZE];
881 struct url_spec **tl;
886 * Don't bother to work on trusted URLs
888 if (!is_untrusted_url(csp))
894 * Else, prepare a response:
896 if (NULL == (rsp = alloc_http_response()))
898 return cgi_error_memory();
901 exports = default_exports(csp, NULL);
904 free_http_response(rsp);
905 return cgi_error_memory();
909 * Export the protocol, host, port, and referrer information
911 err = map(exports, "hostport", 1, csp->http->hostport, 1);
912 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
913 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
915 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
917 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
921 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
927 free_http_response(rsp);
928 return cgi_error_memory();
932 * Export the trust list
935 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
937 sprintf(buf, "<li>%s</li>\n", t->spec);
938 string_append(&p, buf);
940 err = map(exports, "trusted-referrers", 1, p, 0);
945 free_http_response(rsp);
946 return cgi_error_memory();
950 * Export the trust info, if available
952 if (csp->config->trust_info->first)
954 struct list_entry *l;
957 for (l = csp->config->trust_info->first; l ; l = l->next)
959 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
960 string_append(&p, buf);
962 err = map(exports, "trust-info", 1, p, 0);
966 err = map_block_killer(exports, "have-trust-info");
972 free_http_response(rsp);
973 return cgi_error_memory();
977 * Export the force prefix or the force conditional block killer
979 #ifdef FEATURE_FORCE_LOAD
980 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
981 #else /* ifndef FEATURE_FORCE_LOAD */
982 err = map_block_killer(exports, "force-support");
983 #endif /* ndef FEATURE_FORCE_LOAD */
988 free_http_response(rsp);
989 return cgi_error_memory();
995 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
998 free_http_response(rsp);
999 return cgi_error_memory();
1002 return finish_http_response(rsp);
1004 #endif /* def FEATURE_TRUST */
1007 #ifdef FEATURE_FAST_REDIRECTS
1008 /*********************************************************************
1010 * Function : redirect_url
1012 * Description : Checks for redirection URLs and returns a HTTP redirect
1013 * to the destination URL, if necessary
1016 * 1 : csp = Current client state (buffers, headers, etc...)
1018 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1020 *********************************************************************/
1021 struct http_response *redirect_url(struct client_state *csp)
1024 struct http_response *rsp;
1026 p = q = csp->http->path;
1027 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1030 * find the last URL encoded in the request
1032 while ((p = strstr(p, "http://")) != NULL)
1038 * if there was any, generate and return a HTTP redirect
1040 if (q != csp->http->path)
1042 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1044 if (NULL == (rsp = alloc_http_response()))
1046 return cgi_error_memory();
1049 if ( enlist_unique_header(rsp->headers, "Location", q)
1050 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1052 free_http_response(rsp);
1053 return cgi_error_memory();
1056 return finish_http_response(rsp);
1064 #endif /* def FEATURE_FAST_REDIRECTS */
1067 #ifdef FEATURE_IMAGE_BLOCKING
1068 /*********************************************************************
1070 * Function : is_imageurl
1072 * Description : Given a URL, decide whether it is an image or not,
1073 * using either the info from a previous +image action
1074 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, the info from
1075 * the browser's accept header.
1078 * 1 : csp = Current client state (buffers, headers, etc...)
1080 * Returns : True (nonzero) if URL is an image, false (0)
1083 *********************************************************************/
1084 int is_imageurl(struct client_state *csp)
1086 #ifdef FEATURE_IMAGE_DETECT_MSIE
1089 tmp = get_header_value(csp->headers, "User-Agent:");
1090 if (tmp && strstr(tmp, "MSIE"))
1092 tmp = get_header_value(csp->headers, "Accept:");
1093 if (tmp && strstr(tmp, "image/gif"))
1095 /* Client will accept HTML. If this seems counterintuitive,
1105 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1107 return ((csp->action->flags & ACTION_IMAGE) != 0);
1110 #endif /* def FEATURE_IMAGE_BLOCKING */
1113 #ifdef FEATURE_TRUST
1114 /*********************************************************************
1116 * Function : is_untrusted_url
1118 * Description : Should we "distrust" this URL (and block it)?
1120 * Yes if it matches a line in the trustfile, or if the
1121 * referrer matches a line starting with "+" in the
1126 * 1 : csp = Current client state (buffers, headers, etc...)
1128 * Returns : 0 => trusted, 1 => untrusted
1130 *********************************************************************/
1131 int is_untrusted_url(struct client_state *csp)
1133 struct file_list *fl;
1134 struct block_spec *b;
1135 struct url_spec **trusted_url;
1136 struct http_request rhttp[1];
1137 const char * referer;
1141 * If we don't have a trustlist, we trust everybody
1143 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1148 memset(rhttp, '\0', sizeof(*rhttp));
1151 * Do we trust the request URL itself?
1153 for (b = b->next; b ; b = b->next)
1155 if (url_match(b->url, csp->http))
1161 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1163 /* no referrer was supplied */
1169 * If not, do we maybe trust its referrer?
1171 err = parse_http_url(referer, rhttp, csp);
1177 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1179 if (url_match(*trusted_url, rhttp))
1181 /* if the URL's referrer is from a trusted referrer, then
1182 * add the target spec to the trustfile as an unblocked
1183 * domain and return NULL (which means it's OK).
1188 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1192 char * new_entry = strdup("~");
1194 string_append(&new_entry, csp->http->hostport);
1196 path = csp->http->path;
1197 if ( (path[0] == '/')
1199 && ((path_end = strchr(path + 2, '/')) != NULL))
1201 /* since this path points into a user's home space
1202 * be sure to include this spec in the trustfile.
1204 int path_len = path_end - path; /* save offset */
1205 path = strdup(path); /* Copy string */
1208 path_end = path + path_len; /* regenerate ptr to new buffer */
1209 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1211 string_join(&new_entry, path);
1214 if (new_entry != NULL)
1216 fprintf(fp, "%s\n", new_entry);
1221 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1222 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1232 #endif /* def FEATURE_TRUST */
1235 /*********************************************************************
1237 * Function : pcrs_filter_response
1239 * Description : Ecexute all text substitutions from all applying
1240 * +filter actions on the text buffer that's been accumulated
1241 * in csp->iob->buf. If this changes the contents, set
1242 * csp->content_length to the modified size and raise the
1243 * CSP_FLAG_MODIFIED flag.
1246 * 1 : csp = Current client state (buffers, headers, etc...)
1248 * Returns : a pointer to the (newly allocated) modified buffer.
1249 * or NULL if there were no hits or something went wrong
1251 *********************************************************************/
1252 char *pcrs_filter_response(struct client_state *csp)
1257 char *old = csp->iob->cur, *new = NULL;
1260 struct file_list *fl;
1261 struct re_filterfile_spec *b;
1262 struct list_entry *filtername;
1267 if (csp->iob->cur >= csp->iob->eod)
1271 size = csp->iob->eod - csp->iob->cur;
1273 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1275 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1280 * If the body has a "chunked" transfer-encoding,
1281 * get rid of it first, adjusting size and iob->eod
1283 if (csp->flags & CSP_FLAG_CHUNKED)
1285 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1286 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1290 csp->iob->eod = csp->iob->cur + size;
1291 csp->flags |= CSP_FLAG_MODIFIED;
1295 * For all applying +filter actions, look if a filter by that
1296 * name exists and if yes, execute it's pcrs_joblist on the
1299 for (b = fl->f; b; b = b->next)
1301 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1302 filtername ; filtername = filtername->next)
1304 if (strcmp(b->name, filtername->str) == 0)
1306 int current_hits = 0;
1308 if ( NULL == b->joblist )
1310 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1314 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1315 csp->http->hostport, csp->http->path, size, b->name);
1317 /* Apply all jobs from the joblist */
1318 for (job = b->joblist; NULL != job; job = job->next)
1320 current_hits += pcrs_execute(job, old, size, &new, &size);
1321 if (old != csp->iob->cur) free(old);
1325 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1326 hits += current_hits;
1332 * If there were no hits, destroy our copy and let
1333 * chat() use the original in csp->iob
1341 csp->flags |= CSP_FLAG_MODIFIED;
1342 csp->content_length = size;
1350 /*********************************************************************
1352 * Function : gif_deanimate_response
1354 * Description : Deanimate the GIF image that has been accumulated in
1355 * csp->iob->buf, set csp->content_length to the modified
1356 * size and raise the CSP_FLAG_MODIFIED flag.
1359 * 1 : csp = Current client state (buffers, headers, etc...)
1361 * Returns : a pointer to the (newly allocated) modified buffer.
1362 * or NULL in case something went wrong.
1364 *********************************************************************/
1365 char *gif_deanimate_response(struct client_state *csp)
1367 struct binbuffer *in, *out;
1369 size_t size = csp->iob->eod - csp->iob->cur;
1372 * If the body has a "chunked" transfer-encoding,
1373 * get rid of it first, adjusting size and iob->eod
1375 if (csp->flags & CSP_FLAG_CHUNKED)
1377 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1378 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1382 csp->iob->eod = csp->iob->cur + size;
1383 csp->flags |= CSP_FLAG_MODIFIED;
1386 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1387 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1389 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1393 in->buffer = csp->iob->cur;
1396 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1398 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1405 if ((int)size == out->offset)
1407 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1411 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1413 csp->content_length = out->offset;
1414 csp->flags |= CSP_FLAG_MODIFIED;
1424 /*********************************************************************
1426 * Function : remove_chunked_transfer_coding
1428 * Description : In-situ remove the "chunked" transfer coding as defined
1429 * in rfc2616 from a buffer.
1432 * 1 : buffer = Pointer to the text buffer
1433 * 2 : size = Number of bytes to be processed
1435 * Returns : The new size, i.e. the number of bytes from buffer which
1436 * are occupied by the stripped body, or 0 in case something
1439 *********************************************************************/
1440 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1443 unsigned int chunksize = 0;
1444 char *from_p, *to_p;
1447 from_p = to_p = buffer;
1449 if (sscanf(buffer, "%x", &chunksize) != 1)
1451 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1455 while (chunksize > 0)
1457 if (NULL == (from_p = strstr(from_p, "\r\n")))
1459 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1462 newsize += chunksize;
1465 memmove(to_p, from_p, (size_t) chunksize);
1466 to_p = buffer + newsize;
1467 from_p += chunksize + 2;
1469 if (sscanf(from_p, "%x", &chunksize) != 1)
1471 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1476 /* FIXME: Should this get its own loglevel? */
1477 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1483 /*********************************************************************
1485 * Function : url_actions
1487 * Description : Gets the actions for this URL.
1490 * 1 : http = http_request request for blocked URLs
1491 * 2 : csp = Current client state (buffers, headers, etc...)
1495 *********************************************************************/
1496 void url_actions(struct http_request *http,
1497 struct client_state *csp)
1499 struct file_list *fl;
1500 struct url_actions *b;
1502 init_current_action(csp->action);
1504 if (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL))
1509 apply_url_actions(csp->action, http, b);
1514 /*********************************************************************
1516 * Function : apply_url_actions
1518 * Description : Applies a list of URL actions.
1521 * 1 : action = Destination.
1522 * 2 : http = Current URL
1523 * 3 : b = list of URL actions to apply
1527 *********************************************************************/
1528 void apply_url_actions(struct current_action_spec *action,
1529 struct http_request *http,
1530 struct url_actions *b)
1534 /* Should never happen */
1538 for (b = b->next; NULL != b; b = b->next)
1540 if (url_match(b->url, http))
1542 merge_current_action(action, b->action);
1548 /*********************************************************************
1550 * Function : forward_url
1552 * Description : Should we forward this to another proxy?
1555 * 1 : http = http_request request for current URL
1556 * 2 : csp = Current client state (buffers, headers, etc...)
1558 * Returns : Pointer to forwarding information.
1560 *********************************************************************/
1561 const struct forward_spec * forward_url(struct http_request *http,
1562 struct client_state *csp)
1564 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1565 struct forward_spec *fwd = csp->config->forward;
1574 if (url_match(fwd->url, http))