1 const char filters_rcs[] = "$Id: filters.c,v 1.48 2002/03/13 20:25:34 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 * IJBSWA team. http://ijbswa.sourceforge.net
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 1.48 2002/03/13 20:25:34 oes
42 * Better logging for content filters
44 * Revision 1.47 2002/03/13 00:30:52 jongfoster
46 * Added option of always sending redirect for imageblock,
47 * currently disabled with #if 0.
49 * Revision 1.46 2002/03/12 01:42:49 oes
50 * Introduced modular filters
52 * Revision 1.45 2002/03/08 16:47:50 oes
53 * Added choice beween GIF and PNG built-in images
55 * Revision 1.44 2002/03/07 03:49:31 oes
56 * - Fixed compiler warnings etc
57 * - Changed built-in images from GIF to PNG
58 * (with regard to Unisys patent issue)
59 * - Added a 4x4 pattern PNG which is less intrusive
60 * than the logo but also clearly marks the deleted banners
62 * Revision 1.43 2002/01/22 23:51:59 jongfoster
63 * Replacing strsav() with the safer string_append().
65 * Adding missing html_encode() to error message generators. Where encoded
66 * and unencoded versions of a string were provided, removing the unencoded
69 * Revision 1.42 2002/01/17 21:00:32 jongfoster
70 * Moving all our URL and URL pattern parsing code to urlmatch.c.
72 * Using a single, simple url_match(pattern,url) function - rather than
73 * the 3-line match routine which was repeated all over the place.
75 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
77 * Using parse_http_url() to parse URLs without faking a HTTP
78 * request line for parse_http_request().
80 * Revision 1.41 2001/11/13 00:14:07 jongfoster
81 * Fixing stupid bug now I've figured out what || means.
82 * (It always returns 0 or 1, not one of it's paramaters.)
84 * Revision 1.40 2001/10/26 17:37:55 oes
85 * - Re-enabled Netscape 200/404 bug workaround in block_url():
86 * - Removed OS/2 special case
87 * - Made block_url() independant from sed() having been run
88 * - Made trust_url independant from sed() having been run
89 * - Made is_imageurl independant from sed() having been run.
90 * It now checks User-Agent: and Accept: by itself.
93 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
94 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
95 * threads to call select() simultaneously. So, it's time to do a real, live,
96 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
97 * (native). Both versions will work, but using __OS2__ offers multi-threading.
99 * Revision 1.38 2001/10/23 21:32:33 jongfoster
100 * Adding error-checking to selected functions
102 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
103 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
104 * filters.c. Added a FIXME in front of the offending code. I'll gladly
105 * put in a better/more robust fix for all parties if one is presented...
106 * It seems that just returning 200 instead of 404 would pretty much fix
107 * it for everyone, but I don't know all the history of the problem.
109 * Revision 1.36 2001/10/10 16:44:16 oes
110 * Added match_portlist function
112 * Revision 1.35 2001/10/07 15:41:23 oes
113 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
115 * New function remove_chunked_transfer_coding that strips chunked
116 * transfer coding to plain and is called by pcrs_filter_response
117 * and gif_deanimate_response if neccessary
119 * Improved handling of zero-change re_filter runs
121 * pcrs_filter_response and gif_deanimate_response now remove
122 * chunked transfer codeing before processing the body.
124 * Revision 1.34 2001/09/20 15:49:36 steudten
126 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
127 * See cgi.c fill_template().
129 * Revision 1.33 2001/09/16 17:05:14 jongfoster
130 * Removing unused #include showarg.h
132 * Revision 1.32 2001/09/16 13:21:27 jongfoster
133 * Changes to use new list functions.
135 * Revision 1.31 2001/09/16 11:38:02 jongfoster
136 * Splitting fill_template() into 2 functions:
137 * template_load() loads the file
138 * template_fill() performs the PCRS regexps.
139 * This is because the CGI edit interface has a "table row"
140 * template which is used many times in the page - this
141 * change means it's only loaded from disk once.
143 * Revision 1.30 2001/09/16 11:00:10 jongfoster
144 * New function alloc_http_response, for symmetry with free_http_response
146 * Revision 1.29 2001/09/13 23:32:40 jongfoster
147 * Moving image data to cgi.c rather than cgi.h
148 * Fixing a GPF under Win32 (and any other OS that protects global
149 * constants from being written to).
151 * Revision 1.28 2001/09/10 10:18:51 oes
152 * Silenced compiler warnings
154 * Revision 1.27 2001/08/05 16:06:20 jongfoster
155 * Modifiying "struct map" so that there are now separate header and
156 * "map_entry" structures. This means that functions which modify a
157 * map no longer need to return a pointer to the modified map.
158 * Also, it no longer reverses the order of the entries (which may be
159 * important with some advanced template substitutions).
161 * Revision 1.26 2001/07/30 22:08:36 jongfoster
162 * Tidying up #defines:
163 * - All feature #defines are now of the form FEATURE_xxx
164 * - Permanently turned off WIN_GUI_EDIT
165 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
167 * Revision 1.25 2001/07/26 10:09:46 oes
168 * Made browser detection a little less naive
170 * Revision 1.24 2001/07/25 17:22:51 oes
171 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
173 * Revision 1.23 2001/07/23 13:40:12 oes
174 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
176 * Revision 1.22 2001/07/18 12:29:34 oes
177 * - Made gif_deanimate_response respect
178 * csp->action->string[ACTION_STRING_DEANIMATE]
179 * - Logging cosmetics
181 * Revision 1.21 2001/07/13 13:59:53 oes
182 * - Introduced gif_deanimate_response which shares the
183 * generic content modification interface of pcrs_filter_response
184 * and acts as a wrapper to deanimate.c:gif_deanimate()
185 * - Renamed re_process_buffer to pcrs_filter_response
186 * - pcrs_filter_response now returns NULL on failiure
187 * - Removed all #ifdef PCRS
189 * Revision 1.20 2001/07/01 17:01:04 oes
190 * Added comments and missing return statement in is_untrusted_url()
192 * Revision 1.19 2001/06/29 21:45:41 oes
193 * Indentation, CRLF->LF, Tab-> Space
195 * Revision 1.18 2001/06/29 13:27:38 oes
196 * - Cleaned up, renamed and reorderd functions
197 * and improved comments
200 * - Ported to CGI platform. Now delivers
201 * http_response or NULL
202 * - Unified HTML and GIF generation (moved image detection
203 * and GIF generation here from jcc.c:chat())
204 * - Fixed HTTP status to:
205 * - 403 (Forbidden) for the "blocked" HTML message
206 * - 200 (OK) for GIF answers
207 * - 302 (Redirect) for redirect to GIF
210 * - Ported to CGI platform. Now delivers
211 * http_response or NULL
212 * - Separated detection of untrusted URL into
213 * (bool)is_untrusted_url
214 * - Added enforcement of untrusted requests
216 * - Moved redirect_url() from cgi.c to here
217 * and ported it to the CGI platform
219 * - Removed logentry from cancelled commit
221 * Revision 1.17 2001/06/09 10:55:28 jongfoster
222 * Changing BUFSIZ ==> BUFFER_SIZE
224 * Revision 1.16 2001/06/07 23:10:26 jongfoster
225 * Allowing unanchored domain patterns to back off and retry
226 * if they partially match. Optimized right-anchored patterns.
227 * Moving ACL and forward files into config file.
228 * Replacing struct gateway with struct forward_spec
230 * Revision 1.15 2001/06/03 19:12:00 oes
231 * extracted-CGI relevant stuff
233 * Revision 1.14 2001/06/01 10:30:55 oes
234 * Added optional left-anchoring to domaincmp
236 * Revision 1.13 2001/05/31 21:21:30 jongfoster
237 * Permissionsfile / actions file changes:
238 * - Changed "permission" to "action" throughout
239 * - changes to file format to allow string parameters
240 * - Moved helper functions to actions.c
242 * Revision 1.12 2001/05/31 17:35:20 oes
244 * - Enhanced domain part globbing with infix and prefix asterisk
245 * matching and optional unanchored operation
247 * Revision 1.11 2001/05/29 11:53:23 oes
248 * "See why" link added to "blocked" page
250 * Revision 1.10 2001/05/29 09:50:24 jongfoster
251 * Unified blocklist/imagelist/permissionslist.
252 * File format is still under discussion, but the internal changes
255 * Also modified interceptor behaviour:
256 * - We now intercept all URLs beginning with one of the following
257 * prefixes (and *only* these prefixes):
259 * * http://ijbswa.sf.net/config/
260 * * http://ijbswa.sourceforge.net/config/
261 * - New interceptors "home page" - go to http://i.j.b/ to see it.
262 * - Internal changes so that intercepted and fast redirect pages
263 * are not replaced with an image.
264 * - Interceptors now have the option to send a binary page direct
265 * to the client. (i.e. ijb-send-banner uses this)
266 * - Implemented show-url-info interceptor. (Which is why I needed
267 * the above interceptors changes - a typical URL is
268 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
269 * The previous mechanism would not have intercepted that, and
270 * if it had been intercepted then it then it would have replaced
273 * Revision 1.9 2001/05/27 22:17:04 oes
275 * - re_process_buffer no longer writes the modified buffer
276 * to the client, which was very ugly. It now returns the
277 * buffer, which it is then written by chat.
279 * - content_length now adjusts the Content-Length: header
280 * for modified documents rather than crunch()ing it.
281 * (Length info in csp->content_length, which is 0 for
282 * unmodified documents)
284 * - For this to work, sed() is called twice when filtering.
286 * Revision 1.8 2001/05/26 17:13:28 jongfoster
287 * Filled in a function comment.
289 * Revision 1.7 2001/05/26 15:26:15 jongfoster
290 * ACL feature now provides more security by immediately dropping
291 * connections from untrusted hosts.
293 * Revision 1.6 2001/05/26 00:28:36 jongfoster
294 * Automatic reloading of config file.
295 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
296 * Most of the global variables have been moved to a new
297 * struct configuration_spec, accessed through csp->config->globalname
298 * Most of the globals remaining are used by the Win32 GUI.
300 * Revision 1.5 2001/05/25 22:34:30 jongfoster
303 * Revision 1.4 2001/05/22 18:46:04 oes
305 * - Enabled filtering banners by size rather than URL
306 * by adding patterns that replace all standard banner
307 * sizes with the "Junkbuster" gif to the re_filterfile
309 * - Enabled filtering WebBugs by providing a pattern
310 * which kills all 1x1 images
312 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
313 * which is selected by the (nonstandard and therefore
314 * capital) letter 'U' in the option string.
315 * It causes the quantifiers to be ungreedy by default.
316 * Appending a ? turns back to greedy (!).
318 * - Added a new interceptor ijb-send-banner, which
319 * sends back the "Junkbuster" gif. Without imagelist or
320 * MSIE detection support, or if tinygif = 1, or the
321 * URL isn't recognized as an imageurl, a lame HTML
322 * explanation is sent instead.
324 * - Added new feature, which permits blocking remote
325 * script redirects and firing back a local redirect
327 * The feature is conditionally compiled, i.e. it
328 * can be disabled with --disable-fast-redirects,
329 * plus it must be activated by a "fast-redirects"
330 * line in the config file, has its own log level
331 * and of course wants to be displayed by show-proxy-args
332 * Note: Boy, all the #ifdefs in 1001 locations and
333 * all the fumbling with configure.in and acconfig.h
334 * were *way* more work than the feature itself :-(
336 * - Because a generic redirect template was needed for
337 * this, tinygif = 3 now uses the same.
339 * - Moved GIFs, and other static HTTP response templates
344 * - Removed some >400 CRs again (Jon, you really worked
347 * Revision 1.3 2001/05/20 16:44:47 jongfoster
348 * Removing last hardcoded JunkBusters.com URLs.
350 * Revision 1.2 2001/05/20 01:21:20 jongfoster
351 * Version 2.9.4 checkin.
352 * - Merged popupfile and cookiefile, and added control over PCRS
353 * filtering, in new "permissionsfile".
354 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
355 * file error you now get a message box (in the Win32 GUI) rather
356 * than the program exiting with no explanation.
357 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
359 * - Removed tabs from "config"
360 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
361 * - Bumped up version number.
363 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
364 * Initial import of version 2.9.3 source tree
367 *********************************************************************/
373 #include <sys/types.h>
382 #endif /* ndef __OS2__ */
383 #include <netinet/in.h>
385 #include <winsock2.h>
386 #endif /* ndef _WIN32 */
390 #endif /* def __OS2__ */
398 #include "jbsockets.h"
399 #include "miscutil.h"
403 #include "deanimate.h"
404 #include "urlmatch.h"
410 const char filters_h_rcs[] = FILTERS_H_VERSION;
412 /* Fix a problem with Solaris. There should be no effect on other
414 * Solaris's isspace() is a macro which uses it's argument directly
415 * as an array index. Therefore we need to make sure that high-bit
416 * characters generate +ve values, and ideally we also want to make
417 * the argument match the declared parameter type of "int".
419 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
423 /*********************************************************************
425 * Function : block_acl
427 * Description : Block this request?
428 * Decide yes or no based on ACL file.
431 * 1 : dst = The proxy or gateway address this is going to.
432 * Or NULL to check all possible targets.
433 * 2 : csp = Current client state (buffers, headers, etc...)
434 * Also includes the client IP address.
436 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
438 *********************************************************************/
439 int block_acl(struct access_control_addr *dst, struct client_state *csp)
441 struct access_control_list *acl = csp->config->acl;
443 /* if not using an access control list, then permit the connection */
449 /* search the list */
452 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
456 /* Just want to check if they have any access */
457 if (acl->action == ACL_PERMIT)
462 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
463 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
465 if (acl->action == ACL_PERMIT)
483 /*********************************************************************
485 * Function : acl_addr
487 * Description : Called from `load_config' to parse an ACL address.
490 * 1 : aspec = String specifying ACL address.
491 * 2 : aca = struct access_control_addr to fill in.
493 * Returns : 0 => Ok, everything else is an error.
495 *********************************************************************/
496 int acl_addr(char *aspec, struct access_control_addr *aca)
498 int i, masklength, port;
504 if ((p = strchr(aspec, '/')) != NULL)
508 if (ijb_isdigit(*p) == 0)
512 masklength = atoi(p);
515 if ((masklength < 0) || (masklength > 32))
520 if ((p = strchr(aspec, ':')) != NULL)
524 if (ijb_isdigit(*p) == 0)
533 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
535 if (aca->addr == INADDR_NONE)
540 /* build the netmask */
542 for (i=1; i <= masklength ; i++)
544 aca->mask |= (1 << (32 - i));
547 /* now mask off the host portion of the ip address
548 * (i.e. save on the network portion of the address).
550 aca->addr = aca->addr & aca->mask;
555 #endif /* def FEATURE_ACL */
558 /*********************************************************************
560 * Function : match_portlist
562 * Description : Check if a given number is covered by a comma
563 * separated list of numbers and ranges (a,b-c,d,..)
566 * 1 : portlist = String with list
567 * 2 : port = port to check
569 * Returns : 0 => no match
572 *********************************************************************/
573 int match_portlist(const char *portlist, int port)
575 char *min, *max, *next, *portlist_copy;
577 min = next = portlist_copy = strdup(portlist);
580 * Zero-terminate first item and remember offset for next
582 if (NULL != (next = strchr(portlist_copy, (int) ',')))
588 * Loop through all items, checking for match
592 if (NULL == (max = strchr(min, (int) '-')))
595 * No dash, check for equality
597 if (port == atoi(min))
606 * This is a range, so check if between min and max,
607 * or, if max was omitted, between min and 65K
610 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
624 * Zero-terminate next item and remember offset for n+1
626 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
638 /*********************************************************************
640 * Function : block_url
642 * Description : Called from `chat'. Check to see if we need to block this.
645 * 1 : csp = Current client state (buffers, headers, etc...)
647 * Returns : NULL => unblocked, else HTTP block response
649 *********************************************************************/
650 struct http_response *block_url(struct client_state *csp)
652 #ifdef FEATURE_IMAGE_BLOCKING
654 #endif /* def FEATURE_IMAGE_BLOCKING */
655 struct http_response *rsp;
658 * If it's not blocked, don't block it ;-)
660 if ((csp->action->flags & ACTION_BLOCK) == 0)
666 * Else, prepare a response
668 if (NULL == (rsp = alloc_http_response()))
670 return cgi_error_memory();
674 * If it's an image-url, send back an image or redirect
675 * as specified by the relevant +image action
677 #ifdef FEATURE_IMAGE_BLOCKING
678 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
681 /* determine HOW images should be blocked */
682 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
684 #if 1 /* Two alternative strategies, use this one for now: */
686 /* and handle accordingly: */
687 if ((p == NULL) || (0 == strcmpic(p, "logo")))
689 rsp->body = bindup(image_logo_data, image_logo_length);
690 if (rsp->body == NULL)
692 free_http_response(rsp);
693 return cgi_error_memory();
695 rsp->content_length = image_logo_length;
697 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
699 free_http_response(rsp);
700 return cgi_error_memory();
704 else if (0 == strcmpic(p, "blank"))
706 rsp->body = bindup(image_blank_data, image_blank_length);
707 if (rsp->body == NULL)
709 free_http_response(rsp);
710 return cgi_error_memory();
712 rsp->content_length = image_blank_length;
714 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
716 free_http_response(rsp);
717 return cgi_error_memory();
721 else if (0 == strcmpic(p, "pattern"))
723 rsp->body = bindup(image_pattern_data, image_pattern_length);
724 if (rsp->body == NULL)
726 free_http_response(rsp);
727 return cgi_error_memory();
729 rsp->content_length = image_pattern_length;
731 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
733 free_http_response(rsp);
734 return cgi_error_memory();
740 rsp->status = strdup("302 Local Redirect from Junkbuster");
741 if (rsp->status == NULL)
743 free_http_response(rsp);
744 return cgi_error_memory();
747 if (enlist_unique_header(rsp->headers, "Location", p))
749 free_http_response(rsp);
750 return cgi_error_memory();
754 #else /* Following code is disabled for now */
756 /* and handle accordingly: */
757 if ((p == NULL) || (0 == strcmpic(p, "logo")))
759 p = CGI_PREFIX "send-banner?type=logo";
761 else if (0 == strcmpic(p, "blank"))
763 p = CGI_PREFIX "send-banner?type=blank";
765 else if (0 == strcmpic(p, "pattern"))
767 p = CGI_PREFIX "send-banner?type=pattern";
769 rsp->status = strdup("302 Local Redirect from Junkbuster");
770 if (rsp->status == NULL)
772 free_http_response(rsp);
773 return cgi_error_memory();
776 if (enlist_unique_header(rsp->headers, "Location", p))
778 free_http_response(rsp);
779 return cgi_error_memory();
781 #endif /* Preceeding code is disabled for now */
784 #endif /* def FEATURE_IMAGE_BLOCKING */
787 * Else, generate an HTML "blocked" message:
791 struct map * exports;
794 * Workaround for stupid Netscape bug which prevents
795 * pages from being displayed if loading a referenced
796 * JavaScript or style sheet fails. So make it appear
797 * as if it succeeded.
799 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
800 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
801 && !strstr(p, "Gecko") /* save Mozilla, */
802 && !strstr(p, "compatible") /* MSIE */
803 && !strstr(p, "Opera")) /* and Opera. */
805 rsp->status = strdup("200 Request for blocked URL");
809 rsp->status = strdup("404 Request for blocked URL");
812 if (rsp->status == NULL)
814 free_http_response(rsp);
815 return cgi_error_memory();
818 exports = default_exports(csp, NULL);
821 free_http_response(rsp);
822 return cgi_error_memory();
825 #ifdef FEATURE_FORCE_LOAD
826 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
827 #else /* ifndef FEATURE_FORCE_LOAD */
828 err = map_block_killer(exports, "force-support");
829 #endif /* ndef FEATURE_FORCE_LOAD */
831 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
832 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
837 free_http_response(rsp);
838 return cgi_error_memory();
841 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
844 free_http_response(rsp);
845 return cgi_error_memory();
849 return finish_http_response(rsp);
855 /*********************************************************************
857 * Function : trust_url FIXME: I should be called distrust_url
859 * Description : Calls is_untrusted_url to determine if the URL is trusted
860 * and if not, returns a HTTP 304 response with a reject message.
863 * 1 : csp = Current client state (buffers, headers, etc...)
865 * Returns : NULL => trusted, else http_response.
867 *********************************************************************/
868 struct http_response *trust_url(struct client_state *csp)
870 struct http_response *rsp;
871 struct map * exports;
872 char buf[BUFFER_SIZE];
874 struct url_spec **tl;
879 * Don't bother to work on trusted URLs
881 if (!is_untrusted_url(csp))
887 * Else, prepare a response:
889 if (NULL == (rsp = alloc_http_response()))
891 return cgi_error_memory();
894 exports = default_exports(csp, NULL);
897 free_http_response(rsp);
898 return cgi_error_memory();
902 * Export the host, port, and referrer information
904 err = map(exports, "hostport", 1, csp->http->hostport, 1);
905 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
907 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
909 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
913 if (!err) err = map(exports, "referrer", 1, "unknown", 1);
919 free_http_response(rsp);
920 return cgi_error_memory();
924 * Export the trust list
927 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
929 sprintf(buf, "<li>%s</li>\n", t->spec);
930 string_append(&p, buf);
932 err = map(exports, "trusted-referrers", 1, p, 0);
937 free_http_response(rsp);
938 return cgi_error_memory();
942 * Export the trust info, if available
944 if (csp->config->trust_info->first)
946 struct list_entry *l;
949 for (l = csp->config->trust_info->first; l ; l = l->next)
951 sprintf(buf, "<li> <a href=%s>%s</a><br>\n",l->str, l->str);
952 string_append(&p, buf);
954 err = map(exports, "trust-info", 1, p, 0);
958 err = map_block_killer(exports, "have-trust-info");
964 free_http_response(rsp);
965 return cgi_error_memory();
969 * Export the force prefix or the force conditional block killer
971 #ifdef FEATURE_FORCE_LOAD
972 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
973 #else /* ifndef FEATURE_FORCE_LOAD */
974 err = map_block_killer(exports, "force-support");
975 #endif /* ndef FEATURE_FORCE_LOAD */
980 free_http_response(rsp);
981 return cgi_error_memory();
987 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
990 free_http_response(rsp);
991 return cgi_error_memory();
994 return finish_http_response(rsp);
996 #endif /* def FEATURE_TRUST */
999 #ifdef FEATURE_FAST_REDIRECTS
1000 /*********************************************************************
1002 * Function : redirect_url
1004 * Description : Checks for redirection URLs and returns a HTTP redirect
1005 * to the destination URL, if necessary
1008 * 1 : csp = Current client state (buffers, headers, etc...)
1010 * Returns : NULL if URL was clean, HTTP redirect otherwise.
1012 *********************************************************************/
1013 struct http_response *redirect_url(struct client_state *csp)
1016 struct http_response *rsp;
1018 p = q = csp->http->path;
1019 log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
1022 * find the last URL encoded in the request
1024 while ((p = strstr(p, "http://")) != NULL)
1030 * if there was any, generate and return a HTTP redirect
1032 if (q != csp->http->path)
1034 log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
1036 if (NULL == (rsp = alloc_http_response()))
1038 return cgi_error_memory();
1041 if ( enlist_unique_header(rsp->headers, "Location", q)
1042 || (NULL == (rsp->status = strdup("302 Local Redirect from Junkbuster"))) )
1044 free_http_response(rsp);
1045 return cgi_error_memory();
1048 return finish_http_response(rsp);
1056 #endif /* def FEATURE_FAST_REDIRECTS */
1059 #ifdef FEATURE_IMAGE_BLOCKING
1060 /*********************************************************************
1062 * Function : is_imageurl
1064 * Description : Given a URL, decide whether it is an image or not,
1065 * using either the info from a previous +image action
1066 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, the info from
1067 * the browser's accept header.
1070 * 1 : csp = Current client state (buffers, headers, etc...)
1072 * Returns : True (nonzero) if URL is an image, false (0)
1075 *********************************************************************/
1076 int is_imageurl(struct client_state *csp)
1078 #ifdef FEATURE_IMAGE_DETECT_MSIE
1081 tmp = get_header_value(csp->headers, "User-Agent:");
1082 if (tmp && strstr(tmp, "MSIE"))
1084 tmp = get_header_value(csp->headers, "Accept:");
1085 if (tmp && strstr(tmp, "image/gif"))
1087 /* Client will accept HTML. If this seems counterintuitive,
1097 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1099 return ((csp->action->flags & ACTION_IMAGE) != 0);
1102 #endif /* def FEATURE_IMAGE_BLOCKING */
1105 #ifdef FEATURE_COOKIE_JAR
1106 /*********************************************************************
1108 * Function : is_untrusted_url
1110 * Description : Should we "distrust" this URL (and block it)?
1112 * Yes if it matches a line in the trustfile, or if the
1113 * referrer matches a line starting with "+" in the
1118 * 1 : csp = Current client state (buffers, headers, etc...)
1120 * Returns : 0 => trusted, 1 => untrusted
1122 *********************************************************************/
1123 int is_untrusted_url(struct client_state *csp)
1125 struct file_list *fl;
1126 struct block_spec *b;
1127 struct url_spec **trusted_url;
1128 struct http_request rhttp[1];
1129 const char * referer;
1133 * If we don't have a trustlist, we trust everybody
1135 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1140 memset(rhttp, '\0', sizeof(*rhttp));
1143 * Do we trust the request URL itself?
1145 for (b = b->next; b ; b = b->next)
1147 if (url_match(b->url, csp->http))
1153 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1155 /* no referrer was supplied */
1160 * If not, do we maybe trust its referrer?
1165 * Parse the URL from the referrer
1168 err = parse_http_url(referer, rhttp, csp);
1174 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1176 if (url_match(*trusted_url, rhttp))
1178 /* if the URL's referrer is from a trusted referrer, then
1179 * add the target spec to the trustfile as an unblocked
1180 * domain and return NULL (which means it's OK).
1185 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1189 char * new_entry = strdup("~");
1191 string_append(&new_entry, csp->http->hostport);
1193 path = csp->http->path;
1194 if ( (path[0] == '/')
1196 && ((path_end = strchr(path + 2, '/')) != NULL))
1198 /* since this path points into a user's home space
1199 * be sure to include this spec in the trustfile.
1201 int path_len = path_end - path; /* save offset */
1202 path = strdup(path); /* Copy string */
1205 path_end = path + path_len; /* regenerate ptr to new buffer */
1206 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1208 string_join(&new_entry, path);
1211 if (new_entry != NULL)
1213 fprintf(fp, "%s\n", new_entry);
1218 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1219 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1229 #endif /* def FEATURE_COOKIE_JAR */
1232 /*********************************************************************
1234 * Function : pcrs_filter_response
1236 * Description : Ecexute all text substitutions from all applying
1237 * +filter actions on the text buffer that's been accumulated
1238 * in csp->iob->buf. If this changes the contents, set
1239 * csp->content_length to the modified size and raise the
1240 * CSP_FLAG_MODIFIED flag.
1243 * 1 : csp = Current client state (buffers, headers, etc...)
1245 * Returns : a pointer to the (newly allocated) modified buffer.
1246 * or NULL if there were no hits or something went wrong
1248 *********************************************************************/
1249 char *pcrs_filter_response(struct client_state *csp)
1254 char *old = csp->iob->cur, *new = NULL;
1257 struct file_list *fl;
1258 struct re_filterfile_spec *b;
1259 struct list_entry *filtername;
1264 if (csp->iob->cur >= csp->iob->eod)
1268 size = csp->iob->eod - csp->iob->cur;
1270 if ( ( NULL == (fl = csp->rlist) ) || ( NULL == fl->f) )
1272 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1277 * If the body has a "chunked" transfer-encoding,
1278 * get rid of it first, adjusting size and iob->eod
1280 if (csp->flags & CSP_FLAG_CHUNKED)
1282 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1283 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1287 csp->iob->eod = csp->iob->cur + size;
1288 csp->flags |= CSP_FLAG_MODIFIED;
1292 * For all applying +filter actions, look if a filter by that
1293 * name exists and if yes, execute it's pcrs_joblist on the
1296 for (b = fl->f; b; b = b->next)
1298 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1299 filtername ; filtername = filtername->next)
1301 if (strcmp(b->name, filtername->str) == 0)
1303 int current_hits = 0;
1305 if ( NULL == b->joblist )
1307 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1311 log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s...",
1312 csp->http->hostport, csp->http->path, size, b->name);
1314 /* Apply all jobs from the joblist */
1315 for (job = b->joblist; NULL != job; job = job->next)
1317 current_hits += pcrs_execute(job, old, size, &new, &size);
1318 if (old != csp->iob->cur) free(old);
1322 log_error(LOG_LEVEL_RE_FILTER, " ...produced %d hits (new size %d).", current_hits, size);
1323 hits += current_hits;
1329 * If there were no hits, destroy our copy and let
1330 * chat() use the original in csp->iob
1338 csp->flags |= CSP_FLAG_MODIFIED;
1339 csp->content_length = size;
1347 /*********************************************************************
1349 * Function : gif_deanimate_response
1351 * Description : Deanimate the GIF image that has been accumulated in
1352 * csp->iob->buf, set csp->content_length to the modified
1353 * size and raise the CSP_FLAG_MODIFIED flag.
1356 * 1 : csp = Current client state (buffers, headers, etc...)
1358 * Returns : a pointer to the (newly allocated) modified buffer.
1359 * or NULL in case something went wrong.
1361 *********************************************************************/
1362 char *gif_deanimate_response(struct client_state *csp)
1364 struct binbuffer *in, *out;
1366 size_t size = csp->iob->eod - csp->iob->cur;
1369 * If the body has a "chunked" transfer-encoding,
1370 * get rid of it first, adjusting size and iob->eod
1372 if (csp->flags & CSP_FLAG_CHUNKED)
1374 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1375 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1379 csp->iob->eod = csp->iob->cur + size;
1380 csp->flags |= CSP_FLAG_MODIFIED;
1383 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1384 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1386 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1390 in->buffer = csp->iob->cur;
1393 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1395 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1402 if ((int)size == out->offset)
1404 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1408 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1410 csp->content_length = out->offset;
1411 csp->flags |= CSP_FLAG_MODIFIED;
1421 /*********************************************************************
1423 * Function : remove_chunked_transfer_coding
1425 * Description : In-situ remove the "chunked" transfer coding as defined
1426 * in rfc2616 from a buffer.
1429 * 1 : buffer = Pointer to the text buffer
1430 * 2 : size = Number of bytes to be processed
1432 * Returns : The new size, i.e. the number of bytes from buffer which
1433 * are occupied by the stripped body, or 0 in case something
1436 *********************************************************************/
1437 int remove_chunked_transfer_coding(char *buffer, const size_t size)
1440 unsigned int chunksize = 0;
1441 char *from_p, *to_p;
1444 from_p = to_p = buffer;
1446 if (sscanf(buffer, "%x", &chunksize) != 1)
1448 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
1452 while (chunksize > 0)
1454 if (NULL == (from_p = strstr(from_p, "\r\n")))
1456 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1459 newsize += chunksize;
1462 memmove(to_p, from_p, (size_t) chunksize);
1463 to_p = buffer + newsize;
1464 from_p += chunksize + 2;
1466 if (sscanf(from_p, "%x", &chunksize) != 1)
1468 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
1473 /* FIXME: Should this get its own loglevel? */
1474 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
1480 /*********************************************************************
1482 * Function : url_actions
1484 * Description : Gets the actions for this URL.
1487 * 1 : http = http_request request for blocked URLs
1488 * 2 : csp = Current client state (buffers, headers, etc...)
1492 *********************************************************************/
1493 void url_actions(struct http_request *http,
1494 struct client_state *csp)
1496 struct file_list *fl;
1497 struct url_actions *b;
1499 init_current_action(csp->action);
1501 if (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL))
1506 apply_url_actions(csp->action, http, b);
1511 /*********************************************************************
1513 * Function : apply_url_actions
1515 * Description : Applies a list of URL actions.
1518 * 1 : action = Destination.
1519 * 2 : http = Current URL
1520 * 3 : b = list of URL actions to apply
1524 *********************************************************************/
1525 void apply_url_actions(struct current_action_spec *action,
1526 struct http_request *http,
1527 struct url_actions *b)
1531 /* Should never happen */
1535 for (b = b->next; NULL != b; b = b->next)
1537 if (url_match(b->url, http))
1539 merge_current_action(action, b->action);
1545 /*********************************************************************
1547 * Function : forward_url
1549 * Description : Should we forward this to another proxy?
1552 * 1 : http = http_request request for current URL
1553 * 2 : csp = Current client state (buffers, headers, etc...)
1555 * Returns : Pointer to forwarding information.
1557 *********************************************************************/
1558 const struct forward_spec * forward_url(struct http_request *http,
1559 struct client_state *csp)
1561 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
1562 struct forward_spec *fwd = csp->config->forward;
1571 if (url_match(fwd->url, http))