1 const char filters_rcs[] = "$Id: filters.c,v 1.72 2006/12/22 18:52:53 fabiankeil 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', `trust_url', `gif_deanimate_response',
13 * `jpeg_inspect_response', `execute_single_pcrs_command',
14 * `rewrite_url', `get_last_url'
16 * Copyright : Written by and Copyright (C) 2001, 2004-2006 the SourceForge
17 * Privoxy team. http://www.privoxy.org/
19 * Based on the Internet Junkbuster originally written
20 * by and Copyright (C) 1997 Anonymous Coders and
21 * Junkbusters Corporation. http://www.junkbusters.com
23 * This program is free software; you can redistribute it
24 * and/or modify it under the terms of the GNU General
25 * Public License as published by the Free Software
26 * Foundation; either version 2 of the License, or (at
27 * your option) any later version.
29 * This program is distributed in the hope that it will
30 * be useful, but WITHOUT ANY WARRANTY; without even the
31 * implied warranty of MERCHANTABILITY or FITNESS FOR A
32 * PARTICULAR PURPOSE. See the GNU General Public
33 * License for more details.
35 * The GNU General Public License should be included with
36 * this file. If not, you can view it at
37 * http://www.gnu.org/copyleft/gpl.html
38 * or write to the Free Software Foundation, Inc., 59
39 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
43 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
44 * Modified is_untrusted_url to complain in case of
45 * write errors and to give a reason when adding new
46 * entries to the trustfile. Closes FR 1097611.
48 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
49 * Skip empty filter files in pcrs_filter_response,
50 * but don't ignore the ones that come afterwards.
51 * Fixes parts of BR 1619208.
53 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
54 * Added some sanity checks for get_last_url().
55 * Fixed possible segfault caused by my last commit.
57 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
58 * Let get_last_url() catch https URLs as well.
60 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
61 * Make sure get_last_url() behaves like advertised
62 * and fast-redirects{} can be combined with redirect{}.
64 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
65 * Implemented +redirect{s@foo@bar@} to generate
66 * a redirect based on a rewritten version of the
69 * Revision 1.66 2006/09/23 13:26:38 roro
70 * Replace TABs by spaces in source code.
72 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
73 * Fix +redirect{}. Didn't work with -fast-redirects.
75 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
76 * Block requests for untrusted URLs with status
77 * code 403 instead of 200.
79 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
80 * Don't free p which is still in use and will be later
81 * freed by free_map(). Don't claim the referrer is unknown
82 * when the client didn't set one.
84 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
85 * Feature request 595948: Re-Filter logging in single line
87 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
88 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
90 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
91 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
92 * with what was really the latest development (the v_3_0_branch branch)
94 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
95 * Multiple filter file support
97 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
98 * A few memory leaks plugged (mostly on error paths)
100 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
101 * Add the ability to check jpeg images for invalid
102 * lengths of comment blocks. Defensive strategy
103 * against the exploit:
104 * Microsoft Security Bulletin MS04-028
105 * Buffer Overrun in JPEG Processing (GDI+) Could
106 * Allow Code Execution (833987)
107 * Enabled with +inspect-jpegs in actions files.
109 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
110 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
112 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
113 * Fixed bug #839859: "See why" link URL now gets url-encoded.
115 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
118 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
119 * Added basic support for OPTIONS and TRACE HTTP methods:
120 * New function direct_response which handles OPTIONS and
121 * TRACE requests whose Max-Forwards header field is zero.
123 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
124 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
126 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
127 * - Bugfix: Executing a filters without jobs no longer results in
128 * turing off *all* filters.
129 * - Security fix: Malicious web servers can't cause a seg fault
130 * through bogus chunk sizes anymore
132 * Revision 1.58 2002/04/24 02:11:17 oes
133 * Jon's multiple AF patch: url_actions now evaluates rules
136 * Revision 1.57 2002/04/08 20:38:34 swa
139 * Revision 1.56 2002/04/05 15:51:24 oes
140 * - bugfix: error-pages now get correct request protocol
141 * - fix for invalid HTML in trust info
143 * Revision 1.55 2002/04/02 16:13:51 oes
144 * Fix: No "Go there anyway" for SSL
146 * Revision 1.54 2002/04/02 14:55:56 oes
147 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
149 * Revision 1.53 2002/03/26 22:29:54 swa
150 * we have a new homepage!
152 * Revision 1.52 2002/03/24 16:35:57 jongfoster
155 * Revision 1.51 2002/03/24 15:23:33 jongfoster
158 * Revision 1.50 2002/03/24 13:25:43 swa
159 * name change related issues
161 * Revision 1.49 2002/03/16 20:29:14 oes
164 * Revision 1.48 2002/03/13 20:25:34 oes
165 * Better logging for content filters
167 * Revision 1.47 2002/03/13 00:30:52 jongfoster
169 * Added option of always sending redirect for imageblock,
170 * currently disabled with #if 0.
172 * Revision 1.46 2002/03/12 01:42:49 oes
173 * Introduced modular filters
175 * Revision 1.45 2002/03/08 16:47:50 oes
176 * Added choice beween GIF and PNG built-in images
178 * Revision 1.44 2002/03/07 03:49:31 oes
179 * - Fixed compiler warnings etc
180 * - Changed built-in images from GIF to PNG
181 * (with regard to Unisys patent issue)
182 * - Added a 4x4 pattern PNG which is less intrusive
183 * than the logo but also clearly marks the deleted banners
185 * Revision 1.43 2002/01/22 23:51:59 jongfoster
186 * Replacing strsav() with the safer string_append().
188 * Adding missing html_encode() to error message generators. Where encoded
189 * and unencoded versions of a string were provided, removing the unencoded
192 * Revision 1.42 2002/01/17 21:00:32 jongfoster
193 * Moving all our URL and URL pattern parsing code to urlmatch.c.
195 * Using a single, simple url_match(pattern,url) function - rather than
196 * the 3-line match routine which was repeated all over the place.
198 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
200 * Using parse_http_url() to parse URLs without faking a HTTP
201 * request line for parse_http_request().
203 * Revision 1.41 2001/11/13 00:14:07 jongfoster
204 * Fixing stupid bug now I've figured out what || means.
205 * (It always returns 0 or 1, not one of it's paramaters.)
207 * Revision 1.40 2001/10/26 17:37:55 oes
208 * - Re-enabled Netscape 200/404 bug workaround in block_url():
209 * - Removed OS/2 special case
210 * - Made block_url() independant from sed() having been run
211 * - Made trust_url independant from sed() having been run
212 * - Made is_imageurl independant from sed() having been run.
213 * It now checks User-Agent: and Accept: by itself.
216 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
217 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
218 * threads to call select() simultaneously. So, it's time to do a real, live,
219 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
220 * (native). Both versions will work, but using __OS2__ offers multi-threading.
222 * Revision 1.38 2001/10/23 21:32:33 jongfoster
223 * Adding error-checking to selected functions
225 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
226 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
227 * filters.c. Added a FIXME in front of the offending code. I'll gladly
228 * put in a better/more robust fix for all parties if one is presented...
229 * It seems that just returning 200 instead of 404 would pretty much fix
230 * it for everyone, but I don't know all the history of the problem.
232 * Revision 1.36 2001/10/10 16:44:16 oes
233 * Added match_portlist function
235 * Revision 1.35 2001/10/07 15:41:23 oes
236 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
238 * New function remove_chunked_transfer_coding that strips chunked
239 * transfer coding to plain and is called by pcrs_filter_response
240 * and gif_deanimate_response if neccessary
242 * Improved handling of zero-change re_filter runs
244 * pcrs_filter_response and gif_deanimate_response now remove
245 * chunked transfer codeing before processing the body.
247 * Revision 1.34 2001/09/20 15:49:36 steudten
249 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
250 * See cgi.c fill_template().
252 * Revision 1.33 2001/09/16 17:05:14 jongfoster
253 * Removing unused #include showarg.h
255 * Revision 1.32 2001/09/16 13:21:27 jongfoster
256 * Changes to use new list functions.
258 * Revision 1.31 2001/09/16 11:38:02 jongfoster
259 * Splitting fill_template() into 2 functions:
260 * template_load() loads the file
261 * template_fill() performs the PCRS regexps.
262 * This is because the CGI edit interface has a "table row"
263 * template which is used many times in the page - this
264 * change means it's only loaded from disk once.
266 * Revision 1.30 2001/09/16 11:00:10 jongfoster
267 * New function alloc_http_response, for symmetry with free_http_response
269 * Revision 1.29 2001/09/13 23:32:40 jongfoster
270 * Moving image data to cgi.c rather than cgi.h
271 * Fixing a GPF under Win32 (and any other OS that protects global
272 * constants from being written to).
274 * Revision 1.28 2001/09/10 10:18:51 oes
275 * Silenced compiler warnings
277 * Revision 1.27 2001/08/05 16:06:20 jongfoster
278 * Modifiying "struct map" so that there are now separate header and
279 * "map_entry" structures. This means that functions which modify a
280 * map no longer need to return a pointer to the modified map.
281 * Also, it no longer reverses the order of the entries (which may be
282 * important with some advanced template substitutions).
284 * Revision 1.26 2001/07/30 22:08:36 jongfoster
285 * Tidying up #defines:
286 * - All feature #defines are now of the form FEATURE_xxx
287 * - Permanently turned off WIN_GUI_EDIT
288 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
290 * Revision 1.25 2001/07/26 10:09:46 oes
291 * Made browser detection a little less naive
293 * Revision 1.24 2001/07/25 17:22:51 oes
294 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
296 * Revision 1.23 2001/07/23 13:40:12 oes
297 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
299 * Revision 1.22 2001/07/18 12:29:34 oes
300 * - Made gif_deanimate_response respect
301 * csp->action->string[ACTION_STRING_DEANIMATE]
302 * - Logging cosmetics
304 * Revision 1.21 2001/07/13 13:59:53 oes
305 * - Introduced gif_deanimate_response which shares the
306 * generic content modification interface of pcrs_filter_response
307 * and acts as a wrapper to deanimate.c:gif_deanimate()
308 * - Renamed re_process_buffer to pcrs_filter_response
309 * - pcrs_filter_response now returns NULL on failiure
310 * - Removed all #ifdef PCRS
312 * Revision 1.20 2001/07/01 17:01:04 oes
313 * Added comments and missing return statement in is_untrusted_url()
315 * Revision 1.19 2001/06/29 21:45:41 oes
316 * Indentation, CRLF->LF, Tab-> Space
318 * Revision 1.18 2001/06/29 13:27:38 oes
319 * - Cleaned up, renamed and reorderd functions
320 * and improved comments
323 * - Ported to CGI platform. Now delivers
324 * http_response or NULL
325 * - Unified HTML and GIF generation (moved image detection
326 * and GIF generation here from jcc.c:chat())
327 * - Fixed HTTP status to:
328 * - 403 (Forbidden) for the "blocked" HTML message
329 * - 200 (OK) for GIF answers
330 * - 302 (Redirect) for redirect to GIF
333 * - Ported to CGI platform. Now delivers
334 * http_response or NULL
335 * - Separated detection of untrusted URL into
336 * (bool)is_untrusted_url
337 * - Added enforcement of untrusted requests
339 * - Moved redirect_url() from cgi.c to here
340 * and ported it to the CGI platform
342 * - Removed logentry from cancelled commit
344 * Revision 1.17 2001/06/09 10:55:28 jongfoster
345 * Changing BUFSIZ ==> BUFFER_SIZE
347 * Revision 1.16 2001/06/07 23:10:26 jongfoster
348 * Allowing unanchored domain patterns to back off and retry
349 * if they partially match. Optimized right-anchored patterns.
350 * Moving ACL and forward files into config file.
351 * Replacing struct gateway with struct forward_spec
353 * Revision 1.15 2001/06/03 19:12:00 oes
354 * extracted-CGI relevant stuff
356 * Revision 1.14 2001/06/01 10:30:55 oes
357 * Added optional left-anchoring to domaincmp
359 * Revision 1.13 2001/05/31 21:21:30 jongfoster
360 * Permissionsfile / actions file changes:
361 * - Changed "permission" to "action" throughout
362 * - changes to file format to allow string parameters
363 * - Moved helper functions to actions.c
365 * Revision 1.12 2001/05/31 17:35:20 oes
367 * - Enhanced domain part globbing with infix and prefix asterisk
368 * matching and optional unanchored operation
370 * Revision 1.11 2001/05/29 11:53:23 oes
371 * "See why" link added to "blocked" page
373 * Revision 1.10 2001/05/29 09:50:24 jongfoster
374 * Unified blocklist/imagelist/permissionslist.
375 * File format is still under discussion, but the internal changes
378 * Also modified interceptor behaviour:
379 * - We now intercept all URLs beginning with one of the following
380 * prefixes (and *only* these prefixes):
382 * * http://ijbswa.sf.net/config/
383 * * http://ijbswa.sourceforge.net/config/
384 * - New interceptors "home page" - go to http://i.j.b/ to see it.
385 * - Internal changes so that intercepted and fast redirect pages
386 * are not replaced with an image.
387 * - Interceptors now have the option to send a binary page direct
388 * to the client. (i.e. ijb-send-banner uses this)
389 * - Implemented show-url-info interceptor. (Which is why I needed
390 * the above interceptors changes - a typical URL is
391 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
392 * The previous mechanism would not have intercepted that, and
393 * if it had been intercepted then it then it would have replaced
396 * Revision 1.9 2001/05/27 22:17:04 oes
398 * - re_process_buffer no longer writes the modified buffer
399 * to the client, which was very ugly. It now returns the
400 * buffer, which it is then written by chat.
402 * - content_length now adjusts the Content-Length: header
403 * for modified documents rather than crunch()ing it.
404 * (Length info in csp->content_length, which is 0 for
405 * unmodified documents)
407 * - For this to work, sed() is called twice when filtering.
409 * Revision 1.8 2001/05/26 17:13:28 jongfoster
410 * Filled in a function comment.
412 * Revision 1.7 2001/05/26 15:26:15 jongfoster
413 * ACL feature now provides more security by immediately dropping
414 * connections from untrusted hosts.
416 * Revision 1.6 2001/05/26 00:28:36 jongfoster
417 * Automatic reloading of config file.
418 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
419 * Most of the global variables have been moved to a new
420 * struct configuration_spec, accessed through csp->config->globalname
421 * Most of the globals remaining are used by the Win32 GUI.
423 * Revision 1.5 2001/05/25 22:34:30 jongfoster
426 * Revision 1.4 2001/05/22 18:46:04 oes
428 * - Enabled filtering banners by size rather than URL
429 * by adding patterns that replace all standard banner
430 * sizes with the "Junkbuster" gif to the re_filterfile
432 * - Enabled filtering WebBugs by providing a pattern
433 * which kills all 1x1 images
435 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
436 * which is selected by the (nonstandard and therefore
437 * capital) letter 'U' in the option string.
438 * It causes the quantifiers to be ungreedy by default.
439 * Appending a ? turns back to greedy (!).
441 * - Added a new interceptor ijb-send-banner, which
442 * sends back the "Junkbuster" gif. Without imagelist or
443 * MSIE detection support, or if tinygif = 1, or the
444 * URL isn't recognized as an imageurl, a lame HTML
445 * explanation is sent instead.
447 * - Added new feature, which permits blocking remote
448 * script redirects and firing back a local redirect
450 * The feature is conditionally compiled, i.e. it
451 * can be disabled with --disable-fast-redirects,
452 * plus it must be activated by a "fast-redirects"
453 * line in the config file, has its own log level
454 * and of course wants to be displayed by show-proxy-args
455 * Note: Boy, all the #ifdefs in 1001 locations and
456 * all the fumbling with configure.in and acconfig.h
457 * were *way* more work than the feature itself :-(
459 * - Because a generic redirect template was needed for
460 * this, tinygif = 3 now uses the same.
462 * - Moved GIFs, and other static HTTP response templates
467 * - Removed some >400 CRs again (Jon, you really worked
470 * Revision 1.3 2001/05/20 16:44:47 jongfoster
471 * Removing last hardcoded Junkbusters.com URLs.
473 * Revision 1.2 2001/05/20 01:21:20 jongfoster
474 * Version 2.9.4 checkin.
475 * - Merged popupfile and cookiefile, and added control over PCRS
476 * filtering, in new "permissionsfile".
477 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
478 * file error you now get a message box (in the Win32 GUI) rather
479 * than the program exiting with no explanation.
480 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
482 * - Removed tabs from "config"
483 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
484 * - Bumped up version number.
486 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
487 * Initial import of version 2.9.3 source tree
490 *********************************************************************/
496 #include <sys/types.h>
505 #endif /* ndef __OS2__ */
506 #include <netinet/in.h>
508 #include <winsock2.h>
509 #endif /* ndef _WIN32 */
513 #endif /* def __OS2__ */
521 #include "jbsockets.h"
522 #include "miscutil.h"
526 #include "deanimate.h"
527 #include "urlmatch.h"
533 const char filters_h_rcs[] = FILTERS_H_VERSION;
535 /* Fix a problem with Solaris. There should be no effect on other
537 * Solaris's isspace() is a macro which uses it's argument directly
538 * as an array index. Therefore we need to make sure that high-bit
539 * characters generate +ve values, and ideally we also want to make
540 * the argument match the declared parameter type of "int".
542 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
546 /*********************************************************************
548 * Function : block_acl
550 * Description : Block this request?
551 * Decide yes or no based on ACL file.
554 * 1 : dst = The proxy or gateway address this is going to.
555 * Or NULL to check all possible targets.
556 * 2 : csp = Current client state (buffers, headers, etc...)
557 * Also includes the client IP address.
559 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
561 *********************************************************************/
562 int block_acl(struct access_control_addr *dst, struct client_state *csp)
564 struct access_control_list *acl = csp->config->acl;
566 /* if not using an access control list, then permit the connection */
572 /* search the list */
575 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
579 /* Just want to check if they have any access */
580 if (acl->action == ACL_PERMIT)
585 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
586 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
588 if (acl->action == ACL_PERMIT)
606 /*********************************************************************
608 * Function : acl_addr
610 * Description : Called from `load_config' to parse an ACL address.
613 * 1 : aspec = String specifying ACL address.
614 * 2 : aca = struct access_control_addr to fill in.
616 * Returns : 0 => Ok, everything else is an error.
618 *********************************************************************/
619 int acl_addr(char *aspec, struct access_control_addr *aca)
621 int i, masklength, port;
627 if ((p = strchr(aspec, '/')) != NULL)
631 if (ijb_isdigit(*p) == 0)
635 masklength = atoi(p);
638 if ((masklength < 0) || (masklength > 32))
643 if ((p = strchr(aspec, ':')) != NULL)
647 if (ijb_isdigit(*p) == 0)
656 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
658 if (aca->addr == INADDR_NONE)
663 /* build the netmask */
665 for (i=1; i <= masklength ; i++)
667 aca->mask |= (1 << (32 - i));
670 /* now mask off the host portion of the ip address
671 * (i.e. save on the network portion of the address).
673 aca->addr = aca->addr & aca->mask;
678 #endif /* def FEATURE_ACL */
681 /*********************************************************************
683 * Function : match_portlist
685 * Description : Check if a given number is covered by a comma
686 * separated list of numbers and ranges (a,b-c,d,..)
689 * 1 : portlist = String with list
690 * 2 : port = port to check
692 * Returns : 0 => no match
695 *********************************************************************/
696 int match_portlist(const char *portlist, int port)
698 char *min, *max, *next, *portlist_copy;
700 min = next = portlist_copy = strdup(portlist);
703 * Zero-terminate first item and remember offset for next
705 if (NULL != (next = strchr(portlist_copy, (int) ',')))
711 * Loop through all items, checking for match
715 if (NULL == (max = strchr(min, (int) '-')))
718 * No dash, check for equality
720 if (port == atoi(min))
729 * This is a range, so check if between min and max,
730 * or, if max was omitted, between min and 65K
733 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
747 * Zero-terminate next item and remember offset for n+1
749 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
761 /*********************************************************************
763 * Function : block_url
765 * Description : Called from `chat'. Check to see if we need to block this.
768 * 1 : csp = Current client state (buffers, headers, etc...)
770 * Returns : NULL => unblocked, else HTTP block response
772 *********************************************************************/
773 struct http_response *block_url(struct client_state *csp)
775 struct http_response *rsp;
776 const char *new_content_type = NULL;
779 * If it's not blocked, don't block it ;-)
781 if ((csp->action->flags & ACTION_BLOCK) == 0)
785 if (csp->action->flags & ACTION_REDIRECT)
787 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
790 * Else, prepare a response
792 if (NULL == (rsp = alloc_http_response()))
794 return cgi_error_memory();
798 * If it's an image-url, send back an image or redirect
799 * as specified by the relevant +image action
801 #ifdef FEATURE_IMAGE_BLOCKING
802 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
806 /* determine HOW images should be blocked */
807 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
809 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
811 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
813 #if 1 /* Two alternative strategies, use this one for now: */
815 /* and handle accordingly: */
816 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
818 rsp->status = strdup("403 Request blocked by Privoxy");
819 if (rsp->status == NULL)
821 free_http_response(rsp);
822 return cgi_error_memory();
824 rsp->body = bindup(image_pattern_data, image_pattern_length);
825 if (rsp->body == NULL)
827 free_http_response(rsp);
828 return cgi_error_memory();
830 rsp->content_length = image_pattern_length;
832 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
834 free_http_response(rsp);
835 return cgi_error_memory();
839 else if (0 == strcmpic(p, "blank"))
841 rsp->status = strdup("403 Request blocked by Privoxy");
842 if (rsp->status == NULL)
844 free_http_response(rsp);
845 return cgi_error_memory();
847 rsp->body = bindup(image_blank_data, image_blank_length);
848 if (rsp->body == NULL)
850 free_http_response(rsp);
851 return cgi_error_memory();
853 rsp->content_length = image_blank_length;
855 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
857 free_http_response(rsp);
858 return cgi_error_memory();
864 rsp->status = strdup("302 Local Redirect from Privoxy");
865 if (rsp->status == NULL)
867 free_http_response(rsp);
868 return cgi_error_memory();
871 if (enlist_unique_header(rsp->headers, "Location", p))
873 free_http_response(rsp);
874 return cgi_error_memory();
878 #else /* Following code is disabled for now */
880 /* and handle accordingly: */
881 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
883 p = CGI_PREFIX "send-banner?type=pattern";
885 else if (0 == strcmpic(p, "blank"))
887 p = CGI_PREFIX "send-banner?type=blank";
889 rsp->status = strdup("302 Local Redirect from Privoxy");
890 if (rsp->status == NULL)
892 free_http_response(rsp);
893 return cgi_error_memory();
896 if (enlist_unique_header(rsp->headers, "Location", p))
898 free_http_response(rsp);
899 return cgi_error_memory();
901 #endif /* Preceeding code is disabled for now */
903 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
906 * Send empty document.
908 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
911 rsp->body = strdup(" ");
912 rsp->content_length = 1;
914 rsp->status = strdup("403 Request blocked by Privoxy");
915 if (rsp->status == NULL)
917 free_http_response(rsp);
918 return cgi_error_memory();
920 if (new_content_type != 0)
922 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
923 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
925 free_http_response(rsp);
926 return cgi_error_memory();
932 #endif /* def FEATURE_IMAGE_BLOCKING */
935 * Else, generate an HTML "blocked" message:
939 struct map * exports;
943 * Workaround for stupid Netscape bug which prevents
944 * pages from being displayed if loading a referenced
945 * JavaScript or style sheet fails. So make it appear
946 * as if it succeeded.
948 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
949 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
950 && !strstr(p, "Gecko") /* save Mozilla, */
951 && !strstr(p, "compatible") /* MSIE */
952 && !strstr(p, "Opera")) /* and Opera. */
954 rsp->status = strdup("200 Request for blocked URL");
958 rsp->status = strdup("404 Request for blocked URL");
961 if (rsp->status == NULL)
963 free_http_response(rsp);
964 return cgi_error_memory();
967 exports = default_exports(csp, NULL);
970 free_http_response(rsp);
971 return cgi_error_memory();
974 #ifdef FEATURE_FORCE_LOAD
975 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
976 if (csp->http->ssl != 0)
977 #endif /* ndef FEATURE_FORCE_LOAD */
979 err = map_block_killer(exports, "force-support");
982 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
983 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
984 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
985 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
990 free_http_response(rsp);
991 return cgi_error_memory();
994 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
997 free_http_response(rsp);
998 return cgi_error_memory();
1002 return finish_http_response(rsp);
1007 #ifdef FEATURE_TRUST
1008 /*********************************************************************
1010 * Function : trust_url FIXME: I should be called distrust_url
1012 * Description : Calls is_untrusted_url to determine if the URL is trusted
1013 * and if not, returns a HTTP 403 response with a reject message.
1016 * 1 : csp = Current client state (buffers, headers, etc...)
1018 * Returns : NULL => trusted, else http_response.
1020 *********************************************************************/
1021 struct http_response *trust_url(struct client_state *csp)
1023 struct http_response *rsp;
1024 struct map * exports;
1025 char buf[BUFFER_SIZE];
1027 struct url_spec **tl;
1032 * Don't bother to work on trusted URLs
1034 if (!is_untrusted_url(csp))
1040 * Else, prepare a response:
1042 if (NULL == (rsp = alloc_http_response()))
1044 return cgi_error_memory();
1047 rsp->status = strdup("403 Request blocked by Privoxy");
1048 exports = default_exports(csp, NULL);
1049 if (exports == NULL || rsp->status == NULL)
1051 free_http_response(rsp);
1052 return cgi_error_memory();
1056 * Export the protocol, host, port, and referrer information
1058 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1059 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1060 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1062 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1064 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1068 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1074 free_http_response(rsp);
1075 return cgi_error_memory();
1079 * Export the trust list
1082 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1084 sprintf(buf, "<li>%s</li>\n", t->spec);
1085 string_append(&p, buf);
1087 err = map(exports, "trusted-referrers", 1, p, 0);
1092 free_http_response(rsp);
1093 return cgi_error_memory();
1097 * Export the trust info, if available
1099 if (csp->config->trust_info->first)
1101 struct list_entry *l;
1104 for (l = csp->config->trust_info->first; l ; l = l->next)
1106 sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
1107 string_append(&p, buf);
1109 err = map(exports, "trust-info", 1, p, 0);
1113 err = map_block_killer(exports, "have-trust-info");
1119 free_http_response(rsp);
1120 return cgi_error_memory();
1124 * Export the force prefix or the force conditional block killer
1126 #ifdef FEATURE_FORCE_LOAD
1127 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1128 #else /* ifndef FEATURE_FORCE_LOAD */
1129 err = map_block_killer(exports, "force-support");
1130 #endif /* ndef FEATURE_FORCE_LOAD */
1135 free_http_response(rsp);
1136 return cgi_error_memory();
1140 * Build the response
1142 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1145 free_http_response(rsp);
1146 return cgi_error_memory();
1149 return finish_http_response(rsp);
1151 #endif /* def FEATURE_TRUST */
1153 /*********************************************************************
1155 * Function : execute_single_pcrs_command
1157 * Description : Apply single pcrs command to the subject.
1158 * The subject itself is left untouched, memory for the result
1159 * is malloc()ed and it is the caller's responsibility to free
1160 * the result when it's no longer needed.
1163 * 1 : subject = the subject (== original) string
1164 * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
1165 * 3 : hits = int* for returning the number of modifications
1167 * Returns : NULL in case of errors, otherwise the
1168 * result of the pcrs command.
1170 *********************************************************************/
1171 char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
1175 char *result = NULL;
1179 assert(pcrs_command);
1182 size = strlen(subject);
1184 if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
1186 log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
1187 pcrs_command, error);
1189 else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
1191 log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
1198 job = pcrs_free_job(job);
1205 /*********************************************************************
1207 * Function : rewrite_url
1209 * Description : Rewrites a URL with a single pcrs command
1210 * and returns the result if it differs from the
1211 * original and isn't obviously invalid.
1214 * 1 : old_url = URL to rewrite.
1215 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1218 * Returns : NULL if the pcrs_command didn't change the url, or
1219 * the result of the modification.
1221 *********************************************************************/
1222 char *rewrite_url(char *old_url, const char *pcrs_command)
1224 char *new_url = NULL;
1228 assert(pcrs_command);
1230 new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
1234 log_error(LOG_LEVEL_REDIRECTS,
1235 "pcrs command \"%s\" didn't change \"%s\".",
1236 pcrs_command, old_url, new_url);
1239 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1241 log_error(LOG_LEVEL_ERROR,
1242 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1243 "but the result doesn't look like a valid URL and will be ignored.",
1244 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1249 log_error(LOG_LEVEL_REDIRECTS,
1250 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1251 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1259 #ifdef FEATURE_FAST_REDIRECTS
1260 /*********************************************************************
1262 * Function : get_last_url
1264 * Description : Search for the last URL inside a string.
1265 * If the string already is a URL, it will
1266 * be the first URL found.
1269 * 1 : subject = the string to check
1270 * 2 : redirect_mode = +fast-redirect{} mode
1272 * Returns : NULL if no URL was found, or
1273 * the last URL found.
1275 *********************************************************************/
1276 char *get_last_url(char *subject, const char *redirect_mode)
1278 char *new_url = NULL;
1282 assert(redirect_mode);
1284 subject = strdup(subject);
1285 if (subject == NULL)
1287 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1291 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1293 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1294 new_url = url_decode(subject);
1295 if (new_url != NULL)
1302 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1306 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1309 * Find the last URL encoded in the request
1312 while ((tmp = strstr(tmp, "http://")) != NULL)
1316 tmp = (new_url != NULL) ? new_url : subject;
1317 while ((tmp = strstr(tmp, "https://")) != NULL)
1322 if ((new_url != NULL)
1323 && ( (new_url != subject)
1324 || (0 == strncmpic(subject, "http://", 7))
1325 || (0 == strncmpic(subject, "https://", 8))
1329 * Return new URL if we found a redirect
1330 * or if the subject already was a URL.
1332 * The second case makes sure that we can
1333 * chain get_last_url after another redirection check
1334 * (like rewrite_url) without losing earlier redirects.
1336 new_url = strdup(new_url);
1345 #endif /* def FEATURE_FAST_REDIRECTS */
1348 /*********************************************************************
1350 * Function : redirect_url
1352 * Description : Checks if Privoxy should answer the request with
1353 * a HTTP redirect and generates the redirect if
1357 * 1 : csp = Current client state (buffers, headers, etc...)
1359 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1361 *********************************************************************/
1362 struct http_response *redirect_url(struct client_state *csp)
1364 struct http_response *rsp;
1365 #ifdef FEATURE_FAST_REDIRECTS
1367 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1368 * as compile-time option? The user can easily disable
1369 * it in his action file.
1371 char * redirect_mode;
1372 #endif /* def FEATURE_FAST_REDIRECTS */
1373 char *old_url = NULL;
1374 char *new_url = NULL;
1375 char *redirection_string;
1377 if ((csp->action->flags & ACTION_REDIRECT))
1379 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1382 * If the redirection string begins with 's',
1383 * assume it's a pcrs command, otherwise treat it as
1384 * properly formatted URL and use it for the redirection
1387 * According to RFC 2616 section 14.30 the URL
1388 * has to be absolute and if the user tries:
1389 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1390 * she would get undefined results anyway.
1394 if (*redirection_string == 's')
1396 old_url = csp->http->url;
1397 new_url = rewrite_url(old_url, redirection_string);
1401 log_error(LOG_LEVEL_REDIRECTS,
1402 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1403 redirection_string);
1404 new_url = strdup(redirection_string);
1408 #ifdef FEATURE_FAST_REDIRECTS
1409 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1411 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1414 * If it exists, use the previously rewritten URL as input
1415 * otherwise just use the old path.
1417 old_url = new_url ? new_url : strdup(csp->http->path);
1418 new_url = get_last_url(old_url, redirect_mode);
1421 #endif /* def FEATURE_FAST_REDIRECTS */
1423 /* Did any redirect action trigger? */
1426 if (0 == strcmpic(new_url, csp->http->url))
1428 log_error(LOG_LEVEL_ERROR,
1429 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1430 csp->http->url, new_url);
1435 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1437 if (NULL == (rsp = alloc_http_response()))
1440 return cgi_error_memory();
1443 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1444 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1447 free_http_response(rsp);
1448 return cgi_error_memory();
1451 return finish_http_response(rsp);
1455 /* Only reached if no redirect is required */
1461 #ifdef FEATURE_IMAGE_BLOCKING
1462 /*********************************************************************
1464 * Function : is_imageurl
1466 * Description : Given a URL, decide whether it is an image or not,
1467 * using either the info from a previous +image action
1468 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1469 * is MSIE and not on a Mac, tell from the browser's accept
1473 * 1 : csp = Current client state (buffers, headers, etc...)
1475 * Returns : True (nonzero) if URL is an image, false (0)
1478 *********************************************************************/
1479 int is_imageurl(struct client_state *csp)
1481 #ifdef FEATURE_IMAGE_DETECT_MSIE
1484 tmp = get_header_value(csp->headers, "User-Agent:");
1485 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1487 tmp = get_header_value(csp->headers, "Accept:");
1488 if (tmp && strstr(tmp, "image/gif"))
1490 /* Client will accept HTML. If this seems counterintuitive,
1500 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1502 return ((csp->action->flags & ACTION_IMAGE) != 0);
1505 #endif /* def FEATURE_IMAGE_BLOCKING */
1508 #ifdef FEATURE_TRUST
1509 /*********************************************************************
1511 * Function : is_untrusted_url
1513 * Description : Should we "distrust" this URL (and block it)?
1515 * Yes if it matches a line in the trustfile, or if the
1516 * referrer matches a line starting with "+" in the
1521 * 1 : csp = Current client state (buffers, headers, etc...)
1523 * Returns : 0 => trusted, 1 => untrusted
1525 *********************************************************************/
1526 int is_untrusted_url(struct client_state *csp)
1528 struct file_list *fl;
1529 struct block_spec *b;
1530 struct url_spec **trusted_url;
1531 struct http_request rhttp[1];
1532 const char * referer;
1536 * If we don't have a trustlist, we trust everybody
1538 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1543 memset(rhttp, '\0', sizeof(*rhttp));
1546 * Do we trust the request URL itself?
1548 for (b = b->next; b ; b = b->next)
1550 if (url_match(b->url, csp->http))
1556 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1558 /* no referrer was supplied */
1564 * If not, do we maybe trust its referrer?
1566 err = parse_http_url(referer, rhttp, csp);
1572 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1574 if (url_match(*trusted_url, rhttp))
1576 /* if the URL's referrer is from a trusted referrer, then
1577 * add the target spec to the trustfile as an unblocked
1578 * domain and return 0 (which means it's OK).
1583 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1587 char * new_entry = strdup("~");
1589 string_append(&new_entry, csp->http->hostport);
1591 path = csp->http->path;
1592 if ( (path[0] == '/')
1594 && ((path_end = strchr(path + 2, '/')) != NULL))
1596 /* since this path points into a user's home space
1597 * be sure to include this spec in the trustfile.
1599 int path_len = path_end - path; /* save offset */
1600 path = strdup(path); /* Copy string */
1603 path_end = path + path_len; /* regenerate ptr to new buffer */
1604 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1606 string_join(&new_entry, path);
1610 * Give a reason for generating this entry.
1612 string_append(&new_entry, " # Trusted referrer was: ");
1613 string_append(&new_entry, referer);
1615 if (new_entry != NULL)
1617 if (-1 == fprintf(fp, "%s\n", new_entry))
1619 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1620 new_entry, csp->config->trustfile);
1626 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1627 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1634 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1635 csp->http->hostport, csp->config->trustfile);
1642 #endif /* def FEATURE_TRUST */
1645 /*********************************************************************
1647 * Function : pcrs_filter_response
1649 * Description : Execute all text substitutions from all applying
1650 * +filter actions on the text buffer that's been accumulated
1651 * in csp->iob->buf. If this changes the contents, set
1652 * csp->content_length to the modified size and raise the
1653 * CSP_FLAG_MODIFIED flag.
1656 * 1 : csp = Current client state (buffers, headers, etc...)
1658 * Returns : a pointer to the (newly allocated) modified buffer.
1659 * or NULL if there were no hits or something went wrong
1661 *********************************************************************/
1662 char *pcrs_filter_response(struct client_state *csp)
1665 size_t size, prev_size;
1667 char *old = csp->iob->cur, *new = NULL;
1670 struct file_list *fl;
1671 struct re_filterfile_spec *b;
1672 struct list_entry *filtername;
1674 int i, found_filters = 0;
1679 if (csp->iob->cur >= csp->iob->eod)
1683 size = csp->iob->eod - csp->iob->cur;
1686 * Need to check the set of re_filterfiles...
1688 for (i = 0; i < MAX_AF_FILES; i++)
1701 if (0 == found_filters)
1703 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1708 * If the body has a "chunked" transfer-encoding,
1709 * get rid of it first, adjusting size and iob->eod
1711 if (csp->flags & CSP_FLAG_CHUNKED)
1713 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1714 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1718 csp->iob->eod = csp->iob->cur + size;
1719 csp->flags |= CSP_FLAG_MODIFIED;
1722 for (i = 0; i < MAX_AF_FILES; i++)
1725 if ((NULL == fl) || (NULL == fl->f))
1728 * Either there are no filter files
1729 * left, or this filter file just
1730 * contains no valid filters.
1732 * Continue to be sure we don't miss
1733 * valid filter files that are chained
1734 * after empty or invalid ones.
1739 * For all applying +filter actions, look if a filter by that
1740 * name exists and if yes, execute it's pcrs_joblist on the
1743 for (b = fl->f; b; b = b->next)
1745 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1746 filtername ; filtername = filtername->next)
1748 if (strcmp(b->name, filtername->str) == 0)
1750 int current_hits = 0; /* Number of hits caused by this filter */
1751 int job_number = 0; /* Which job we're currently executing */
1752 int job_hits = 0; /* How many hits the current job caused */
1754 if ( NULL == b->joblist )
1756 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1761 /* Apply all jobs from the joblist */
1762 for (job = b->joblist; NULL != job; job = job->next)
1765 job_hits = pcrs_execute(job, old, size, &new, &size);
1770 * That went well. Continue filtering
1771 * and use the result of this job as
1772 * input for the next one.
1774 current_hits += job_hits;
1775 if (old != csp->iob->cur)
1784 * The job caused an unexpected error. Inform the user
1785 * and skip the rest of jobs in this filter. We could
1786 * continue with the next job, but usually the jobs
1787 * depend on each other or are similar enough to
1788 * fail with the same reason.
1790 * XXX: In theory pcrs_strerror() would
1791 * return a proper error message here.
1793 * At the moment, however, our pcrs expects the
1794 * error codes of pcre 3.4 and newer pcre version
1795 * return different error codes. As a result
1796 * pcrs_strerror()'s error message might be bogus,
1797 * therefore we print the numerical value as well.
1799 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1800 b->name, job_number, pcrs_strerror (job_hits), job_hits);
1805 log_error(LOG_LEVEL_RE_FILTER,
1806 "re_filtering %s%s (size %d) with filter %s produced %d hits (new size %d).",
1807 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1809 hits += current_hits;
1816 * If there were no hits, destroy our copy and let
1817 * chat() use the original in csp->iob
1825 csp->flags |= CSP_FLAG_MODIFIED;
1826 csp->content_length = size;
1834 /*********************************************************************
1836 * Function : gif_deanimate_response
1838 * Description : Deanimate the GIF image that has been accumulated in
1839 * csp->iob->buf, set csp->content_length to the modified
1840 * size and raise the CSP_FLAG_MODIFIED flag.
1843 * 1 : csp = Current client state (buffers, headers, etc...)
1845 * Returns : a pointer to the (newly allocated) modified buffer.
1846 * or NULL in case something went wrong.
1848 *********************************************************************/
1849 char *gif_deanimate_response(struct client_state *csp)
1851 struct binbuffer *in, *out;
1853 size_t size = csp->iob->eod - csp->iob->cur;
1856 * If the body has a "chunked" transfer-encoding,
1857 * get rid of it first, adjusting size and iob->eod
1859 if (csp->flags & CSP_FLAG_CHUNKED)
1861 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1862 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1866 csp->iob->eod = csp->iob->cur + size;
1867 csp->flags |= CSP_FLAG_MODIFIED;
1870 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1871 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1873 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1877 in->buffer = csp->iob->cur;
1880 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1882 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1889 if ((int)size == out->offset)
1891 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1895 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1897 csp->content_length = out->offset;
1898 csp->flags |= CSP_FLAG_MODIFIED;
1908 /*********************************************************************
1910 * Function : jpeg_inspect_response
1915 * 1 : csp = Current client state (buffers, headers, etc...)
1917 * Returns : a pointer to the (newly allocated) modified buffer
1918 * or NULL in case something went wrong.
1920 *********************************************************************/
1921 char *jpeg_inspect_response(struct client_state *csp)
1923 struct binbuffer *in = NULL, *out = NULL;
1925 size_t size = csp->iob->eod - csp->iob->cur;
1928 * If the body has a "chunked" transfer-encoding,
1929 * get rid of it first, adjusting size and iob->eod
1931 if (csp->flags & CSP_FLAG_CHUNKED)
1933 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1934 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1938 csp->iob->eod = csp->iob->cur + size;
1939 csp->flags |= CSP_FLAG_MODIFIED;
1942 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1944 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
1948 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
1950 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
1954 in->buffer = csp->iob->cur;
1958 * Calling jpeg_inspect has the side-effect of creating and
1959 * modifying the image buffer of "out" directly.
1961 if (jpeg_inspect(in, out))
1963 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
1971 csp->content_length = out->offset;
1972 csp->flags |= CSP_FLAG_MODIFIED;
1982 /*********************************************************************
1984 * Function : remove_chunked_transfer_coding
1986 * Description : In-situ remove the "chunked" transfer coding as defined
1987 * in rfc2616 from a buffer.
1990 * 1 : buffer = Pointer to the text buffer
1991 * 2 : size = Number of bytes to be processed
1993 * Returns : The new size, i.e. the number of bytes from buffer which
1994 * are occupied by the stripped body, or 0 in case something
1997 *********************************************************************/
1998 int remove_chunked_transfer_coding(char *buffer, const size_t size)
2001 unsigned int chunksize = 0;
2002 char *from_p, *to_p;
2005 from_p = to_p = buffer;
2007 if (sscanf(buffer, "%x", &chunksize) != 1)
2009 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2013 while (chunksize > 0)
2015 if (NULL == (from_p = strstr(from_p, "\r\n")))
2017 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2021 if ((newsize += chunksize) >= size)
2023 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2028 memmove(to_p, from_p, (size_t) chunksize);
2029 to_p = buffer + newsize;
2030 from_p += chunksize + 2;
2032 if (sscanf(from_p, "%x", &chunksize) != 1)
2034 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2039 /* FIXME: Should this get its own loglevel? */
2040 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
2046 /*********************************************************************
2048 * Function : url_actions
2050 * Description : Gets the actions for this URL.
2053 * 1 : http = http_request request for blocked URLs
2054 * 2 : csp = Current client state (buffers, headers, etc...)
2058 *********************************************************************/
2059 void url_actions(struct http_request *http,
2060 struct client_state *csp)
2062 struct file_list *fl;
2063 struct url_actions *b;
2066 init_current_action(csp->action);
2068 for (i = 0; i < MAX_AF_FILES; i++)
2070 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2075 apply_url_actions(csp->action, http, b);
2082 /*********************************************************************
2084 * Function : apply_url_actions
2086 * Description : Applies a list of URL actions.
2089 * 1 : action = Destination.
2090 * 2 : http = Current URL
2091 * 3 : b = list of URL actions to apply
2095 *********************************************************************/
2096 void apply_url_actions(struct current_action_spec *action,
2097 struct http_request *http,
2098 struct url_actions *b)
2102 /* Should never happen */
2106 for (b = b->next; NULL != b; b = b->next)
2108 if (url_match(b->url, http))
2110 merge_current_action(action, b->action);
2116 /*********************************************************************
2118 * Function : forward_url
2120 * Description : Should we forward this to another proxy?
2123 * 1 : http = http_request request for current URL
2124 * 2 : csp = Current client state (buffers, headers, etc...)
2126 * Returns : Pointer to forwarding information.
2128 *********************************************************************/
2129 const struct forward_spec * forward_url(struct http_request *http,
2130 struct client_state *csp)
2132 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2133 struct forward_spec *fwd = csp->config->forward;
2142 if (url_match(fwd->url, http))
2153 /*********************************************************************
2155 * Function : direct_response
2157 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2158 * request and if so, return a HTTP 501 to the client.
2160 * FIXME: I have a stupid name and I should handle the
2161 * requests properly. Still, what we do here is rfc-
2162 * compliant, whereas ignoring or forwarding are not.
2165 * 1 : csp = Current client state (buffers, headers, etc...)
2167 * Returns : http_response if , NULL if nonmatch or handler fail
2169 *********************************************************************/
2170 struct http_response *direct_response(struct client_state *csp)
2172 struct http_response *rsp;
2173 struct list_entry *p;
2175 if ((0 == strcmpic(csp->http->gpc, "trace"))
2176 || (0 == strcmpic(csp->http->gpc, "options")))
2178 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2180 if (!strncmp("Max-Forwards:", p->str, 13)
2181 && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
2183 /* FIXME: We could handle at least TRACE here,
2184 but that would require a verbatim copy of
2185 the request which we don't have anymore */
2187 log_error(LOG_LEVEL_HEADER, "Found Max-Forwards:0 in OPTIONS or TRACE request -- Returning 501");
2189 /* Get mem for response or fail*/
2190 if (NULL == (rsp = alloc_http_response()))
2192 return cgi_error_memory();
2195 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2197 free_http_response(rsp);
2198 return cgi_error_memory();
2202 return(finish_http_response(rsp));