1 const char filters_rcs[] = "$Id: filters.c,v 1.74 2006/12/24 17:37:38 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 1.74 2006/12/24 17:37:38 fabiankeil
44 * Adjust comment in pcrs_filter_response()
45 * to recent pcrs changes. Hohoho.
47 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
48 * Don't crash if pcre returns an error code
49 * that pcrs didn't expect. Fixes BR 1621173.
51 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
52 * Modified is_untrusted_url to complain in case of
53 * write errors and to give a reason when adding new
54 * entries to the trustfile. Closes FR 1097611.
56 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
57 * Skip empty filter files in pcrs_filter_response,
58 * but don't ignore the ones that come afterwards.
59 * Fixes parts of BR 1619208.
61 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
62 * Added some sanity checks for get_last_url().
63 * Fixed possible segfault caused by my last commit.
65 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
66 * Let get_last_url() catch https URLs as well.
68 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
69 * Make sure get_last_url() behaves like advertised
70 * and fast-redirects{} can be combined with redirect{}.
72 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
73 * Implemented +redirect{s@foo@bar@} to generate
74 * a redirect based on a rewritten version of the
77 * Revision 1.66 2006/09/23 13:26:38 roro
78 * Replace TABs by spaces in source code.
80 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
81 * Fix +redirect{}. Didn't work with -fast-redirects.
83 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
84 * Block requests for untrusted URLs with status
85 * code 403 instead of 200.
87 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
88 * Don't free p which is still in use and will be later
89 * freed by free_map(). Don't claim the referrer is unknown
90 * when the client didn't set one.
92 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
93 * Feature request 595948: Re-Filter logging in single line
95 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
96 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
98 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
99 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
100 * with what was really the latest development (the v_3_0_branch branch)
102 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
103 * Multiple filter file support
105 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
106 * A few memory leaks plugged (mostly on error paths)
108 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
109 * Add the ability to check jpeg images for invalid
110 * lengths of comment blocks. Defensive strategy
111 * against the exploit:
112 * Microsoft Security Bulletin MS04-028
113 * Buffer Overrun in JPEG Processing (GDI+) Could
114 * Allow Code Execution (833987)
115 * Enabled with +inspect-jpegs in actions files.
117 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
118 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
120 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
121 * Fixed bug #839859: "See why" link URL now gets url-encoded.
123 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
126 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
127 * Added basic support for OPTIONS and TRACE HTTP methods:
128 * New function direct_response which handles OPTIONS and
129 * TRACE requests whose Max-Forwards header field is zero.
131 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
132 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
134 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
135 * - Bugfix: Executing a filters without jobs no longer results in
136 * turing off *all* filters.
137 * - Security fix: Malicious web servers can't cause a seg fault
138 * through bogus chunk sizes anymore
140 * Revision 1.58 2002/04/24 02:11:17 oes
141 * Jon's multiple AF patch: url_actions now evaluates rules
144 * Revision 1.57 2002/04/08 20:38:34 swa
147 * Revision 1.56 2002/04/05 15:51:24 oes
148 * - bugfix: error-pages now get correct request protocol
149 * - fix for invalid HTML in trust info
151 * Revision 1.55 2002/04/02 16:13:51 oes
152 * Fix: No "Go there anyway" for SSL
154 * Revision 1.54 2002/04/02 14:55:56 oes
155 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
157 * Revision 1.53 2002/03/26 22:29:54 swa
158 * we have a new homepage!
160 * Revision 1.52 2002/03/24 16:35:57 jongfoster
163 * Revision 1.51 2002/03/24 15:23:33 jongfoster
166 * Revision 1.50 2002/03/24 13:25:43 swa
167 * name change related issues
169 * Revision 1.49 2002/03/16 20:29:14 oes
172 * Revision 1.48 2002/03/13 20:25:34 oes
173 * Better logging for content filters
175 * Revision 1.47 2002/03/13 00:30:52 jongfoster
177 * Added option of always sending redirect for imageblock,
178 * currently disabled with #if 0.
180 * Revision 1.46 2002/03/12 01:42:49 oes
181 * Introduced modular filters
183 * Revision 1.45 2002/03/08 16:47:50 oes
184 * Added choice beween GIF and PNG built-in images
186 * Revision 1.44 2002/03/07 03:49:31 oes
187 * - Fixed compiler warnings etc
188 * - Changed built-in images from GIF to PNG
189 * (with regard to Unisys patent issue)
190 * - Added a 4x4 pattern PNG which is less intrusive
191 * than the logo but also clearly marks the deleted banners
193 * Revision 1.43 2002/01/22 23:51:59 jongfoster
194 * Replacing strsav() with the safer string_append().
196 * Adding missing html_encode() to error message generators. Where encoded
197 * and unencoded versions of a string were provided, removing the unencoded
200 * Revision 1.42 2002/01/17 21:00:32 jongfoster
201 * Moving all our URL and URL pattern parsing code to urlmatch.c.
203 * Using a single, simple url_match(pattern,url) function - rather than
204 * the 3-line match routine which was repeated all over the place.
206 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
208 * Using parse_http_url() to parse URLs without faking a HTTP
209 * request line for parse_http_request().
211 * Revision 1.41 2001/11/13 00:14:07 jongfoster
212 * Fixing stupid bug now I've figured out what || means.
213 * (It always returns 0 or 1, not one of it's paramaters.)
215 * Revision 1.40 2001/10/26 17:37:55 oes
216 * - Re-enabled Netscape 200/404 bug workaround in block_url():
217 * - Removed OS/2 special case
218 * - Made block_url() independant from sed() having been run
219 * - Made trust_url independant from sed() having been run
220 * - Made is_imageurl independant from sed() having been run.
221 * It now checks User-Agent: and Accept: by itself.
224 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
225 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
226 * threads to call select() simultaneously. So, it's time to do a real, live,
227 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
228 * (native). Both versions will work, but using __OS2__ offers multi-threading.
230 * Revision 1.38 2001/10/23 21:32:33 jongfoster
231 * Adding error-checking to selected functions
233 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
234 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
235 * filters.c. Added a FIXME in front of the offending code. I'll gladly
236 * put in a better/more robust fix for all parties if one is presented...
237 * It seems that just returning 200 instead of 404 would pretty much fix
238 * it for everyone, but I don't know all the history of the problem.
240 * Revision 1.36 2001/10/10 16:44:16 oes
241 * Added match_portlist function
243 * Revision 1.35 2001/10/07 15:41:23 oes
244 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
246 * New function remove_chunked_transfer_coding that strips chunked
247 * transfer coding to plain and is called by pcrs_filter_response
248 * and gif_deanimate_response if neccessary
250 * Improved handling of zero-change re_filter runs
252 * pcrs_filter_response and gif_deanimate_response now remove
253 * chunked transfer codeing before processing the body.
255 * Revision 1.34 2001/09/20 15:49:36 steudten
257 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
258 * See cgi.c fill_template().
260 * Revision 1.33 2001/09/16 17:05:14 jongfoster
261 * Removing unused #include showarg.h
263 * Revision 1.32 2001/09/16 13:21:27 jongfoster
264 * Changes to use new list functions.
266 * Revision 1.31 2001/09/16 11:38:02 jongfoster
267 * Splitting fill_template() into 2 functions:
268 * template_load() loads the file
269 * template_fill() performs the PCRS regexps.
270 * This is because the CGI edit interface has a "table row"
271 * template which is used many times in the page - this
272 * change means it's only loaded from disk once.
274 * Revision 1.30 2001/09/16 11:00:10 jongfoster
275 * New function alloc_http_response, for symmetry with free_http_response
277 * Revision 1.29 2001/09/13 23:32:40 jongfoster
278 * Moving image data to cgi.c rather than cgi.h
279 * Fixing a GPF under Win32 (and any other OS that protects global
280 * constants from being written to).
282 * Revision 1.28 2001/09/10 10:18:51 oes
283 * Silenced compiler warnings
285 * Revision 1.27 2001/08/05 16:06:20 jongfoster
286 * Modifiying "struct map" so that there are now separate header and
287 * "map_entry" structures. This means that functions which modify a
288 * map no longer need to return a pointer to the modified map.
289 * Also, it no longer reverses the order of the entries (which may be
290 * important with some advanced template substitutions).
292 * Revision 1.26 2001/07/30 22:08:36 jongfoster
293 * Tidying up #defines:
294 * - All feature #defines are now of the form FEATURE_xxx
295 * - Permanently turned off WIN_GUI_EDIT
296 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
298 * Revision 1.25 2001/07/26 10:09:46 oes
299 * Made browser detection a little less naive
301 * Revision 1.24 2001/07/25 17:22:51 oes
302 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
304 * Revision 1.23 2001/07/23 13:40:12 oes
305 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
307 * Revision 1.22 2001/07/18 12:29:34 oes
308 * - Made gif_deanimate_response respect
309 * csp->action->string[ACTION_STRING_DEANIMATE]
310 * - Logging cosmetics
312 * Revision 1.21 2001/07/13 13:59:53 oes
313 * - Introduced gif_deanimate_response which shares the
314 * generic content modification interface of pcrs_filter_response
315 * and acts as a wrapper to deanimate.c:gif_deanimate()
316 * - Renamed re_process_buffer to pcrs_filter_response
317 * - pcrs_filter_response now returns NULL on failiure
318 * - Removed all #ifdef PCRS
320 * Revision 1.20 2001/07/01 17:01:04 oes
321 * Added comments and missing return statement in is_untrusted_url()
323 * Revision 1.19 2001/06/29 21:45:41 oes
324 * Indentation, CRLF->LF, Tab-> Space
326 * Revision 1.18 2001/06/29 13:27:38 oes
327 * - Cleaned up, renamed and reorderd functions
328 * and improved comments
331 * - Ported to CGI platform. Now delivers
332 * http_response or NULL
333 * - Unified HTML and GIF generation (moved image detection
334 * and GIF generation here from jcc.c:chat())
335 * - Fixed HTTP status to:
336 * - 403 (Forbidden) for the "blocked" HTML message
337 * - 200 (OK) for GIF answers
338 * - 302 (Redirect) for redirect to GIF
341 * - Ported to CGI platform. Now delivers
342 * http_response or NULL
343 * - Separated detection of untrusted URL into
344 * (bool)is_untrusted_url
345 * - Added enforcement of untrusted requests
347 * - Moved redirect_url() from cgi.c to here
348 * and ported it to the CGI platform
350 * - Removed logentry from cancelled commit
352 * Revision 1.17 2001/06/09 10:55:28 jongfoster
353 * Changing BUFSIZ ==> BUFFER_SIZE
355 * Revision 1.16 2001/06/07 23:10:26 jongfoster
356 * Allowing unanchored domain patterns to back off and retry
357 * if they partially match. Optimized right-anchored patterns.
358 * Moving ACL and forward files into config file.
359 * Replacing struct gateway with struct forward_spec
361 * Revision 1.15 2001/06/03 19:12:00 oes
362 * extracted-CGI relevant stuff
364 * Revision 1.14 2001/06/01 10:30:55 oes
365 * Added optional left-anchoring to domaincmp
367 * Revision 1.13 2001/05/31 21:21:30 jongfoster
368 * Permissionsfile / actions file changes:
369 * - Changed "permission" to "action" throughout
370 * - changes to file format to allow string parameters
371 * - Moved helper functions to actions.c
373 * Revision 1.12 2001/05/31 17:35:20 oes
375 * - Enhanced domain part globbing with infix and prefix asterisk
376 * matching and optional unanchored operation
378 * Revision 1.11 2001/05/29 11:53:23 oes
379 * "See why" link added to "blocked" page
381 * Revision 1.10 2001/05/29 09:50:24 jongfoster
382 * Unified blocklist/imagelist/permissionslist.
383 * File format is still under discussion, but the internal changes
386 * Also modified interceptor behaviour:
387 * - We now intercept all URLs beginning with one of the following
388 * prefixes (and *only* these prefixes):
390 * * http://ijbswa.sf.net/config/
391 * * http://ijbswa.sourceforge.net/config/
392 * - New interceptors "home page" - go to http://i.j.b/ to see it.
393 * - Internal changes so that intercepted and fast redirect pages
394 * are not replaced with an image.
395 * - Interceptors now have the option to send a binary page direct
396 * to the client. (i.e. ijb-send-banner uses this)
397 * - Implemented show-url-info interceptor. (Which is why I needed
398 * the above interceptors changes - a typical URL is
399 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
400 * The previous mechanism would not have intercepted that, and
401 * if it had been intercepted then it then it would have replaced
404 * Revision 1.9 2001/05/27 22:17:04 oes
406 * - re_process_buffer no longer writes the modified buffer
407 * to the client, which was very ugly. It now returns the
408 * buffer, which it is then written by chat.
410 * - content_length now adjusts the Content-Length: header
411 * for modified documents rather than crunch()ing it.
412 * (Length info in csp->content_length, which is 0 for
413 * unmodified documents)
415 * - For this to work, sed() is called twice when filtering.
417 * Revision 1.8 2001/05/26 17:13:28 jongfoster
418 * Filled in a function comment.
420 * Revision 1.7 2001/05/26 15:26:15 jongfoster
421 * ACL feature now provides more security by immediately dropping
422 * connections from untrusted hosts.
424 * Revision 1.6 2001/05/26 00:28:36 jongfoster
425 * Automatic reloading of config file.
426 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
427 * Most of the global variables have been moved to a new
428 * struct configuration_spec, accessed through csp->config->globalname
429 * Most of the globals remaining are used by the Win32 GUI.
431 * Revision 1.5 2001/05/25 22:34:30 jongfoster
434 * Revision 1.4 2001/05/22 18:46:04 oes
436 * - Enabled filtering banners by size rather than URL
437 * by adding patterns that replace all standard banner
438 * sizes with the "Junkbuster" gif to the re_filterfile
440 * - Enabled filtering WebBugs by providing a pattern
441 * which kills all 1x1 images
443 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
444 * which is selected by the (nonstandard and therefore
445 * capital) letter 'U' in the option string.
446 * It causes the quantifiers to be ungreedy by default.
447 * Appending a ? turns back to greedy (!).
449 * - Added a new interceptor ijb-send-banner, which
450 * sends back the "Junkbuster" gif. Without imagelist or
451 * MSIE detection support, or if tinygif = 1, or the
452 * URL isn't recognized as an imageurl, a lame HTML
453 * explanation is sent instead.
455 * - Added new feature, which permits blocking remote
456 * script redirects and firing back a local redirect
458 * The feature is conditionally compiled, i.e. it
459 * can be disabled with --disable-fast-redirects,
460 * plus it must be activated by a "fast-redirects"
461 * line in the config file, has its own log level
462 * and of course wants to be displayed by show-proxy-args
463 * Note: Boy, all the #ifdefs in 1001 locations and
464 * all the fumbling with configure.in and acconfig.h
465 * were *way* more work than the feature itself :-(
467 * - Because a generic redirect template was needed for
468 * this, tinygif = 3 now uses the same.
470 * - Moved GIFs, and other static HTTP response templates
475 * - Removed some >400 CRs again (Jon, you really worked
478 * Revision 1.3 2001/05/20 16:44:47 jongfoster
479 * Removing last hardcoded Junkbusters.com URLs.
481 * Revision 1.2 2001/05/20 01:21:20 jongfoster
482 * Version 2.9.4 checkin.
483 * - Merged popupfile and cookiefile, and added control over PCRS
484 * filtering, in new "permissionsfile".
485 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
486 * file error you now get a message box (in the Win32 GUI) rather
487 * than the program exiting with no explanation.
488 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
490 * - Removed tabs from "config"
491 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
492 * - Bumped up version number.
494 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
495 * Initial import of version 2.9.3 source tree
498 *********************************************************************/
504 #include <sys/types.h>
513 #endif /* ndef __OS2__ */
514 #include <netinet/in.h>
516 #include <winsock2.h>
517 #endif /* ndef _WIN32 */
521 #endif /* def __OS2__ */
529 #include "jbsockets.h"
530 #include "miscutil.h"
534 #include "deanimate.h"
535 #include "urlmatch.h"
541 const char filters_h_rcs[] = FILTERS_H_VERSION;
543 /* Fix a problem with Solaris. There should be no effect on other
545 * Solaris's isspace() is a macro which uses it's argument directly
546 * as an array index. Therefore we need to make sure that high-bit
547 * characters generate +ve values, and ideally we also want to make
548 * the argument match the declared parameter type of "int".
550 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
554 /*********************************************************************
556 * Function : block_acl
558 * Description : Block this request?
559 * Decide yes or no based on ACL file.
562 * 1 : dst = The proxy or gateway address this is going to.
563 * Or NULL to check all possible targets.
564 * 2 : csp = Current client state (buffers, headers, etc...)
565 * Also includes the client IP address.
567 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
569 *********************************************************************/
570 int block_acl(struct access_control_addr *dst, struct client_state *csp)
572 struct access_control_list *acl = csp->config->acl;
574 /* if not using an access control list, then permit the connection */
580 /* search the list */
583 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
587 /* Just want to check if they have any access */
588 if (acl->action == ACL_PERMIT)
593 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
594 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
596 if (acl->action == ACL_PERMIT)
614 /*********************************************************************
616 * Function : acl_addr
618 * Description : Called from `load_config' to parse an ACL address.
621 * 1 : aspec = String specifying ACL address.
622 * 2 : aca = struct access_control_addr to fill in.
624 * Returns : 0 => Ok, everything else is an error.
626 *********************************************************************/
627 int acl_addr(char *aspec, struct access_control_addr *aca)
629 int i, masklength, port;
635 if ((p = strchr(aspec, '/')) != NULL)
639 if (ijb_isdigit(*p) == 0)
643 masklength = atoi(p);
646 if ((masklength < 0) || (masklength > 32))
651 if ((p = strchr(aspec, ':')) != NULL)
655 if (ijb_isdigit(*p) == 0)
664 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
666 if (aca->addr == INADDR_NONE)
671 /* build the netmask */
673 for (i=1; i <= masklength ; i++)
675 aca->mask |= (1 << (32 - i));
678 /* now mask off the host portion of the ip address
679 * (i.e. save on the network portion of the address).
681 aca->addr = aca->addr & aca->mask;
686 #endif /* def FEATURE_ACL */
689 /*********************************************************************
691 * Function : match_portlist
693 * Description : Check if a given number is covered by a comma
694 * separated list of numbers and ranges (a,b-c,d,..)
697 * 1 : portlist = String with list
698 * 2 : port = port to check
700 * Returns : 0 => no match
703 *********************************************************************/
704 int match_portlist(const char *portlist, int port)
706 char *min, *max, *next, *portlist_copy;
708 min = next = portlist_copy = strdup(portlist);
711 * Zero-terminate first item and remember offset for next
713 if (NULL != (next = strchr(portlist_copy, (int) ',')))
719 * Loop through all items, checking for match
723 if (NULL == (max = strchr(min, (int) '-')))
726 * No dash, check for equality
728 if (port == atoi(min))
737 * This is a range, so check if between min and max,
738 * or, if max was omitted, between min and 65K
741 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
755 * Zero-terminate next item and remember offset for n+1
757 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
769 /*********************************************************************
771 * Function : block_url
773 * Description : Called from `chat'. Check to see if we need to block this.
776 * 1 : csp = Current client state (buffers, headers, etc...)
778 * Returns : NULL => unblocked, else HTTP block response
780 *********************************************************************/
781 struct http_response *block_url(struct client_state *csp)
783 struct http_response *rsp;
784 const char *new_content_type = NULL;
787 * If it's not blocked, don't block it ;-)
789 if ((csp->action->flags & ACTION_BLOCK) == 0)
793 if (csp->action->flags & ACTION_REDIRECT)
795 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
798 * Else, prepare a response
800 if (NULL == (rsp = alloc_http_response()))
802 return cgi_error_memory();
806 * If it's an image-url, send back an image or redirect
807 * as specified by the relevant +image action
809 #ifdef FEATURE_IMAGE_BLOCKING
810 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
814 /* determine HOW images should be blocked */
815 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
817 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
819 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
821 #if 1 /* Two alternative strategies, use this one for now: */
823 /* and handle accordingly: */
824 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
826 rsp->status = strdup("403 Request blocked by Privoxy");
827 if (rsp->status == NULL)
829 free_http_response(rsp);
830 return cgi_error_memory();
832 rsp->body = bindup(image_pattern_data, image_pattern_length);
833 if (rsp->body == NULL)
835 free_http_response(rsp);
836 return cgi_error_memory();
838 rsp->content_length = image_pattern_length;
840 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
842 free_http_response(rsp);
843 return cgi_error_memory();
847 else if (0 == strcmpic(p, "blank"))
849 rsp->status = strdup("403 Request blocked by Privoxy");
850 if (rsp->status == NULL)
852 free_http_response(rsp);
853 return cgi_error_memory();
855 rsp->body = bindup(image_blank_data, image_blank_length);
856 if (rsp->body == NULL)
858 free_http_response(rsp);
859 return cgi_error_memory();
861 rsp->content_length = image_blank_length;
863 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
865 free_http_response(rsp);
866 return cgi_error_memory();
872 rsp->status = strdup("302 Local Redirect from Privoxy");
873 if (rsp->status == NULL)
875 free_http_response(rsp);
876 return cgi_error_memory();
879 if (enlist_unique_header(rsp->headers, "Location", p))
881 free_http_response(rsp);
882 return cgi_error_memory();
886 #else /* Following code is disabled for now */
888 /* and handle accordingly: */
889 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
891 p = CGI_PREFIX "send-banner?type=pattern";
893 else if (0 == strcmpic(p, "blank"))
895 p = CGI_PREFIX "send-banner?type=blank";
897 rsp->status = strdup("302 Local Redirect from Privoxy");
898 if (rsp->status == NULL)
900 free_http_response(rsp);
901 return cgi_error_memory();
904 if (enlist_unique_header(rsp->headers, "Location", p))
906 free_http_response(rsp);
907 return cgi_error_memory();
909 #endif /* Preceeding code is disabled for now */
911 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
914 * Send empty document.
916 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
919 rsp->body = strdup(" ");
920 rsp->content_length = 1;
922 rsp->status = strdup("403 Request blocked by Privoxy");
923 if (rsp->status == NULL)
925 free_http_response(rsp);
926 return cgi_error_memory();
928 if (new_content_type != 0)
930 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
931 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
933 free_http_response(rsp);
934 return cgi_error_memory();
940 #endif /* def FEATURE_IMAGE_BLOCKING */
943 * Else, generate an HTML "blocked" message:
947 struct map * exports;
951 * Workaround for stupid Netscape bug which prevents
952 * pages from being displayed if loading a referenced
953 * JavaScript or style sheet fails. So make it appear
954 * as if it succeeded.
956 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
957 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
958 && !strstr(p, "Gecko") /* save Mozilla, */
959 && !strstr(p, "compatible") /* MSIE */
960 && !strstr(p, "Opera")) /* and Opera. */
962 rsp->status = strdup("200 Request for blocked URL");
966 rsp->status = strdup("404 Request for blocked URL");
969 if (rsp->status == NULL)
971 free_http_response(rsp);
972 return cgi_error_memory();
975 exports = default_exports(csp, NULL);
978 free_http_response(rsp);
979 return cgi_error_memory();
982 #ifdef FEATURE_FORCE_LOAD
983 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
984 if (csp->http->ssl != 0)
985 #endif /* ndef FEATURE_FORCE_LOAD */
987 err = map_block_killer(exports, "force-support");
990 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
991 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
992 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
993 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
998 free_http_response(rsp);
999 return cgi_error_memory();
1002 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1005 free_http_response(rsp);
1006 return cgi_error_memory();
1010 return finish_http_response(rsp);
1015 #ifdef FEATURE_TRUST
1016 /*********************************************************************
1018 * Function : trust_url FIXME: I should be called distrust_url
1020 * Description : Calls is_untrusted_url to determine if the URL is trusted
1021 * and if not, returns a HTTP 403 response with a reject message.
1024 * 1 : csp = Current client state (buffers, headers, etc...)
1026 * Returns : NULL => trusted, else http_response.
1028 *********************************************************************/
1029 struct http_response *trust_url(struct client_state *csp)
1031 struct http_response *rsp;
1032 struct map * exports;
1033 char buf[BUFFER_SIZE];
1035 struct url_spec **tl;
1040 * Don't bother to work on trusted URLs
1042 if (!is_untrusted_url(csp))
1048 * Else, prepare a response:
1050 if (NULL == (rsp = alloc_http_response()))
1052 return cgi_error_memory();
1055 rsp->status = strdup("403 Request blocked by Privoxy");
1056 exports = default_exports(csp, NULL);
1057 if (exports == NULL || rsp->status == NULL)
1059 free_http_response(rsp);
1060 return cgi_error_memory();
1064 * Export the protocol, host, port, and referrer information
1066 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1067 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1068 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1070 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1072 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1076 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1082 free_http_response(rsp);
1083 return cgi_error_memory();
1087 * Export the trust list
1090 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1092 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1093 string_append(&p, buf);
1095 err = map(exports, "trusted-referrers", 1, p, 0);
1100 free_http_response(rsp);
1101 return cgi_error_memory();
1105 * Export the trust info, if available
1107 if (csp->config->trust_info->first)
1109 struct list_entry *l;
1112 for (l = csp->config->trust_info->first; l ; l = l->next)
1114 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1115 string_append(&p, buf);
1117 err = map(exports, "trust-info", 1, p, 0);
1121 err = map_block_killer(exports, "have-trust-info");
1127 free_http_response(rsp);
1128 return cgi_error_memory();
1132 * Export the force prefix or the force conditional block killer
1134 #ifdef FEATURE_FORCE_LOAD
1135 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1136 #else /* ifndef FEATURE_FORCE_LOAD */
1137 err = map_block_killer(exports, "force-support");
1138 #endif /* ndef FEATURE_FORCE_LOAD */
1143 free_http_response(rsp);
1144 return cgi_error_memory();
1148 * Build the response
1150 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1153 free_http_response(rsp);
1154 return cgi_error_memory();
1157 return finish_http_response(rsp);
1159 #endif /* def FEATURE_TRUST */
1161 /*********************************************************************
1163 * Function : execute_single_pcrs_command
1165 * Description : Apply single pcrs command to the subject.
1166 * The subject itself is left untouched, memory for the result
1167 * is malloc()ed and it is the caller's responsibility to free
1168 * the result when it's no longer needed.
1171 * 1 : subject = the subject (== original) string
1172 * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
1173 * 3 : hits = int* for returning the number of modifications
1175 * Returns : NULL in case of errors, otherwise the
1176 * result of the pcrs command.
1178 *********************************************************************/
1179 char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
1183 char *result = NULL;
1187 assert(pcrs_command);
1190 size = strlen(subject);
1192 if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
1194 log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
1195 pcrs_command, error);
1197 else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
1199 log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
1206 job = pcrs_free_job(job);
1213 /*********************************************************************
1215 * Function : rewrite_url
1217 * Description : Rewrites a URL with a single pcrs command
1218 * and returns the result if it differs from the
1219 * original and isn't obviously invalid.
1222 * 1 : old_url = URL to rewrite.
1223 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1226 * Returns : NULL if the pcrs_command didn't change the url, or
1227 * the result of the modification.
1229 *********************************************************************/
1230 char *rewrite_url(char *old_url, const char *pcrs_command)
1232 char *new_url = NULL;
1236 assert(pcrs_command);
1238 new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
1242 log_error(LOG_LEVEL_REDIRECTS,
1243 "pcrs command \"%s\" didn't change \"%s\".",
1244 pcrs_command, old_url, new_url);
1247 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1249 log_error(LOG_LEVEL_ERROR,
1250 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1251 "but the result doesn't look like a valid URL and will be ignored.",
1252 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1257 log_error(LOG_LEVEL_REDIRECTS,
1258 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1259 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1267 #ifdef FEATURE_FAST_REDIRECTS
1268 /*********************************************************************
1270 * Function : get_last_url
1272 * Description : Search for the last URL inside a string.
1273 * If the string already is a URL, it will
1274 * be the first URL found.
1277 * 1 : subject = the string to check
1278 * 2 : redirect_mode = +fast-redirect{} mode
1280 * Returns : NULL if no URL was found, or
1281 * the last URL found.
1283 *********************************************************************/
1284 char *get_last_url(char *subject, const char *redirect_mode)
1286 char *new_url = NULL;
1290 assert(redirect_mode);
1292 subject = strdup(subject);
1293 if (subject == NULL)
1295 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1299 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1301 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1302 new_url = url_decode(subject);
1303 if (new_url != NULL)
1310 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1314 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1317 * Find the last URL encoded in the request
1320 while ((tmp = strstr(tmp, "http://")) != NULL)
1324 tmp = (new_url != NULL) ? new_url : subject;
1325 while ((tmp = strstr(tmp, "https://")) != NULL)
1330 if ((new_url != NULL)
1331 && ( (new_url != subject)
1332 || (0 == strncmpic(subject, "http://", 7))
1333 || (0 == strncmpic(subject, "https://", 8))
1337 * Return new URL if we found a redirect
1338 * or if the subject already was a URL.
1340 * The second case makes sure that we can
1341 * chain get_last_url after another redirection check
1342 * (like rewrite_url) without losing earlier redirects.
1344 new_url = strdup(new_url);
1353 #endif /* def FEATURE_FAST_REDIRECTS */
1356 /*********************************************************************
1358 * Function : redirect_url
1360 * Description : Checks if Privoxy should answer the request with
1361 * a HTTP redirect and generates the redirect if
1365 * 1 : csp = Current client state (buffers, headers, etc...)
1367 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1369 *********************************************************************/
1370 struct http_response *redirect_url(struct client_state *csp)
1372 struct http_response *rsp;
1373 #ifdef FEATURE_FAST_REDIRECTS
1375 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1376 * as compile-time option? The user can easily disable
1377 * it in his action file.
1379 char * redirect_mode;
1380 #endif /* def FEATURE_FAST_REDIRECTS */
1381 char *old_url = NULL;
1382 char *new_url = NULL;
1383 char *redirection_string;
1385 if ((csp->action->flags & ACTION_REDIRECT))
1387 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1390 * If the redirection string begins with 's',
1391 * assume it's a pcrs command, otherwise treat it as
1392 * properly formatted URL and use it for the redirection
1395 * According to RFC 2616 section 14.30 the URL
1396 * has to be absolute and if the user tries:
1397 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1398 * she would get undefined results anyway.
1402 if (*redirection_string == 's')
1404 old_url = csp->http->url;
1405 new_url = rewrite_url(old_url, redirection_string);
1409 log_error(LOG_LEVEL_REDIRECTS,
1410 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1411 redirection_string);
1412 new_url = strdup(redirection_string);
1416 #ifdef FEATURE_FAST_REDIRECTS
1417 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1419 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1422 * If it exists, use the previously rewritten URL as input
1423 * otherwise just use the old path.
1425 old_url = new_url ? new_url : strdup(csp->http->path);
1426 new_url = get_last_url(old_url, redirect_mode);
1429 #endif /* def FEATURE_FAST_REDIRECTS */
1431 /* Did any redirect action trigger? */
1434 if (0 == strcmpic(new_url, csp->http->url))
1436 log_error(LOG_LEVEL_ERROR,
1437 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1438 csp->http->url, new_url);
1443 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1445 if (NULL == (rsp = alloc_http_response()))
1448 return cgi_error_memory();
1451 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1452 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1455 free_http_response(rsp);
1456 return cgi_error_memory();
1459 return finish_http_response(rsp);
1463 /* Only reached if no redirect is required */
1469 #ifdef FEATURE_IMAGE_BLOCKING
1470 /*********************************************************************
1472 * Function : is_imageurl
1474 * Description : Given a URL, decide whether it is an image or not,
1475 * using either the info from a previous +image action
1476 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1477 * is MSIE and not on a Mac, tell from the browser's accept
1481 * 1 : csp = Current client state (buffers, headers, etc...)
1483 * Returns : True (nonzero) if URL is an image, false (0)
1486 *********************************************************************/
1487 int is_imageurl(struct client_state *csp)
1489 #ifdef FEATURE_IMAGE_DETECT_MSIE
1492 tmp = get_header_value(csp->headers, "User-Agent:");
1493 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1495 tmp = get_header_value(csp->headers, "Accept:");
1496 if (tmp && strstr(tmp, "image/gif"))
1498 /* Client will accept HTML. If this seems counterintuitive,
1508 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1510 return ((csp->action->flags & ACTION_IMAGE) != 0);
1513 #endif /* def FEATURE_IMAGE_BLOCKING */
1516 #ifdef FEATURE_TRUST
1517 /*********************************************************************
1519 * Function : is_untrusted_url
1521 * Description : Should we "distrust" this URL (and block it)?
1523 * Yes if it matches a line in the trustfile, or if the
1524 * referrer matches a line starting with "+" in the
1529 * 1 : csp = Current client state (buffers, headers, etc...)
1531 * Returns : 0 => trusted, 1 => untrusted
1533 *********************************************************************/
1534 int is_untrusted_url(struct client_state *csp)
1536 struct file_list *fl;
1537 struct block_spec *b;
1538 struct url_spec **trusted_url;
1539 struct http_request rhttp[1];
1540 const char * referer;
1544 * If we don't have a trustlist, we trust everybody
1546 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1551 memset(rhttp, '\0', sizeof(*rhttp));
1554 * Do we trust the request URL itself?
1556 for (b = b->next; b ; b = b->next)
1558 if (url_match(b->url, csp->http))
1564 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1566 /* no referrer was supplied */
1572 * If not, do we maybe trust its referrer?
1574 err = parse_http_url(referer, rhttp, csp);
1580 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1582 if (url_match(*trusted_url, rhttp))
1584 /* if the URL's referrer is from a trusted referrer, then
1585 * add the target spec to the trustfile as an unblocked
1586 * domain and return 0 (which means it's OK).
1591 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1595 char * new_entry = strdup("~");
1597 string_append(&new_entry, csp->http->hostport);
1599 path = csp->http->path;
1600 if ( (path[0] == '/')
1602 && ((path_end = strchr(path + 2, '/')) != NULL))
1604 /* since this path points into a user's home space
1605 * be sure to include this spec in the trustfile.
1607 int path_len = path_end - path; /* save offset */
1608 path = strdup(path); /* Copy string */
1611 path_end = path + path_len; /* regenerate ptr to new buffer */
1612 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1614 string_join(&new_entry, path);
1618 * Give a reason for generating this entry.
1620 string_append(&new_entry, " # Trusted referrer was: ");
1621 string_append(&new_entry, referer);
1623 if (new_entry != NULL)
1625 if (-1 == fprintf(fp, "%s\n", new_entry))
1627 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1628 new_entry, csp->config->trustfile);
1634 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1635 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1642 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1643 csp->http->hostport, csp->config->trustfile);
1650 #endif /* def FEATURE_TRUST */
1653 /*********************************************************************
1655 * Function : pcrs_filter_response
1657 * Description : Execute all text substitutions from all applying
1658 * +filter actions on the text buffer that's been accumulated
1659 * in csp->iob->buf. If this changes the contents, set
1660 * csp->content_length to the modified size and raise the
1661 * CSP_FLAG_MODIFIED flag.
1664 * 1 : csp = Current client state (buffers, headers, etc...)
1666 * Returns : a pointer to the (newly allocated) modified buffer.
1667 * or NULL if there were no hits or something went wrong
1669 *********************************************************************/
1670 char *pcrs_filter_response(struct client_state *csp)
1673 size_t size, prev_size;
1675 char *old = csp->iob->cur, *new = NULL;
1678 struct file_list *fl;
1679 struct re_filterfile_spec *b;
1680 struct list_entry *filtername;
1682 int i, found_filters = 0;
1687 if (csp->iob->cur >= csp->iob->eod)
1691 size = (size_t)(csp->iob->eod - csp->iob->cur);
1694 * Need to check the set of re_filterfiles...
1696 for (i = 0; i < MAX_AF_FILES; i++)
1709 if (0 == found_filters)
1711 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1716 * If the body has a "chunked" transfer-encoding,
1717 * get rid of it first, adjusting size and iob->eod
1719 if (csp->flags & CSP_FLAG_CHUNKED)
1721 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1722 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1726 csp->iob->eod = csp->iob->cur + size;
1727 csp->flags |= CSP_FLAG_MODIFIED;
1730 for (i = 0; i < MAX_AF_FILES; i++)
1733 if ((NULL == fl) || (NULL == fl->f))
1736 * Either there are no filter files
1737 * left, or this filter file just
1738 * contains no valid filters.
1740 * Continue to be sure we don't miss
1741 * valid filter files that are chained
1742 * after empty or invalid ones.
1747 * For all applying +filter actions, look if a filter by that
1748 * name exists and if yes, execute it's pcrs_joblist on the
1751 for (b = fl->f; b; b = b->next)
1753 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1754 filtername ; filtername = filtername->next)
1756 if (strcmp(b->name, filtername->str) == 0)
1758 int current_hits = 0; /* Number of hits caused by this filter */
1759 int job_number = 0; /* Which job we're currently executing */
1760 int job_hits = 0; /* How many hits the current job caused */
1762 if ( NULL == b->joblist )
1764 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1769 /* Apply all jobs from the joblist */
1770 for (job = b->joblist; NULL != job; job = job->next)
1773 job_hits = pcrs_execute(job, old, size, &new, &size);
1778 * That went well. Continue filtering
1779 * and use the result of this job as
1780 * input for the next one.
1782 current_hits += job_hits;
1783 if (old != csp->iob->cur)
1792 * This job caused an unexpected error. Inform the user
1793 * and skip the rest of the jobs in this filter. We could
1794 * continue with the next job, but usually the jobs
1795 * depend on each other or are similar enough to
1796 * fail for the same reason.
1798 * At the moment our pcrs expects the error codes of pcre 3.4,
1799 * but newer pcre versions can return additional error codes.
1800 * As a result pcrs_strerror()'s error message might be
1801 * "Unknown error ...", therefore we print the numerical value
1804 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1805 * should we use LOG_LEVEL_RE_FILTER instead?
1807 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1808 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1813 log_error(LOG_LEVEL_RE_FILTER,
1814 "re_filtering %s%s (size %d) with filter %s produced %d hits (new size %d).",
1815 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1817 hits += current_hits;
1824 * If there were no hits, destroy our copy and let
1825 * chat() use the original in csp->iob
1833 csp->flags |= CSP_FLAG_MODIFIED;
1834 csp->content_length = size;
1842 /*********************************************************************
1844 * Function : gif_deanimate_response
1846 * Description : Deanimate the GIF image that has been accumulated in
1847 * csp->iob->buf, set csp->content_length to the modified
1848 * size and raise the CSP_FLAG_MODIFIED flag.
1851 * 1 : csp = Current client state (buffers, headers, etc...)
1853 * Returns : a pointer to the (newly allocated) modified buffer.
1854 * or NULL in case something went wrong.
1856 *********************************************************************/
1857 char *gif_deanimate_response(struct client_state *csp)
1859 struct binbuffer *in, *out;
1863 size = (size_t)(csp->iob->eod - csp->iob->cur);
1866 * If the body has a "chunked" transfer-encoding,
1867 * get rid of it first, adjusting size and iob->eod
1869 if (csp->flags & CSP_FLAG_CHUNKED)
1871 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1872 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1876 csp->iob->eod = csp->iob->cur + size;
1877 csp->flags |= CSP_FLAG_MODIFIED;
1880 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1881 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1883 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1887 in->buffer = csp->iob->cur;
1890 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1892 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1899 if ((int)size == out->offset)
1901 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1905 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1907 csp->content_length = out->offset;
1908 csp->flags |= CSP_FLAG_MODIFIED;
1918 /*********************************************************************
1920 * Function : jpeg_inspect_response
1925 * 1 : csp = Current client state (buffers, headers, etc...)
1927 * Returns : a pointer to the (newly allocated) modified buffer
1928 * or NULL in case something went wrong.
1930 *********************************************************************/
1931 char *jpeg_inspect_response(struct client_state *csp)
1933 struct binbuffer *in = NULL;
1934 struct binbuffer *out = NULL;
1938 size = (size_t)(csp->iob->eod - csp->iob->cur);
1941 * If the body has a "chunked" transfer-encoding,
1942 * get rid of it first, adjusting size and iob->eod
1944 if (csp->flags & CSP_FLAG_CHUNKED)
1946 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1947 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1951 csp->iob->eod = csp->iob->cur + size;
1952 csp->flags |= CSP_FLAG_MODIFIED;
1955 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1957 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
1961 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
1963 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
1967 in->buffer = csp->iob->cur;
1971 * Calling jpeg_inspect has the side-effect of creating and
1972 * modifying the image buffer of "out" directly.
1974 if (jpeg_inspect(in, out))
1976 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
1984 csp->content_length = out->offset;
1985 csp->flags |= CSP_FLAG_MODIFIED;
1995 /*********************************************************************
1997 * Function : remove_chunked_transfer_coding
1999 * Description : In-situ remove the "chunked" transfer coding as defined
2000 * in rfc2616 from a buffer.
2003 * 1 : buffer = Pointer to the text buffer
2004 * 2 : size = Number of bytes to be processed
2006 * Returns : The new size, i.e. the number of bytes from buffer which
2007 * are occupied by the stripped body, or 0 in case something
2010 *********************************************************************/
2011 size_t remove_chunked_transfer_coding(char *buffer, const size_t size)
2014 unsigned int chunksize = 0;
2015 char *from_p, *to_p;
2018 from_p = to_p = buffer;
2020 if (sscanf(buffer, "%x", &chunksize) != 1)
2022 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2026 while (chunksize > 0)
2028 if (NULL == (from_p = strstr(from_p, "\r\n")))
2030 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2034 if ((newsize += chunksize) >= size)
2036 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2041 memmove(to_p, from_p, (size_t) chunksize);
2042 to_p = buffer + newsize;
2043 from_p += chunksize + 2;
2045 if (sscanf(from_p, "%x", &chunksize) != 1)
2047 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2052 /* FIXME: Should this get its own loglevel? */
2053 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
2059 /*********************************************************************
2061 * Function : url_actions
2063 * Description : Gets the actions for this URL.
2066 * 1 : http = http_request request for blocked URLs
2067 * 2 : csp = Current client state (buffers, headers, etc...)
2071 *********************************************************************/
2072 void url_actions(struct http_request *http,
2073 struct client_state *csp)
2075 struct file_list *fl;
2076 struct url_actions *b;
2079 init_current_action(csp->action);
2081 for (i = 0; i < MAX_AF_FILES; i++)
2083 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2088 apply_url_actions(csp->action, http, b);
2095 /*********************************************************************
2097 * Function : apply_url_actions
2099 * Description : Applies a list of URL actions.
2102 * 1 : action = Destination.
2103 * 2 : http = Current URL
2104 * 3 : b = list of URL actions to apply
2108 *********************************************************************/
2109 void apply_url_actions(struct current_action_spec *action,
2110 struct http_request *http,
2111 struct url_actions *b)
2115 /* Should never happen */
2119 for (b = b->next; NULL != b; b = b->next)
2121 if (url_match(b->url, http))
2123 merge_current_action(action, b->action);
2129 /*********************************************************************
2131 * Function : forward_url
2133 * Description : Should we forward this to another proxy?
2136 * 1 : http = http_request request for current URL
2137 * 2 : csp = Current client state (buffers, headers, etc...)
2139 * Returns : Pointer to forwarding information.
2141 *********************************************************************/
2142 const struct forward_spec * forward_url(struct http_request *http,
2143 struct client_state *csp)
2145 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2146 struct forward_spec *fwd = csp->config->forward;
2155 if (url_match(fwd->url, http))
2166 /*********************************************************************
2168 * Function : direct_response
2170 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2171 * request and if so, return a HTTP 501 to the client.
2173 * FIXME: I have a stupid name and I should handle the
2174 * requests properly. Still, what we do here is rfc-
2175 * compliant, whereas ignoring or forwarding are not.
2178 * 1 : csp = Current client state (buffers, headers, etc...)
2180 * Returns : http_response if , NULL if nonmatch or handler fail
2182 *********************************************************************/
2183 struct http_response *direct_response(struct client_state *csp)
2185 struct http_response *rsp;
2186 struct list_entry *p;
2188 if ((0 == strcmpic(csp->http->gpc, "trace"))
2189 || (0 == strcmpic(csp->http->gpc, "options")))
2191 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2193 if (!strncmp("Max-Forwards:", p->str, 13)
2194 && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
2196 /* FIXME: We could handle at least TRACE here,
2197 but that would require a verbatim copy of
2198 the request which we don't have anymore */
2200 log_error(LOG_LEVEL_HEADER, "Found Max-Forwards:0 in OPTIONS or TRACE request -- Returning 501");
2202 /* Get mem for response or fail*/
2203 if (NULL == (rsp = alloc_http_response()))
2205 return cgi_error_memory();
2208 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2210 free_http_response(rsp);
2211 return cgi_error_memory();
2215 return(finish_http_response(rsp));