1 const char filters_rcs[] = "$Id: filters.c,v 1.80 2007/02/07 10:55:20 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, 2004-2007 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 1.80 2007/02/07 10:55:20 fabiankeil
44 * - Save the reason for generating http_responses.
45 * - Block (+block) with status code 403 instead of 404.
46 * - Use a different kludge to remember a failed decompression.
48 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
49 * Search for Max-Forwards headers case-insensitive,
50 * don't generate the "501 unsupported" message for invalid
51 * Max-Forwards values and don't increase negative ones.
53 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
54 * - Add HEAD support to finish_http_response.
55 * - Add error favicon to internal HTML error messages.
57 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
58 * Mark *csp as immutable for is_untrusted_url()
59 * and is_imageurl(). Closes FR 1237736.
61 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
62 * Integrate a modified version of Wil Mahan's
63 * zlib patch (PR #895531).
65 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
66 * Fixed gcc43 conversion warnings,
67 * changed sprintf calls to snprintf.
69 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
70 * Adjust comment in pcrs_filter_response()
71 * to recent pcrs changes. Hohoho.
73 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
74 * Don't crash if pcre returns an error code
75 * that pcrs didn't expect. Fixes BR 1621173.
77 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
78 * Modified is_untrusted_url to complain in case of
79 * write errors and to give a reason when adding new
80 * entries to the trustfile. Closes FR 1097611.
82 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
83 * Skip empty filter files in pcrs_filter_response,
84 * but don't ignore the ones that come afterwards.
85 * Fixes parts of BR 1619208.
87 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
88 * Added some sanity checks for get_last_url().
89 * Fixed possible segfault caused by my last commit.
91 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
92 * Let get_last_url() catch https URLs as well.
94 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
95 * Make sure get_last_url() behaves like advertised
96 * and fast-redirects{} can be combined with redirect{}.
98 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
99 * Implemented +redirect{s@foo@bar@} to generate
100 * a redirect based on a rewritten version of the
103 * Revision 1.66 2006/09/23 13:26:38 roro
104 * Replace TABs by spaces in source code.
106 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
107 * Fix +redirect{}. Didn't work with -fast-redirects.
109 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
110 * Block requests for untrusted URLs with status
111 * code 403 instead of 200.
113 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
114 * Don't free p which is still in use and will be later
115 * freed by free_map(). Don't claim the referrer is unknown
116 * when the client didn't set one.
118 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
119 * Feature request 595948: Re-Filter logging in single line
121 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
122 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
124 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
125 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
126 * with what was really the latest development (the v_3_0_branch branch)
128 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
129 * Multiple filter file support
131 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
132 * A few memory leaks plugged (mostly on error paths)
134 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
135 * Add the ability to check jpeg images for invalid
136 * lengths of comment blocks. Defensive strategy
137 * against the exploit:
138 * Microsoft Security Bulletin MS04-028
139 * Buffer Overrun in JPEG Processing (GDI+) Could
140 * Allow Code Execution (833987)
141 * Enabled with +inspect-jpegs in actions files.
143 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
144 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
146 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
147 * Fixed bug #839859: "See why" link URL now gets url-encoded.
149 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
152 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
153 * Added basic support for OPTIONS and TRACE HTTP methods:
154 * New function direct_response which handles OPTIONS and
155 * TRACE requests whose Max-Forwards header field is zero.
157 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
158 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
160 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
161 * - Bugfix: Executing a filters without jobs no longer results in
162 * turing off *all* filters.
163 * - Security fix: Malicious web servers can't cause a seg fault
164 * through bogus chunk sizes anymore
166 * Revision 1.58 2002/04/24 02:11:17 oes
167 * Jon's multiple AF patch: url_actions now evaluates rules
170 * Revision 1.57 2002/04/08 20:38:34 swa
173 * Revision 1.56 2002/04/05 15:51:24 oes
174 * - bugfix: error-pages now get correct request protocol
175 * - fix for invalid HTML in trust info
177 * Revision 1.55 2002/04/02 16:13:51 oes
178 * Fix: No "Go there anyway" for SSL
180 * Revision 1.54 2002/04/02 14:55:56 oes
181 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
183 * Revision 1.53 2002/03/26 22:29:54 swa
184 * we have a new homepage!
186 * Revision 1.52 2002/03/24 16:35:57 jongfoster
189 * Revision 1.51 2002/03/24 15:23:33 jongfoster
192 * Revision 1.50 2002/03/24 13:25:43 swa
193 * name change related issues
195 * Revision 1.49 2002/03/16 20:29:14 oes
198 * Revision 1.48 2002/03/13 20:25:34 oes
199 * Better logging for content filters
201 * Revision 1.47 2002/03/13 00:30:52 jongfoster
203 * Added option of always sending redirect for imageblock,
204 * currently disabled with #if 0.
206 * Revision 1.46 2002/03/12 01:42:49 oes
207 * Introduced modular filters
209 * Revision 1.45 2002/03/08 16:47:50 oes
210 * Added choice beween GIF and PNG built-in images
212 * Revision 1.44 2002/03/07 03:49:31 oes
213 * - Fixed compiler warnings etc
214 * - Changed built-in images from GIF to PNG
215 * (with regard to Unisys patent issue)
216 * - Added a 4x4 pattern PNG which is less intrusive
217 * than the logo but also clearly marks the deleted banners
219 * Revision 1.43 2002/01/22 23:51:59 jongfoster
220 * Replacing strsav() with the safer string_append().
222 * Adding missing html_encode() to error message generators. Where encoded
223 * and unencoded versions of a string were provided, removing the unencoded
226 * Revision 1.42 2002/01/17 21:00:32 jongfoster
227 * Moving all our URL and URL pattern parsing code to urlmatch.c.
229 * Using a single, simple url_match(pattern,url) function - rather than
230 * the 3-line match routine which was repeated all over the place.
232 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
234 * Using parse_http_url() to parse URLs without faking a HTTP
235 * request line for parse_http_request().
237 * Revision 1.41 2001/11/13 00:14:07 jongfoster
238 * Fixing stupid bug now I've figured out what || means.
239 * (It always returns 0 or 1, not one of it's paramaters.)
241 * Revision 1.40 2001/10/26 17:37:55 oes
242 * - Re-enabled Netscape 200/404 bug workaround in block_url():
243 * - Removed OS/2 special case
244 * - Made block_url() independant from sed() having been run
245 * - Made trust_url independant from sed() having been run
246 * - Made is_imageurl independant from sed() having been run.
247 * It now checks User-Agent: and Accept: by itself.
250 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
251 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
252 * threads to call select() simultaneously. So, it's time to do a real, live,
253 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
254 * (native). Both versions will work, but using __OS2__ offers multi-threading.
256 * Revision 1.38 2001/10/23 21:32:33 jongfoster
257 * Adding error-checking to selected functions
259 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
260 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
261 * filters.c. Added a FIXME in front of the offending code. I'll gladly
262 * put in a better/more robust fix for all parties if one is presented...
263 * It seems that just returning 200 instead of 404 would pretty much fix
264 * it for everyone, but I don't know all the history of the problem.
266 * Revision 1.36 2001/10/10 16:44:16 oes
267 * Added match_portlist function
269 * Revision 1.35 2001/10/07 15:41:23 oes
270 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
272 * New function remove_chunked_transfer_coding that strips chunked
273 * transfer coding to plain and is called by pcrs_filter_response
274 * and gif_deanimate_response if neccessary
276 * Improved handling of zero-change re_filter runs
278 * pcrs_filter_response and gif_deanimate_response now remove
279 * chunked transfer codeing before processing the body.
281 * Revision 1.34 2001/09/20 15:49:36 steudten
283 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
284 * See cgi.c fill_template().
286 * Revision 1.33 2001/09/16 17:05:14 jongfoster
287 * Removing unused #include showarg.h
289 * Revision 1.32 2001/09/16 13:21:27 jongfoster
290 * Changes to use new list functions.
292 * Revision 1.31 2001/09/16 11:38:02 jongfoster
293 * Splitting fill_template() into 2 functions:
294 * template_load() loads the file
295 * template_fill() performs the PCRS regexps.
296 * This is because the CGI edit interface has a "table row"
297 * template which is used many times in the page - this
298 * change means it's only loaded from disk once.
300 * Revision 1.30 2001/09/16 11:00:10 jongfoster
301 * New function alloc_http_response, for symmetry with free_http_response
303 * Revision 1.29 2001/09/13 23:32:40 jongfoster
304 * Moving image data to cgi.c rather than cgi.h
305 * Fixing a GPF under Win32 (and any other OS that protects global
306 * constants from being written to).
308 * Revision 1.28 2001/09/10 10:18:51 oes
309 * Silenced compiler warnings
311 * Revision 1.27 2001/08/05 16:06:20 jongfoster
312 * Modifiying "struct map" so that there are now separate header and
313 * "map_entry" structures. This means that functions which modify a
314 * map no longer need to return a pointer to the modified map.
315 * Also, it no longer reverses the order of the entries (which may be
316 * important with some advanced template substitutions).
318 * Revision 1.26 2001/07/30 22:08:36 jongfoster
319 * Tidying up #defines:
320 * - All feature #defines are now of the form FEATURE_xxx
321 * - Permanently turned off WIN_GUI_EDIT
322 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
324 * Revision 1.25 2001/07/26 10:09:46 oes
325 * Made browser detection a little less naive
327 * Revision 1.24 2001/07/25 17:22:51 oes
328 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
330 * Revision 1.23 2001/07/23 13:40:12 oes
331 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
333 * Revision 1.22 2001/07/18 12:29:34 oes
334 * - Made gif_deanimate_response respect
335 * csp->action->string[ACTION_STRING_DEANIMATE]
336 * - Logging cosmetics
338 * Revision 1.21 2001/07/13 13:59:53 oes
339 * - Introduced gif_deanimate_response which shares the
340 * generic content modification interface of pcrs_filter_response
341 * and acts as a wrapper to deanimate.c:gif_deanimate()
342 * - Renamed re_process_buffer to pcrs_filter_response
343 * - pcrs_filter_response now returns NULL on failiure
344 * - Removed all #ifdef PCRS
346 * Revision 1.20 2001/07/01 17:01:04 oes
347 * Added comments and missing return statement in is_untrusted_url()
349 * Revision 1.19 2001/06/29 21:45:41 oes
350 * Indentation, CRLF->LF, Tab-> Space
352 * Revision 1.18 2001/06/29 13:27:38 oes
353 * - Cleaned up, renamed and reorderd functions
354 * and improved comments
357 * - Ported to CGI platform. Now delivers
358 * http_response or NULL
359 * - Unified HTML and GIF generation (moved image detection
360 * and GIF generation here from jcc.c:chat())
361 * - Fixed HTTP status to:
362 * - 403 (Forbidden) for the "blocked" HTML message
363 * - 200 (OK) for GIF answers
364 * - 302 (Redirect) for redirect to GIF
367 * - Ported to CGI platform. Now delivers
368 * http_response or NULL
369 * - Separated detection of untrusted URL into
370 * (bool)is_untrusted_url
371 * - Added enforcement of untrusted requests
373 * - Moved redirect_url() from cgi.c to here
374 * and ported it to the CGI platform
376 * - Removed logentry from cancelled commit
378 * Revision 1.17 2001/06/09 10:55:28 jongfoster
379 * Changing BUFSIZ ==> BUFFER_SIZE
381 * Revision 1.16 2001/06/07 23:10:26 jongfoster
382 * Allowing unanchored domain patterns to back off and retry
383 * if they partially match. Optimized right-anchored patterns.
384 * Moving ACL and forward files into config file.
385 * Replacing struct gateway with struct forward_spec
387 * Revision 1.15 2001/06/03 19:12:00 oes
388 * extracted-CGI relevant stuff
390 * Revision 1.14 2001/06/01 10:30:55 oes
391 * Added optional left-anchoring to domaincmp
393 * Revision 1.13 2001/05/31 21:21:30 jongfoster
394 * Permissionsfile / actions file changes:
395 * - Changed "permission" to "action" throughout
396 * - changes to file format to allow string parameters
397 * - Moved helper functions to actions.c
399 * Revision 1.12 2001/05/31 17:35:20 oes
401 * - Enhanced domain part globbing with infix and prefix asterisk
402 * matching and optional unanchored operation
404 * Revision 1.11 2001/05/29 11:53:23 oes
405 * "See why" link added to "blocked" page
407 * Revision 1.10 2001/05/29 09:50:24 jongfoster
408 * Unified blocklist/imagelist/permissionslist.
409 * File format is still under discussion, but the internal changes
412 * Also modified interceptor behaviour:
413 * - We now intercept all URLs beginning with one of the following
414 * prefixes (and *only* these prefixes):
416 * * http://ijbswa.sf.net/config/
417 * * http://ijbswa.sourceforge.net/config/
418 * - New interceptors "home page" - go to http://i.j.b/ to see it.
419 * - Internal changes so that intercepted and fast redirect pages
420 * are not replaced with an image.
421 * - Interceptors now have the option to send a binary page direct
422 * to the client. (i.e. ijb-send-banner uses this)
423 * - Implemented show-url-info interceptor. (Which is why I needed
424 * the above interceptors changes - a typical URL is
425 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
426 * The previous mechanism would not have intercepted that, and
427 * if it had been intercepted then it then it would have replaced
430 * Revision 1.9 2001/05/27 22:17:04 oes
432 * - re_process_buffer no longer writes the modified buffer
433 * to the client, which was very ugly. It now returns the
434 * buffer, which it is then written by chat.
436 * - content_length now adjusts the Content-Length: header
437 * for modified documents rather than crunch()ing it.
438 * (Length info in csp->content_length, which is 0 for
439 * unmodified documents)
441 * - For this to work, sed() is called twice when filtering.
443 * Revision 1.8 2001/05/26 17:13:28 jongfoster
444 * Filled in a function comment.
446 * Revision 1.7 2001/05/26 15:26:15 jongfoster
447 * ACL feature now provides more security by immediately dropping
448 * connections from untrusted hosts.
450 * Revision 1.6 2001/05/26 00:28:36 jongfoster
451 * Automatic reloading of config file.
452 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
453 * Most of the global variables have been moved to a new
454 * struct configuration_spec, accessed through csp->config->globalname
455 * Most of the globals remaining are used by the Win32 GUI.
457 * Revision 1.5 2001/05/25 22:34:30 jongfoster
460 * Revision 1.4 2001/05/22 18:46:04 oes
462 * - Enabled filtering banners by size rather than URL
463 * by adding patterns that replace all standard banner
464 * sizes with the "Junkbuster" gif to the re_filterfile
466 * - Enabled filtering WebBugs by providing a pattern
467 * which kills all 1x1 images
469 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
470 * which is selected by the (nonstandard and therefore
471 * capital) letter 'U' in the option string.
472 * It causes the quantifiers to be ungreedy by default.
473 * Appending a ? turns back to greedy (!).
475 * - Added a new interceptor ijb-send-banner, which
476 * sends back the "Junkbuster" gif. Without imagelist or
477 * MSIE detection support, or if tinygif = 1, or the
478 * URL isn't recognized as an imageurl, a lame HTML
479 * explanation is sent instead.
481 * - Added new feature, which permits blocking remote
482 * script redirects and firing back a local redirect
484 * The feature is conditionally compiled, i.e. it
485 * can be disabled with --disable-fast-redirects,
486 * plus it must be activated by a "fast-redirects"
487 * line in the config file, has its own log level
488 * and of course wants to be displayed by show-proxy-args
489 * Note: Boy, all the #ifdefs in 1001 locations and
490 * all the fumbling with configure.in and acconfig.h
491 * were *way* more work than the feature itself :-(
493 * - Because a generic redirect template was needed for
494 * this, tinygif = 3 now uses the same.
496 * - Moved GIFs, and other static HTTP response templates
501 * - Removed some >400 CRs again (Jon, you really worked
504 * Revision 1.3 2001/05/20 16:44:47 jongfoster
505 * Removing last hardcoded Junkbusters.com URLs.
507 * Revision 1.2 2001/05/20 01:21:20 jongfoster
508 * Version 2.9.4 checkin.
509 * - Merged popupfile and cookiefile, and added control over PCRS
510 * filtering, in new "permissionsfile".
511 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
512 * file error you now get a message box (in the Win32 GUI) rather
513 * than the program exiting with no explanation.
514 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
516 * - Removed tabs from "config"
517 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
518 * - Bumped up version number.
520 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
521 * Initial import of version 2.9.3 source tree
524 *********************************************************************/
530 #include <sys/types.h>
539 #endif /* ndef __OS2__ */
540 #include <netinet/in.h>
542 #include <winsock2.h>
543 #endif /* ndef _WIN32 */
547 #endif /* def __OS2__ */
555 #include "jbsockets.h"
556 #include "miscutil.h"
560 #include "deanimate.h"
561 #include "urlmatch.h"
567 const char filters_h_rcs[] = FILTERS_H_VERSION;
569 /* Fix a problem with Solaris. There should be no effect on other
571 * Solaris's isspace() is a macro which uses it's argument directly
572 * as an array index. Therefore we need to make sure that high-bit
573 * characters generate +ve values, and ideally we also want to make
574 * the argument match the declared parameter type of "int".
576 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
580 /*********************************************************************
582 * Function : block_acl
584 * Description : Block this request?
585 * Decide yes or no based on ACL file.
588 * 1 : dst = The proxy or gateway address this is going to.
589 * Or NULL to check all possible targets.
590 * 2 : csp = Current client state (buffers, headers, etc...)
591 * Also includes the client IP address.
593 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
595 *********************************************************************/
596 int block_acl(struct access_control_addr *dst, struct client_state *csp)
598 struct access_control_list *acl = csp->config->acl;
600 /* if not using an access control list, then permit the connection */
606 /* search the list */
609 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
613 /* Just want to check if they have any access */
614 if (acl->action == ACL_PERMIT)
619 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
620 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
622 if (acl->action == ACL_PERMIT)
640 /*********************************************************************
642 * Function : acl_addr
644 * Description : Called from `load_config' to parse an ACL address.
647 * 1 : aspec = String specifying ACL address.
648 * 2 : aca = struct access_control_addr to fill in.
650 * Returns : 0 => Ok, everything else is an error.
652 *********************************************************************/
653 int acl_addr(char *aspec, struct access_control_addr *aca)
655 int i, masklength, port;
661 if ((p = strchr(aspec, '/')) != NULL)
665 if (ijb_isdigit(*p) == 0)
669 masklength = atoi(p);
672 if ((masklength < 0) || (masklength > 32))
677 if ((p = strchr(aspec, ':')) != NULL)
681 if (ijb_isdigit(*p) == 0)
690 aca->addr = ntohl(resolve_hostname_to_ip(aspec));
692 if (aca->addr == INADDR_NONE)
697 /* build the netmask */
699 for (i=1; i <= masklength ; i++)
701 aca->mask |= (1 << (32 - i));
704 /* now mask off the host portion of the ip address
705 * (i.e. save on the network portion of the address).
707 aca->addr = aca->addr & aca->mask;
712 #endif /* def FEATURE_ACL */
715 /*********************************************************************
717 * Function : match_portlist
719 * Description : Check if a given number is covered by a comma
720 * separated list of numbers and ranges (a,b-c,d,..)
723 * 1 : portlist = String with list
724 * 2 : port = port to check
726 * Returns : 0 => no match
729 *********************************************************************/
730 int match_portlist(const char *portlist, int port)
732 char *min, *max, *next, *portlist_copy;
734 min = next = portlist_copy = strdup(portlist);
737 * Zero-terminate first item and remember offset for next
739 if (NULL != (next = strchr(portlist_copy, (int) ',')))
745 * Loop through all items, checking for match
749 if (NULL == (max = strchr(min, (int) '-')))
752 * No dash, check for equality
754 if (port == atoi(min))
763 * This is a range, so check if between min and max,
764 * or, if max was omitted, between min and 65K
767 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
781 * Zero-terminate next item and remember offset for n+1
783 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
795 /*********************************************************************
797 * Function : block_url
799 * Description : Called from `chat'. Check to see if we need to block this.
802 * 1 : csp = Current client state (buffers, headers, etc...)
804 * Returns : NULL => unblocked, else HTTP block response
806 *********************************************************************/
807 struct http_response *block_url(struct client_state *csp)
809 struct http_response *rsp;
810 const char *new_content_type = NULL;
813 * If it's not blocked, don't block it ;-)
815 if ((csp->action->flags & ACTION_BLOCK) == 0)
819 if (csp->action->flags & ACTION_REDIRECT)
821 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
824 * Else, prepare a response
826 if (NULL == (rsp = alloc_http_response()))
828 return cgi_error_memory();
832 * If it's an image-url, send back an image or redirect
833 * as specified by the relevant +image action
835 #ifdef FEATURE_IMAGE_BLOCKING
836 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
840 /* determine HOW images should be blocked */
841 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
843 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
845 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
847 #if 1 /* Two alternative strategies, use this one for now: */
849 /* and handle accordingly: */
850 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
852 rsp->status = strdup("403 Request blocked by Privoxy");
853 if (rsp->status == NULL)
855 free_http_response(rsp);
856 return cgi_error_memory();
858 rsp->body = bindup(image_pattern_data, image_pattern_length);
859 if (rsp->body == NULL)
861 free_http_response(rsp);
862 return cgi_error_memory();
864 rsp->content_length = image_pattern_length;
866 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
868 free_http_response(rsp);
869 return cgi_error_memory();
873 else if (0 == strcmpic(p, "blank"))
875 rsp->status = strdup("403 Request blocked by Privoxy");
876 if (rsp->status == NULL)
878 free_http_response(rsp);
879 return cgi_error_memory();
881 rsp->body = bindup(image_blank_data, image_blank_length);
882 if (rsp->body == NULL)
884 free_http_response(rsp);
885 return cgi_error_memory();
887 rsp->content_length = image_blank_length;
889 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
891 free_http_response(rsp);
892 return cgi_error_memory();
898 rsp->status = strdup("302 Local Redirect from Privoxy");
899 if (rsp->status == NULL)
901 free_http_response(rsp);
902 return cgi_error_memory();
905 if (enlist_unique_header(rsp->headers, "Location", p))
907 free_http_response(rsp);
908 return cgi_error_memory();
912 #else /* Following code is disabled for now */
914 /* and handle accordingly: */
915 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
917 p = CGI_PREFIX "send-banner?type=pattern";
919 else if (0 == strcmpic(p, "blank"))
921 p = CGI_PREFIX "send-banner?type=blank";
923 rsp->status = strdup("302 Local Redirect from Privoxy");
924 if (rsp->status == NULL)
926 free_http_response(rsp);
927 return cgi_error_memory();
930 if (enlist_unique_header(rsp->headers, "Location", p))
932 free_http_response(rsp);
933 return cgi_error_memory();
935 #endif /* Preceeding code is disabled for now */
937 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
940 * Send empty document.
942 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
945 rsp->body = strdup(" ");
946 rsp->content_length = 1;
948 rsp->status = strdup("403 Request blocked by Privoxy");
949 if (rsp->status == NULL)
951 free_http_response(rsp);
952 return cgi_error_memory();
954 if (new_content_type != 0)
956 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
957 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
959 free_http_response(rsp);
960 return cgi_error_memory();
965 #endif /* def FEATURE_IMAGE_BLOCKING */
968 * Else, generate an HTML "blocked" message:
972 struct map * exports;
976 * Workaround for stupid Netscape bug which prevents
977 * pages from being displayed if loading a referenced
978 * JavaScript or style sheet fails. So make it appear
979 * as if it succeeded.
981 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
982 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
983 && !strstr(p, "Gecko") /* save Mozilla, */
984 && !strstr(p, "compatible") /* MSIE */
985 && !strstr(p, "Opera")) /* and Opera. */
987 rsp->status = strdup("200 Request for blocked URL");
991 rsp->status = strdup("403 Request for blocked URL");
994 if (rsp->status == NULL)
996 free_http_response(rsp);
997 return cgi_error_memory();
1000 exports = default_exports(csp, NULL);
1001 if (exports == NULL)
1003 free_http_response(rsp);
1004 return cgi_error_memory();
1007 #ifdef FEATURE_FORCE_LOAD
1008 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1009 if (csp->http->ssl != 0 || 0 == strcmpic(csp->http->gpc, "connect"))
1010 #endif /* ndef FEATURE_FORCE_LOAD */
1012 err = map_block_killer(exports, "force-support");
1015 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1016 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1017 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1018 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1023 free_http_response(rsp);
1024 return cgi_error_memory();
1027 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1030 free_http_response(rsp);
1031 return cgi_error_memory();
1034 rsp->reason = RSP_REASON_BLOCKED;
1036 return finish_http_response(csp, rsp);
1041 #ifdef FEATURE_TRUST
1042 /*********************************************************************
1044 * Function : trust_url FIXME: I should be called distrust_url
1046 * Description : Calls is_untrusted_url to determine if the URL is trusted
1047 * and if not, returns a HTTP 403 response with a reject message.
1050 * 1 : csp = Current client state (buffers, headers, etc...)
1052 * Returns : NULL => trusted, else http_response.
1054 *********************************************************************/
1055 struct http_response *trust_url(struct client_state *csp)
1057 struct http_response *rsp;
1058 struct map * exports;
1059 char buf[BUFFER_SIZE];
1061 struct url_spec **tl;
1066 * Don't bother to work on trusted URLs
1068 if (!is_untrusted_url(csp))
1074 * Else, prepare a response:
1076 if (NULL == (rsp = alloc_http_response()))
1078 return cgi_error_memory();
1081 rsp->status = strdup("403 Request blocked by Privoxy");
1082 exports = default_exports(csp, NULL);
1083 if (exports == NULL || rsp->status == NULL)
1085 free_http_response(rsp);
1086 return cgi_error_memory();
1090 * Export the protocol, host, port, and referrer information
1092 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1093 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1094 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1096 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1098 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1102 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1108 free_http_response(rsp);
1109 return cgi_error_memory();
1113 * Export the trust list
1116 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1118 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1119 string_append(&p, buf);
1121 err = map(exports, "trusted-referrers", 1, p, 0);
1126 free_http_response(rsp);
1127 return cgi_error_memory();
1131 * Export the trust info, if available
1133 if (csp->config->trust_info->first)
1135 struct list_entry *l;
1138 for (l = csp->config->trust_info->first; l ; l = l->next)
1140 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1141 string_append(&p, buf);
1143 err = map(exports, "trust-info", 1, p, 0);
1147 err = map_block_killer(exports, "have-trust-info");
1153 free_http_response(rsp);
1154 return cgi_error_memory();
1158 * Export the force prefix or the force conditional block killer
1160 #ifdef FEATURE_FORCE_LOAD
1161 if (0 == strcmpic(csp->http->gpc, "connect"))
1163 err = map_block_killer(exports, "force-support");
1167 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1169 #else /* ifndef FEATURE_FORCE_LOAD */
1170 err = map_block_killer(exports, "force-support");
1171 #endif /* ndef FEATURE_FORCE_LOAD */
1176 free_http_response(rsp);
1177 return cgi_error_memory();
1181 * Build the response
1183 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1186 free_http_response(rsp);
1187 return cgi_error_memory();
1189 rsp->reason = RSP_REASON_UNTRUSTED;
1191 return finish_http_response(csp, rsp);
1193 #endif /* def FEATURE_TRUST */
1195 /*********************************************************************
1197 * Function : execute_single_pcrs_command
1199 * Description : Apply single pcrs command to the subject.
1200 * The subject itself is left untouched, memory for the result
1201 * is malloc()ed and it is the caller's responsibility to free
1202 * the result when it's no longer needed.
1205 * 1 : subject = the subject (== original) string
1206 * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
1207 * 3 : hits = int* for returning the number of modifications
1209 * Returns : NULL in case of errors, otherwise the
1210 * result of the pcrs command.
1212 *********************************************************************/
1213 char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
1217 char *result = NULL;
1221 assert(pcrs_command);
1224 size = strlen(subject);
1226 if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
1228 log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
1229 pcrs_command, error);
1231 else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
1233 log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
1240 job = pcrs_free_job(job);
1247 /*********************************************************************
1249 * Function : rewrite_url
1251 * Description : Rewrites a URL with a single pcrs command
1252 * and returns the result if it differs from the
1253 * original and isn't obviously invalid.
1256 * 1 : old_url = URL to rewrite.
1257 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1260 * Returns : NULL if the pcrs_command didn't change the url, or
1261 * the result of the modification.
1263 *********************************************************************/
1264 char *rewrite_url(char *old_url, const char *pcrs_command)
1266 char *new_url = NULL;
1270 assert(pcrs_command);
1272 new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
1276 log_error(LOG_LEVEL_REDIRECTS,
1277 "pcrs command \"%s\" didn't change \"%s\".",
1278 pcrs_command, old_url, new_url);
1281 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1283 log_error(LOG_LEVEL_ERROR,
1284 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1285 "but the result doesn't look like a valid URL and will be ignored.",
1286 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1291 log_error(LOG_LEVEL_REDIRECTS,
1292 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1293 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1301 #ifdef FEATURE_FAST_REDIRECTS
1302 /*********************************************************************
1304 * Function : get_last_url
1306 * Description : Search for the last URL inside a string.
1307 * If the string already is a URL, it will
1308 * be the first URL found.
1311 * 1 : subject = the string to check
1312 * 2 : redirect_mode = +fast-redirect{} mode
1314 * Returns : NULL if no URL was found, or
1315 * the last URL found.
1317 *********************************************************************/
1318 char *get_last_url(char *subject, const char *redirect_mode)
1320 char *new_url = NULL;
1324 assert(redirect_mode);
1326 subject = strdup(subject);
1327 if (subject == NULL)
1329 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1333 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1335 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1336 new_url = url_decode(subject);
1337 if (new_url != NULL)
1344 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1348 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1351 * Find the last URL encoded in the request
1354 while ((tmp = strstr(tmp, "http://")) != NULL)
1358 tmp = (new_url != NULL) ? new_url : subject;
1359 while ((tmp = strstr(tmp, "https://")) != NULL)
1364 if ((new_url != NULL)
1365 && ( (new_url != subject)
1366 || (0 == strncmpic(subject, "http://", 7))
1367 || (0 == strncmpic(subject, "https://", 8))
1371 * Return new URL if we found a redirect
1372 * or if the subject already was a URL.
1374 * The second case makes sure that we can
1375 * chain get_last_url after another redirection check
1376 * (like rewrite_url) without losing earlier redirects.
1378 new_url = strdup(new_url);
1387 #endif /* def FEATURE_FAST_REDIRECTS */
1390 /*********************************************************************
1392 * Function : redirect_url
1394 * Description : Checks if Privoxy should answer the request with
1395 * a HTTP redirect and generates the redirect if
1399 * 1 : csp = Current client state (buffers, headers, etc...)
1401 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1403 *********************************************************************/
1404 struct http_response *redirect_url(struct client_state *csp)
1406 struct http_response *rsp;
1407 #ifdef FEATURE_FAST_REDIRECTS
1409 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1410 * as compile-time option? The user can easily disable
1411 * it in his action file.
1413 char * redirect_mode;
1414 #endif /* def FEATURE_FAST_REDIRECTS */
1415 char *old_url = NULL;
1416 char *new_url = NULL;
1417 char *redirection_string;
1419 if ((csp->action->flags & ACTION_REDIRECT))
1421 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1424 * If the redirection string begins with 's',
1425 * assume it's a pcrs command, otherwise treat it as
1426 * properly formatted URL and use it for the redirection
1429 * According to RFC 2616 section 14.30 the URL
1430 * has to be absolute and if the user tries:
1431 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1432 * she would get undefined results anyway.
1436 if (*redirection_string == 's')
1438 old_url = csp->http->url;
1439 new_url = rewrite_url(old_url, redirection_string);
1443 log_error(LOG_LEVEL_REDIRECTS,
1444 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1445 redirection_string);
1446 new_url = strdup(redirection_string);
1450 #ifdef FEATURE_FAST_REDIRECTS
1451 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1453 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1456 * If it exists, use the previously rewritten URL as input
1457 * otherwise just use the old path.
1459 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1460 new_url = get_last_url(old_url, redirect_mode);
1463 #endif /* def FEATURE_FAST_REDIRECTS */
1465 /* Did any redirect action trigger? */
1468 if (0 == strcmpic(new_url, csp->http->url))
1470 log_error(LOG_LEVEL_ERROR,
1471 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1472 csp->http->url, new_url);
1477 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1479 if (NULL == (rsp = alloc_http_response()))
1482 return cgi_error_memory();
1485 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1486 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1489 free_http_response(rsp);
1490 return cgi_error_memory();
1492 rsp->reason = RSP_REASON_REDIRECTED;
1495 return finish_http_response(csp, rsp);
1499 /* Only reached if no redirect is required */
1505 #ifdef FEATURE_IMAGE_BLOCKING
1506 /*********************************************************************
1508 * Function : is_imageurl
1510 * Description : Given a URL, decide whether it is an image or not,
1511 * using either the info from a previous +image action
1512 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1513 * is MSIE and not on a Mac, tell from the browser's accept
1517 * 1 : csp = Current client state (buffers, headers, etc...)
1519 * Returns : True (nonzero) if URL is an image, false (0)
1522 *********************************************************************/
1523 int is_imageurl(const struct client_state *csp)
1525 #ifdef FEATURE_IMAGE_DETECT_MSIE
1528 tmp = get_header_value(csp->headers, "User-Agent:");
1529 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1531 tmp = get_header_value(csp->headers, "Accept:");
1532 if (tmp && strstr(tmp, "image/gif"))
1534 /* Client will accept HTML. If this seems counterintuitive,
1544 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1546 return ((csp->action->flags & ACTION_IMAGE) != 0);
1549 #endif /* def FEATURE_IMAGE_BLOCKING */
1552 #ifdef FEATURE_TRUST
1553 /*********************************************************************
1555 * Function : is_untrusted_url
1557 * Description : Should we "distrust" this URL (and block it)?
1559 * Yes if it matches a line in the trustfile, or if the
1560 * referrer matches a line starting with "+" in the
1565 * 1 : csp = Current client state (buffers, headers, etc...)
1567 * Returns : 0 => trusted, 1 => untrusted
1569 *********************************************************************/
1570 int is_untrusted_url(const struct client_state *csp)
1572 struct file_list *fl;
1573 struct block_spec *b;
1574 struct url_spec **trusted_url;
1575 struct http_request rhttp[1];
1576 const char * referer;
1580 * If we don't have a trustlist, we trust everybody
1582 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1587 memset(rhttp, '\0', sizeof(*rhttp));
1590 * Do we trust the request URL itself?
1592 for (b = b->next; b ; b = b->next)
1594 if (url_match(b->url, csp->http))
1600 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1602 /* no referrer was supplied */
1608 * If not, do we maybe trust its referrer?
1610 err = parse_http_url(referer, rhttp, csp);
1616 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1618 if (url_match(*trusted_url, rhttp))
1620 /* if the URL's referrer is from a trusted referrer, then
1621 * add the target spec to the trustfile as an unblocked
1622 * domain and return 0 (which means it's OK).
1627 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1631 char * new_entry = strdup("~");
1633 string_append(&new_entry, csp->http->hostport);
1635 path = csp->http->path;
1636 if ( (path[0] == '/')
1638 && ((path_end = strchr(path + 2, '/')) != NULL))
1640 /* since this path points into a user's home space
1641 * be sure to include this spec in the trustfile.
1643 int path_len = path_end - path; /* save offset */
1644 path = strdup(path); /* Copy string */
1647 path_end = path + path_len; /* regenerate ptr to new buffer */
1648 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1650 string_join(&new_entry, path);
1654 * Give a reason for generating this entry.
1656 string_append(&new_entry, " # Trusted referrer was: ");
1657 string_append(&new_entry, referer);
1659 if (new_entry != NULL)
1661 if (-1 == fprintf(fp, "%s\n", new_entry))
1663 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1664 new_entry, csp->config->trustfile);
1670 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1671 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1678 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1679 csp->http->hostport, csp->config->trustfile);
1687 #endif /* def FEATURE_TRUST */
1690 /*********************************************************************
1692 * Function : pcrs_filter_response
1694 * Description : Execute all text substitutions from all applying
1695 * +filter actions on the text buffer that's been accumulated
1696 * in csp->iob->buf. If this changes the contents, set
1697 * csp->content_length to the modified size and raise the
1698 * CSP_FLAG_MODIFIED flag.
1700 * XXX: Currently pcrs_filter_response is also responsible
1701 * for dechunking and decompressing. Both should be
1702 * done in separate functions so other content modifiers
1703 * profit as well, even if pcrs filtering is disabled.
1706 * 1 : csp = Current client state (buffers, headers, etc...)
1708 * Returns : a pointer to the (newly allocated) modified buffer.
1709 * or NULL if there were no hits or something went wrong
1711 *********************************************************************/
1712 char *pcrs_filter_response(struct client_state *csp)
1715 size_t size, prev_size;
1717 char *old = csp->iob->cur, *new = NULL;
1720 struct file_list *fl;
1721 struct re_filterfile_spec *b;
1722 struct list_entry *filtername;
1724 int i, found_filters = 0;
1729 if (csp->iob->cur >= csp->iob->eod)
1733 size = (size_t)(csp->iob->eod - csp->iob->cur);
1736 * Need to check the set of re_filterfiles...
1738 for (i = 0; i < MAX_AF_FILES; i++)
1751 if (0 == found_filters)
1753 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1758 * If the body has a "chunked" transfer-encoding,
1759 * get rid of it first, adjusting size and iob->eod
1761 if (csp->flags & CSP_FLAG_CHUNKED)
1763 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1764 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1768 csp->iob->eod = csp->iob->cur + size;
1769 csp->flags |= CSP_FLAG_MODIFIED;
1774 * If the body has a compressed transfer-encoding,
1775 * uncompress it first, adjusting size and iob->eod.
1776 * Note that decompression occurs after de-chunking.
1778 if (csp->content_type & (CT_GZIP | CT_DEFLATE))
1780 /* Notice that we at least tried to decompress. */
1781 if (JB_ERR_OK != decompress_iob(csp))
1784 * We failed to decompress the data; there's no point
1785 * in continuing since we can't filter.
1787 * XXX: Actually the Accept-Encoding header may
1788 * just be incorrect in which case we could continue
1791 * Unset CT_GZIP and CT_DEFLATE to remember not
1792 * to modify the Content-Encoding header later.
1794 csp->content_type &= ~CT_GZIP;
1795 csp->content_type &= ~CT_DEFLATE;
1798 log_error(LOG_LEVEL_RE_FILTER, "Decompression successful");
1801 * Decompression gives us a completely new iob,
1802 * so we need to update.
1804 size = (size_t)(csp->iob->eod - csp->iob->cur);
1805 old = csp->iob->cur;
1807 csp->flags |= CSP_FLAG_MODIFIED;
1808 csp->content_type &= ~CT_TABOO;
1812 for (i = 0; i < MAX_AF_FILES; i++)
1815 if ((NULL == fl) || (NULL == fl->f))
1818 * Either there are no filter files
1819 * left, or this filter file just
1820 * contains no valid filters.
1822 * Continue to be sure we don't miss
1823 * valid filter files that are chained
1824 * after empty or invalid ones.
1829 * For all applying +filter actions, look if a filter by that
1830 * name exists and if yes, execute it's pcrs_joblist on the
1833 for (b = fl->f; b; b = b->next)
1835 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1836 filtername ; filtername = filtername->next)
1838 if (strcmp(b->name, filtername->str) == 0)
1840 int current_hits = 0; /* Number of hits caused by this filter */
1841 int job_number = 0; /* Which job we're currently executing */
1842 int job_hits = 0; /* How many hits the current job caused */
1844 if ( NULL == b->joblist )
1846 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1851 /* Apply all jobs from the joblist */
1852 for (job = b->joblist; NULL != job; job = job->next)
1855 job_hits = pcrs_execute(job, old, size, &new, &size);
1860 * That went well. Continue filtering
1861 * and use the result of this job as
1862 * input for the next one.
1864 current_hits += job_hits;
1865 if (old != csp->iob->cur)
1874 * This job caused an unexpected error. Inform the user
1875 * and skip the rest of the jobs in this filter. We could
1876 * continue with the next job, but usually the jobs
1877 * depend on each other or are similar enough to
1878 * fail for the same reason.
1880 * At the moment our pcrs expects the error codes of pcre 3.4,
1881 * but newer pcre versions can return additional error codes.
1882 * As a result pcrs_strerror()'s error message might be
1883 * "Unknown error ...", therefore we print the numerical value
1886 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1887 * should we use LOG_LEVEL_RE_FILTER instead?
1889 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1890 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1895 log_error(LOG_LEVEL_RE_FILTER,
1896 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
1897 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1899 hits += current_hits;
1906 * If there were no hits, destroy our copy and let
1907 * chat() use the original in csp->iob
1915 csp->flags |= CSP_FLAG_MODIFIED;
1916 csp->content_length = size;
1924 /*********************************************************************
1926 * Function : gif_deanimate_response
1928 * Description : Deanimate the GIF image that has been accumulated in
1929 * csp->iob->buf, set csp->content_length to the modified
1930 * size and raise the CSP_FLAG_MODIFIED flag.
1933 * 1 : csp = Current client state (buffers, headers, etc...)
1935 * Returns : a pointer to the (newly allocated) modified buffer.
1936 * or NULL in case something went wrong.
1938 *********************************************************************/
1939 char *gif_deanimate_response(struct client_state *csp)
1941 struct binbuffer *in, *out;
1945 size = (size_t)(csp->iob->eod - csp->iob->cur);
1948 * If the body has a "chunked" transfer-encoding,
1949 * get rid of it first, adjusting size and iob->eod
1951 if (csp->flags & CSP_FLAG_CHUNKED)
1953 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1954 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1958 csp->iob->eod = csp->iob->cur + size;
1959 csp->flags |= CSP_FLAG_MODIFIED;
1962 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1963 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1965 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1969 in->buffer = csp->iob->cur;
1972 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
1974 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
1981 if ((int)size == out->offset)
1983 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
1987 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
1989 csp->content_length = out->offset;
1990 csp->flags |= CSP_FLAG_MODIFIED;
2000 /*********************************************************************
2002 * Function : jpeg_inspect_response
2007 * 1 : csp = Current client state (buffers, headers, etc...)
2009 * Returns : a pointer to the (newly allocated) modified buffer
2010 * or NULL in case something went wrong.
2012 *********************************************************************/
2013 char *jpeg_inspect_response(struct client_state *csp)
2015 struct binbuffer *in = NULL;
2016 struct binbuffer *out = NULL;
2020 size = (size_t)(csp->iob->eod - csp->iob->cur);
2023 * If the body has a "chunked" transfer-encoding,
2024 * get rid of it first, adjusting size and iob->eod
2026 if (csp->flags & CSP_FLAG_CHUNKED)
2028 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2029 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2033 csp->iob->eod = csp->iob->cur + size;
2034 csp->flags |= CSP_FLAG_MODIFIED;
2037 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2039 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
2043 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
2045 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
2049 in->buffer = csp->iob->cur;
2053 * Calling jpeg_inspect has the side-effect of creating and
2054 * modifying the image buffer of "out" directly.
2056 if (jpeg_inspect(in, out))
2058 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
2066 csp->content_length = out->offset;
2067 csp->flags |= CSP_FLAG_MODIFIED;
2077 /*********************************************************************
2079 * Function : remove_chunked_transfer_coding
2081 * Description : In-situ remove the "chunked" transfer coding as defined
2082 * in rfc2616 from a buffer.
2085 * 1 : buffer = Pointer to the text buffer
2086 * 2 : size = Number of bytes to be processed
2088 * Returns : The new size, i.e. the number of bytes from buffer which
2089 * are occupied by the stripped body, or 0 in case something
2092 *********************************************************************/
2093 size_t remove_chunked_transfer_coding(char *buffer, const size_t size)
2096 unsigned int chunksize = 0;
2097 char *from_p, *to_p;
2100 from_p = to_p = buffer;
2102 if (sscanf(buffer, "%x", &chunksize) != 1)
2104 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2108 while (chunksize > 0)
2110 if (NULL == (from_p = strstr(from_p, "\r\n")))
2112 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2116 if ((newsize += chunksize) >= size)
2118 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2123 memmove(to_p, from_p, (size_t) chunksize);
2124 to_p = buffer + newsize;
2125 from_p += chunksize + 2;
2127 if (sscanf(from_p, "%x", &chunksize) != 1)
2129 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2134 /* FIXME: Should this get its own loglevel? */
2135 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
2141 /*********************************************************************
2143 * Function : url_actions
2145 * Description : Gets the actions for this URL.
2148 * 1 : http = http_request request for blocked URLs
2149 * 2 : csp = Current client state (buffers, headers, etc...)
2153 *********************************************************************/
2154 void url_actions(struct http_request *http,
2155 struct client_state *csp)
2157 struct file_list *fl;
2158 struct url_actions *b;
2161 init_current_action(csp->action);
2163 for (i = 0; i < MAX_AF_FILES; i++)
2165 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2170 apply_url_actions(csp->action, http, b);
2177 /*********************************************************************
2179 * Function : apply_url_actions
2181 * Description : Applies a list of URL actions.
2184 * 1 : action = Destination.
2185 * 2 : http = Current URL
2186 * 3 : b = list of URL actions to apply
2190 *********************************************************************/
2191 void apply_url_actions(struct current_action_spec *action,
2192 struct http_request *http,
2193 struct url_actions *b)
2197 /* Should never happen */
2201 for (b = b->next; NULL != b; b = b->next)
2203 if (url_match(b->url, http))
2205 merge_current_action(action, b->action);
2211 /*********************************************************************
2213 * Function : forward_url
2215 * Description : Should we forward this to another proxy?
2218 * 1 : http = http_request request for current URL
2219 * 2 : csp = Current client state (buffers, headers, etc...)
2221 * Returns : Pointer to forwarding information.
2223 *********************************************************************/
2224 const struct forward_spec * forward_url(struct http_request *http,
2225 struct client_state *csp)
2227 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2228 struct forward_spec *fwd = csp->config->forward;
2237 if (url_match(fwd->url, http))
2248 /*********************************************************************
2250 * Function : direct_response
2252 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2253 * request and if so, return a HTTP 501 to the client.
2255 * FIXME: I have a stupid name and I should handle the
2256 * requests properly. Still, what we do here is rfc-
2257 * compliant, whereas ignoring or forwarding are not.
2260 * 1 : csp = Current client state (buffers, headers, etc...)
2262 * Returns : http_response if , NULL if nonmatch or handler fail
2264 *********************************************************************/
2265 struct http_response *direct_response(struct client_state *csp)
2267 struct http_response *rsp;
2268 struct list_entry *p;
2270 if ((0 == strcmpic(csp->http->gpc, "trace"))
2271 || (0 == strcmpic(csp->http->gpc, "options")))
2273 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2275 if (!strncmpic("Max-Forwards:", p->str, 13))
2277 unsigned int max_forwards;
2280 * If it's a Max-Forwards value of zero,
2281 * we have to intercept the request.
2283 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2286 * FIXME: We could handle at least TRACE here,
2287 * but that would require a verbatim copy of
2288 * the request which we don't have anymore
2290 log_error(LOG_LEVEL_HEADER,
2291 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2294 /* Get mem for response or fail*/
2295 if (NULL == (rsp = alloc_http_response()))
2297 return cgi_error_memory();
2300 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2302 free_http_response(rsp);
2303 return cgi_error_memory();
2307 rsp->reason = RSP_REASON_UNSUPPORTED;
2309 return(finish_http_response(csp, rsp));