1 const char filters_rcs[] = "$Id: filters.c,v 1.81 2007/03/05 14:40:53 fabiankeil Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/filters.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
7 * Functions declared include:
8 * `acl_addr', `add_stats', `block_acl', `block_imageurl',
9 * `block_url', `url_actions', `domain_split',
10 * `filter_popups', `forward_url', 'redirect_url',
11 * `ij_untrusted_url', `intercept_url', `pcrs_filter_respose',
12 * `ijb_send_banner', `trust_url', `gif_deanimate_response',
13 * `jpeg_inspect_response', `execute_single_pcrs_command',
14 * `rewrite_url', `get_last_url'
16 * Copyright : Written by and Copyright (C) 2001, 2004-2007 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 1.81 2007/03/05 14:40:53 fabiankeil
44 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
45 * - Hide the "Go there anyway" link for blocked CONNECT
46 * requests where going there anyway doesn't work anyway.
48 * Revision 1.80 2007/02/07 10:55:20 fabiankeil
49 * - Save the reason for generating http_responses.
50 * - Block (+block) with status code 403 instead of 404.
51 * - Use a different kludge to remember a failed decompression.
53 * Revision 1.79 2007/01/31 16:21:38 fabiankeil
54 * Search for Max-Forwards headers case-insensitive,
55 * don't generate the "501 unsupported" message for invalid
56 * Max-Forwards values and don't increase negative ones.
58 * Revision 1.78 2007/01/28 13:41:18 fabiankeil
59 * - Add HEAD support to finish_http_response.
60 * - Add error favicon to internal HTML error messages.
62 * Revision 1.77 2007/01/12 15:36:44 fabiankeil
63 * Mark *csp as immutable for is_untrusted_url()
64 * and is_imageurl(). Closes FR 1237736.
66 * Revision 1.76 2007/01/01 19:36:37 fabiankeil
67 * Integrate a modified version of Wil Mahan's
68 * zlib patch (PR #895531).
70 * Revision 1.75 2006/12/29 18:30:46 fabiankeil
71 * Fixed gcc43 conversion warnings,
72 * changed sprintf calls to snprintf.
74 * Revision 1.74 2006/12/24 17:37:38 fabiankeil
75 * Adjust comment in pcrs_filter_response()
76 * to recent pcrs changes. Hohoho.
78 * Revision 1.73 2006/12/23 16:01:02 fabiankeil
79 * Don't crash if pcre returns an error code
80 * that pcrs didn't expect. Fixes BR 1621173.
82 * Revision 1.72 2006/12/22 18:52:53 fabiankeil
83 * Modified is_untrusted_url to complain in case of
84 * write errors and to give a reason when adding new
85 * entries to the trustfile. Closes FR 1097611.
87 * Revision 1.71 2006/12/22 14:24:52 fabiankeil
88 * Skip empty filter files in pcrs_filter_response,
89 * but don't ignore the ones that come afterwards.
90 * Fixes parts of BR 1619208.
92 * Revision 1.70 2006/12/09 13:33:15 fabiankeil
93 * Added some sanity checks for get_last_url().
94 * Fixed possible segfault caused by my last commit.
96 * Revision 1.69 2006/12/08 12:39:13 fabiankeil
97 * Let get_last_url() catch https URLs as well.
99 * Revision 1.68 2006/12/05 14:45:48 fabiankeil
100 * Make sure get_last_url() behaves like advertised
101 * and fast-redirects{} can be combined with redirect{}.
103 * Revision 1.67 2006/11/28 15:19:43 fabiankeil
104 * Implemented +redirect{s@foo@bar@} to generate
105 * a redirect based on a rewritten version of the
108 * Revision 1.66 2006/09/23 13:26:38 roro
109 * Replace TABs by spaces in source code.
111 * Revision 1.65 2006/09/21 12:54:43 fabiankeil
112 * Fix +redirect{}. Didn't work with -fast-redirects.
114 * Revision 1.64 2006/08/31 10:55:49 fabiankeil
115 * Block requests for untrusted URLs with status
116 * code 403 instead of 200.
118 * Revision 1.63 2006/08/31 10:11:28 fabiankeil
119 * Don't free p which is still in use and will be later
120 * freed by free_map(). Don't claim the referrer is unknown
121 * when the client didn't set one.
123 * Revision 1.62 2006/08/14 00:27:47 david__schmidt
124 * Feature request 595948: Re-Filter logging in single line
126 * Revision 1.61 2006/08/03 02:46:41 david__schmidt
127 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
129 * Revision 1.60 2006/07/18 14:48:46 david__schmidt
130 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
131 * with what was really the latest development (the v_3_0_branch branch)
133 * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
134 * Multiple filter file support
136 * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
137 * A few memory leaks plugged (mostly on error paths)
139 * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
140 * Add the ability to check jpeg images for invalid
141 * lengths of comment blocks. Defensive strategy
142 * against the exploit:
143 * Microsoft Security Bulletin MS04-028
144 * Buffer Overrun in JPEG Processing (GDI+) Could
145 * Allow Code Execution (833987)
146 * Enabled with +inspect-jpegs in actions files.
148 * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
149 * Correcting compile problem with FEATURE_IMAGE_BLOCKING
151 * Revision 1.58.2.5 2003/11/11 13:10:31 oes
152 * Fixed bug #839859: "See why" link URL now gets url-encoded.
154 * Revision 1.58.2.4 2003/02/28 12:52:45 oes
157 * Revision 1.58.2.3 2002/09/25 14:51:51 oes
158 * Added basic support for OPTIONS and TRACE HTTP methods:
159 * New function direct_response which handles OPTIONS and
160 * TRACE requests whose Max-Forwards header field is zero.
162 * Revision 1.58.2.2 2002/08/01 17:18:28 oes
163 * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
165 * Revision 1.58.2.1 2002/07/26 15:18:53 oes
166 * - Bugfix: Executing a filters without jobs no longer results in
167 * turing off *all* filters.
168 * - Security fix: Malicious web servers can't cause a seg fault
169 * through bogus chunk sizes anymore
171 * Revision 1.58 2002/04/24 02:11:17 oes
172 * Jon's multiple AF patch: url_actions now evaluates rules
175 * Revision 1.57 2002/04/08 20:38:34 swa
178 * Revision 1.56 2002/04/05 15:51:24 oes
179 * - bugfix: error-pages now get correct request protocol
180 * - fix for invalid HTML in trust info
182 * Revision 1.55 2002/04/02 16:13:51 oes
183 * Fix: No "Go there anyway" for SSL
185 * Revision 1.54 2002/04/02 14:55:56 oes
186 * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
188 * Revision 1.53 2002/03/26 22:29:54 swa
189 * we have a new homepage!
191 * Revision 1.52 2002/03/24 16:35:57 jongfoster
194 * Revision 1.51 2002/03/24 15:23:33 jongfoster
197 * Revision 1.50 2002/03/24 13:25:43 swa
198 * name change related issues
200 * Revision 1.49 2002/03/16 20:29:14 oes
203 * Revision 1.48 2002/03/13 20:25:34 oes
204 * Better logging for content filters
206 * Revision 1.47 2002/03/13 00:30:52 jongfoster
208 * Added option of always sending redirect for imageblock,
209 * currently disabled with #if 0.
211 * Revision 1.46 2002/03/12 01:42:49 oes
212 * Introduced modular filters
214 * Revision 1.45 2002/03/08 16:47:50 oes
215 * Added choice beween GIF and PNG built-in images
217 * Revision 1.44 2002/03/07 03:49:31 oes
218 * - Fixed compiler warnings etc
219 * - Changed built-in images from GIF to PNG
220 * (with regard to Unisys patent issue)
221 * - Added a 4x4 pattern PNG which is less intrusive
222 * than the logo but also clearly marks the deleted banners
224 * Revision 1.43 2002/01/22 23:51:59 jongfoster
225 * Replacing strsav() with the safer string_append().
227 * Adding missing html_encode() to error message generators. Where encoded
228 * and unencoded versions of a string were provided, removing the unencoded
231 * Revision 1.42 2002/01/17 21:00:32 jongfoster
232 * Moving all our URL and URL pattern parsing code to urlmatch.c.
234 * Using a single, simple url_match(pattern,url) function - rather than
235 * the 3-line match routine which was repeated all over the place.
237 * Renaming free_url to free_url_spec, since it frees a struct url_spec.
239 * Using parse_http_url() to parse URLs without faking a HTTP
240 * request line for parse_http_request().
242 * Revision 1.41 2001/11/13 00:14:07 jongfoster
243 * Fixing stupid bug now I've figured out what || means.
244 * (It always returns 0 or 1, not one of it's paramaters.)
246 * Revision 1.40 2001/10/26 17:37:55 oes
247 * - Re-enabled Netscape 200/404 bug workaround in block_url():
248 * - Removed OS/2 special case
249 * - Made block_url() independant from sed() having been run
250 * - Made trust_url independant from sed() having been run
251 * - Made is_imageurl independant from sed() having been run.
252 * It now checks User-Agent: and Accept: by itself.
255 * Revision 1.39 2001/10/25 03:40:48 david__schmidt
256 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
257 * threads to call select() simultaneously. So, it's time to do a real, live,
258 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
259 * (native). Both versions will work, but using __OS2__ offers multi-threading.
261 * Revision 1.38 2001/10/23 21:32:33 jongfoster
262 * Adding error-checking to selected functions
264 * Revision 1.37 2001/10/22 15:33:56 david__schmidt
265 * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
266 * filters.c. Added a FIXME in front of the offending code. I'll gladly
267 * put in a better/more robust fix for all parties if one is presented...
268 * It seems that just returning 200 instead of 404 would pretty much fix
269 * it for everyone, but I don't know all the history of the problem.
271 * Revision 1.36 2001/10/10 16:44:16 oes
272 * Added match_portlist function
274 * Revision 1.35 2001/10/07 15:41:23 oes
275 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
277 * New function remove_chunked_transfer_coding that strips chunked
278 * transfer coding to plain and is called by pcrs_filter_response
279 * and gif_deanimate_response if neccessary
281 * Improved handling of zero-change re_filter runs
283 * pcrs_filter_response and gif_deanimate_response now remove
284 * chunked transfer codeing before processing the body.
286 * Revision 1.34 2001/09/20 15:49:36 steudten
288 * Fix BUG: Change int size to size_t size in pcrs_filter_response().
289 * See cgi.c fill_template().
291 * Revision 1.33 2001/09/16 17:05:14 jongfoster
292 * Removing unused #include showarg.h
294 * Revision 1.32 2001/09/16 13:21:27 jongfoster
295 * Changes to use new list functions.
297 * Revision 1.31 2001/09/16 11:38:02 jongfoster
298 * Splitting fill_template() into 2 functions:
299 * template_load() loads the file
300 * template_fill() performs the PCRS regexps.
301 * This is because the CGI edit interface has a "table row"
302 * template which is used many times in the page - this
303 * change means it's only loaded from disk once.
305 * Revision 1.30 2001/09/16 11:00:10 jongfoster
306 * New function alloc_http_response, for symmetry with free_http_response
308 * Revision 1.29 2001/09/13 23:32:40 jongfoster
309 * Moving image data to cgi.c rather than cgi.h
310 * Fixing a GPF under Win32 (and any other OS that protects global
311 * constants from being written to).
313 * Revision 1.28 2001/09/10 10:18:51 oes
314 * Silenced compiler warnings
316 * Revision 1.27 2001/08/05 16:06:20 jongfoster
317 * Modifiying "struct map" so that there are now separate header and
318 * "map_entry" structures. This means that functions which modify a
319 * map no longer need to return a pointer to the modified map.
320 * Also, it no longer reverses the order of the entries (which may be
321 * important with some advanced template substitutions).
323 * Revision 1.26 2001/07/30 22:08:36 jongfoster
324 * Tidying up #defines:
325 * - All feature #defines are now of the form FEATURE_xxx
326 * - Permanently turned off WIN_GUI_EDIT
327 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
329 * Revision 1.25 2001/07/26 10:09:46 oes
330 * Made browser detection a little less naive
332 * Revision 1.24 2001/07/25 17:22:51 oes
333 * Added workaround for Netscape bug that prevents display of page when loading a component fails.
335 * Revision 1.23 2001/07/23 13:40:12 oes
336 * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
338 * Revision 1.22 2001/07/18 12:29:34 oes
339 * - Made gif_deanimate_response respect
340 * csp->action->string[ACTION_STRING_DEANIMATE]
341 * - Logging cosmetics
343 * Revision 1.21 2001/07/13 13:59:53 oes
344 * - Introduced gif_deanimate_response which shares the
345 * generic content modification interface of pcrs_filter_response
346 * and acts as a wrapper to deanimate.c:gif_deanimate()
347 * - Renamed re_process_buffer to pcrs_filter_response
348 * - pcrs_filter_response now returns NULL on failiure
349 * - Removed all #ifdef PCRS
351 * Revision 1.20 2001/07/01 17:01:04 oes
352 * Added comments and missing return statement in is_untrusted_url()
354 * Revision 1.19 2001/06/29 21:45:41 oes
355 * Indentation, CRLF->LF, Tab-> Space
357 * Revision 1.18 2001/06/29 13:27:38 oes
358 * - Cleaned up, renamed and reorderd functions
359 * and improved comments
362 * - Ported to CGI platform. Now delivers
363 * http_response or NULL
364 * - Unified HTML and GIF generation (moved image detection
365 * and GIF generation here from jcc.c:chat())
366 * - Fixed HTTP status to:
367 * - 403 (Forbidden) for the "blocked" HTML message
368 * - 200 (OK) for GIF answers
369 * - 302 (Redirect) for redirect to GIF
372 * - Ported to CGI platform. Now delivers
373 * http_response or NULL
374 * - Separated detection of untrusted URL into
375 * (bool)is_untrusted_url
376 * - Added enforcement of untrusted requests
378 * - Moved redirect_url() from cgi.c to here
379 * and ported it to the CGI platform
381 * - Removed logentry from cancelled commit
383 * Revision 1.17 2001/06/09 10:55:28 jongfoster
384 * Changing BUFSIZ ==> BUFFER_SIZE
386 * Revision 1.16 2001/06/07 23:10:26 jongfoster
387 * Allowing unanchored domain patterns to back off and retry
388 * if they partially match. Optimized right-anchored patterns.
389 * Moving ACL and forward files into config file.
390 * Replacing struct gateway with struct forward_spec
392 * Revision 1.15 2001/06/03 19:12:00 oes
393 * extracted-CGI relevant stuff
395 * Revision 1.14 2001/06/01 10:30:55 oes
396 * Added optional left-anchoring to domaincmp
398 * Revision 1.13 2001/05/31 21:21:30 jongfoster
399 * Permissionsfile / actions file changes:
400 * - Changed "permission" to "action" throughout
401 * - changes to file format to allow string parameters
402 * - Moved helper functions to actions.c
404 * Revision 1.12 2001/05/31 17:35:20 oes
406 * - Enhanced domain part globbing with infix and prefix asterisk
407 * matching and optional unanchored operation
409 * Revision 1.11 2001/05/29 11:53:23 oes
410 * "See why" link added to "blocked" page
412 * Revision 1.10 2001/05/29 09:50:24 jongfoster
413 * Unified blocklist/imagelist/permissionslist.
414 * File format is still under discussion, but the internal changes
417 * Also modified interceptor behaviour:
418 * - We now intercept all URLs beginning with one of the following
419 * prefixes (and *only* these prefixes):
421 * * http://ijbswa.sf.net/config/
422 * * http://ijbswa.sourceforge.net/config/
423 * - New interceptors "home page" - go to http://i.j.b/ to see it.
424 * - Internal changes so that intercepted and fast redirect pages
425 * are not replaced with an image.
426 * - Interceptors now have the option to send a binary page direct
427 * to the client. (i.e. ijb-send-banner uses this)
428 * - Implemented show-url-info interceptor. (Which is why I needed
429 * the above interceptors changes - a typical URL is
430 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
431 * The previous mechanism would not have intercepted that, and
432 * if it had been intercepted then it then it would have replaced
435 * Revision 1.9 2001/05/27 22:17:04 oes
437 * - re_process_buffer no longer writes the modified buffer
438 * to the client, which was very ugly. It now returns the
439 * buffer, which it is then written by chat.
441 * - content_length now adjusts the Content-Length: header
442 * for modified documents rather than crunch()ing it.
443 * (Length info in csp->content_length, which is 0 for
444 * unmodified documents)
446 * - For this to work, sed() is called twice when filtering.
448 * Revision 1.8 2001/05/26 17:13:28 jongfoster
449 * Filled in a function comment.
451 * Revision 1.7 2001/05/26 15:26:15 jongfoster
452 * ACL feature now provides more security by immediately dropping
453 * connections from untrusted hosts.
455 * Revision 1.6 2001/05/26 00:28:36 jongfoster
456 * Automatic reloading of config file.
457 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
458 * Most of the global variables have been moved to a new
459 * struct configuration_spec, accessed through csp->config->globalname
460 * Most of the globals remaining are used by the Win32 GUI.
462 * Revision 1.5 2001/05/25 22:34:30 jongfoster
465 * Revision 1.4 2001/05/22 18:46:04 oes
467 * - Enabled filtering banners by size rather than URL
468 * by adding patterns that replace all standard banner
469 * sizes with the "Junkbuster" gif to the re_filterfile
471 * - Enabled filtering WebBugs by providing a pattern
472 * which kills all 1x1 images
474 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
475 * which is selected by the (nonstandard and therefore
476 * capital) letter 'U' in the option string.
477 * It causes the quantifiers to be ungreedy by default.
478 * Appending a ? turns back to greedy (!).
480 * - Added a new interceptor ijb-send-banner, which
481 * sends back the "Junkbuster" gif. Without imagelist or
482 * MSIE detection support, or if tinygif = 1, or the
483 * URL isn't recognized as an imageurl, a lame HTML
484 * explanation is sent instead.
486 * - Added new feature, which permits blocking remote
487 * script redirects and firing back a local redirect
489 * The feature is conditionally compiled, i.e. it
490 * can be disabled with --disable-fast-redirects,
491 * plus it must be activated by a "fast-redirects"
492 * line in the config file, has its own log level
493 * and of course wants to be displayed by show-proxy-args
494 * Note: Boy, all the #ifdefs in 1001 locations and
495 * all the fumbling with configure.in and acconfig.h
496 * were *way* more work than the feature itself :-(
498 * - Because a generic redirect template was needed for
499 * this, tinygif = 3 now uses the same.
501 * - Moved GIFs, and other static HTTP response templates
506 * - Removed some >400 CRs again (Jon, you really worked
509 * Revision 1.3 2001/05/20 16:44:47 jongfoster
510 * Removing last hardcoded Junkbusters.com URLs.
512 * Revision 1.2 2001/05/20 01:21:20 jongfoster
513 * Version 2.9.4 checkin.
514 * - Merged popupfile and cookiefile, and added control over PCRS
515 * filtering, in new "permissionsfile".
516 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
517 * file error you now get a message box (in the Win32 GUI) rather
518 * than the program exiting with no explanation.
519 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
521 * - Removed tabs from "config"
522 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
523 * - Bumped up version number.
525 * Revision 1.1.1.1 2001/05/15 13:58:52 oes
526 * Initial import of version 2.9.3 source tree
529 *********************************************************************/
535 #include <sys/types.h>
544 #endif /* ndef __OS2__ */
545 #include <netinet/in.h>
547 #include <winsock2.h>
548 #endif /* ndef _WIN32 */
552 #endif /* def __OS2__ */
560 #include "jbsockets.h"
561 #include "miscutil.h"
565 #include "deanimate.h"
566 #include "urlmatch.h"
572 const char filters_h_rcs[] = FILTERS_H_VERSION;
574 /* Fix a problem with Solaris. There should be no effect on other
576 * Solaris's isspace() is a macro which uses it's argument directly
577 * as an array index. Therefore we need to make sure that high-bit
578 * characters generate +ve values, and ideally we also want to make
579 * the argument match the declared parameter type of "int".
581 #define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
585 /*********************************************************************
587 * Function : block_acl
589 * Description : Block this request?
590 * Decide yes or no based on ACL file.
593 * 1 : dst = The proxy or gateway address this is going to.
594 * Or NULL to check all possible targets.
595 * 2 : csp = Current client state (buffers, headers, etc...)
596 * Also includes the client IP address.
598 * Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
600 *********************************************************************/
601 int block_acl(struct access_control_addr *dst, struct client_state *csp)
603 struct access_control_list *acl = csp->config->acl;
605 /* if not using an access control list, then permit the connection */
611 /* search the list */
614 if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
618 /* Just want to check if they have any access */
619 if (acl->action == ACL_PERMIT)
624 else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
625 && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
627 if (acl->action == ACL_PERMIT)
645 /*********************************************************************
647 * Function : acl_addr
649 * Description : Called from `load_config' to parse an ACL address.
652 * 1 : aspec = String specifying ACL address.
653 * 2 : aca = struct access_control_addr to fill in.
655 * Returns : 0 => Ok, everything else is an error.
657 *********************************************************************/
658 int acl_addr(const char *aspec, struct access_control_addr *aca)
663 char *acl_spec = NULL;
669 * Use a temporary acl spec copy so we can log
670 * the unmodified original in case of parse errors.
672 acl_spec = strdup(aspec);
673 if (acl_spec == NULL)
675 /* XXX: This will be logged as parse error. */
679 if ((p = strchr(acl_spec, '/')) != NULL)
682 if (ijb_isdigit(*p) == 0)
687 masklength = atoi(p);
690 if ((masklength < 0) || (masklength > 32))
696 if ((p = strchr(acl_spec, ':')) != NULL)
701 port = strtol(p, &endptr, 10);
703 if (port <= 0 || port > 65535 || *endptr != '\0')
710 aca->port = (unsigned long)port;
712 aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
715 if (aca->addr == INADDR_NONE)
717 /* XXX: This will be logged as parse error. */
721 /* build the netmask */
723 for (i=1; i <= masklength ; i++)
725 aca->mask |= (1 << (32 - i));
728 /* now mask off the host portion of the ip address
729 * (i.e. save on the network portion of the address).
731 aca->addr = aca->addr & aca->mask;
736 #endif /* def FEATURE_ACL */
739 /*********************************************************************
741 * Function : match_portlist
743 * Description : Check if a given number is covered by a comma
744 * separated list of numbers and ranges (a,b-c,d,..)
747 * 1 : portlist = String with list
748 * 2 : port = port to check
750 * Returns : 0 => no match
753 *********************************************************************/
754 int match_portlist(const char *portlist, int port)
756 char *min, *max, *next, *portlist_copy;
758 min = next = portlist_copy = strdup(portlist);
761 * Zero-terminate first item and remember offset for next
763 if (NULL != (next = strchr(portlist_copy, (int) ',')))
769 * Loop through all items, checking for match
773 if (NULL == (max = strchr(min, (int) '-')))
776 * No dash, check for equality
778 if (port == atoi(min))
787 * This is a range, so check if between min and max,
788 * or, if max was omitted, between min and 65K
791 if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
805 * Zero-terminate next item and remember offset for n+1
807 if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
819 /*********************************************************************
821 * Function : block_url
823 * Description : Called from `chat'. Check to see if we need to block this.
826 * 1 : csp = Current client state (buffers, headers, etc...)
828 * Returns : NULL => unblocked, else HTTP block response
830 *********************************************************************/
831 struct http_response *block_url(struct client_state *csp)
833 struct http_response *rsp;
834 const char *new_content_type = NULL;
837 * If it's not blocked, don't block it ;-)
839 if ((csp->action->flags & ACTION_BLOCK) == 0)
843 if (csp->action->flags & ACTION_REDIRECT)
845 log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
848 * Else, prepare a response
850 if (NULL == (rsp = alloc_http_response()))
852 return cgi_error_memory();
856 * If it's an image-url, send back an image or redirect
857 * as specified by the relevant +image action
859 #ifdef FEATURE_IMAGE_BLOCKING
860 if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
864 /* determine HOW images should be blocked */
865 p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
867 if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
869 log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
871 #if 1 /* Two alternative strategies, use this one for now: */
873 /* and handle accordingly: */
874 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
876 rsp->status = strdup("403 Request blocked by Privoxy");
877 if (rsp->status == NULL)
879 free_http_response(rsp);
880 return cgi_error_memory();
882 rsp->body = bindup(image_pattern_data, image_pattern_length);
883 if (rsp->body == NULL)
885 free_http_response(rsp);
886 return cgi_error_memory();
888 rsp->content_length = image_pattern_length;
890 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
892 free_http_response(rsp);
893 return cgi_error_memory();
897 else if (0 == strcmpic(p, "blank"))
899 rsp->status = strdup("403 Request blocked by Privoxy");
900 if (rsp->status == NULL)
902 free_http_response(rsp);
903 return cgi_error_memory();
905 rsp->body = bindup(image_blank_data, image_blank_length);
906 if (rsp->body == NULL)
908 free_http_response(rsp);
909 return cgi_error_memory();
911 rsp->content_length = image_blank_length;
913 if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
915 free_http_response(rsp);
916 return cgi_error_memory();
922 rsp->status = strdup("302 Local Redirect from Privoxy");
923 if (rsp->status == NULL)
925 free_http_response(rsp);
926 return cgi_error_memory();
929 if (enlist_unique_header(rsp->headers, "Location", p))
931 free_http_response(rsp);
932 return cgi_error_memory();
936 #else /* Following code is disabled for now */
938 /* and handle accordingly: */
939 if ((p == NULL) || (0 == strcmpic(p, "pattern")))
941 p = CGI_PREFIX "send-banner?type=pattern";
943 else if (0 == strcmpic(p, "blank"))
945 p = CGI_PREFIX "send-banner?type=blank";
947 rsp->status = strdup("302 Local Redirect from Privoxy");
948 if (rsp->status == NULL)
950 free_http_response(rsp);
951 return cgi_error_memory();
954 if (enlist_unique_header(rsp->headers, "Location", p))
956 free_http_response(rsp);
957 return cgi_error_memory();
959 #endif /* Preceeding code is disabled for now */
961 else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
964 * Send empty document.
966 new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
969 rsp->body = strdup(" ");
970 rsp->content_length = 1;
972 rsp->status = strdup("403 Request blocked by Privoxy");
973 if (rsp->status == NULL)
975 free_http_response(rsp);
976 return cgi_error_memory();
978 if (new_content_type != 0)
980 log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
981 if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
983 free_http_response(rsp);
984 return cgi_error_memory();
989 #endif /* def FEATURE_IMAGE_BLOCKING */
992 * Else, generate an HTML "blocked" message:
996 struct map * exports;
1000 * Workaround for stupid Netscape bug which prevents
1001 * pages from being displayed if loading a referenced
1002 * JavaScript or style sheet fails. So make it appear
1003 * as if it succeeded.
1005 if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
1006 && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
1007 && !strstr(p, "Gecko") /* save Mozilla, */
1008 && !strstr(p, "compatible") /* MSIE */
1009 && !strstr(p, "Opera")) /* and Opera. */
1011 rsp->status = strdup("200 Request for blocked URL");
1015 rsp->status = strdup("403 Request for blocked URL");
1018 if (rsp->status == NULL)
1020 free_http_response(rsp);
1021 return cgi_error_memory();
1024 exports = default_exports(csp, NULL);
1025 if (exports == NULL)
1027 free_http_response(rsp);
1028 return cgi_error_memory();
1031 #ifdef FEATURE_FORCE_LOAD
1032 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1033 if (csp->http->ssl != 0 || 0 == strcmpic(csp->http->gpc, "connect"))
1034 #endif /* ndef FEATURE_FORCE_LOAD */
1036 err = map_block_killer(exports, "force-support");
1039 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1040 if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
1041 if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
1042 if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
1047 free_http_response(rsp);
1048 return cgi_error_memory();
1051 err = template_fill_for_cgi(csp, "blocked", exports, rsp);
1054 free_http_response(rsp);
1055 return cgi_error_memory();
1058 rsp->reason = RSP_REASON_BLOCKED;
1060 return finish_http_response(csp, rsp);
1065 #ifdef FEATURE_TRUST
1066 /*********************************************************************
1068 * Function : trust_url FIXME: I should be called distrust_url
1070 * Description : Calls is_untrusted_url to determine if the URL is trusted
1071 * and if not, returns a HTTP 403 response with a reject message.
1074 * 1 : csp = Current client state (buffers, headers, etc...)
1076 * Returns : NULL => trusted, else http_response.
1078 *********************************************************************/
1079 struct http_response *trust_url(struct client_state *csp)
1081 struct http_response *rsp;
1082 struct map * exports;
1083 char buf[BUFFER_SIZE];
1085 struct url_spec **tl;
1090 * Don't bother to work on trusted URLs
1092 if (!is_untrusted_url(csp))
1098 * Else, prepare a response:
1100 if (NULL == (rsp = alloc_http_response()))
1102 return cgi_error_memory();
1105 rsp->status = strdup("403 Request blocked by Privoxy");
1106 exports = default_exports(csp, NULL);
1107 if (exports == NULL || rsp->status == NULL)
1109 free_http_response(rsp);
1110 return cgi_error_memory();
1114 * Export the protocol, host, port, and referrer information
1116 err = map(exports, "hostport", 1, csp->http->hostport, 1);
1117 if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
1118 if (!err) err = map(exports, "path", 1, csp->http->path, 1);
1120 if (NULL != (p = get_header_value(csp->headers, "Referer:")))
1122 if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
1126 if (!err) err = map(exports, "referrer", 1, "none set", 1);
1132 free_http_response(rsp);
1133 return cgi_error_memory();
1137 * Export the trust list
1140 for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
1142 snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
1143 string_append(&p, buf);
1145 err = map(exports, "trusted-referrers", 1, p, 0);
1150 free_http_response(rsp);
1151 return cgi_error_memory();
1155 * Export the trust info, if available
1157 if (csp->config->trust_info->first)
1159 struct list_entry *l;
1162 for (l = csp->config->trust_info->first; l ; l = l->next)
1164 snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
1165 string_append(&p, buf);
1167 err = map(exports, "trust-info", 1, p, 0);
1171 err = map_block_killer(exports, "have-trust-info");
1177 free_http_response(rsp);
1178 return cgi_error_memory();
1182 * Export the force prefix or the force conditional block killer
1184 #ifdef FEATURE_FORCE_LOAD
1185 if (0 == strcmpic(csp->http->gpc, "connect"))
1187 err = map_block_killer(exports, "force-support");
1191 err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
1193 #else /* ifndef FEATURE_FORCE_LOAD */
1194 err = map_block_killer(exports, "force-support");
1195 #endif /* ndef FEATURE_FORCE_LOAD */
1200 free_http_response(rsp);
1201 return cgi_error_memory();
1205 * Build the response
1207 err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
1210 free_http_response(rsp);
1211 return cgi_error_memory();
1213 rsp->reason = RSP_REASON_UNTRUSTED;
1215 return finish_http_response(csp, rsp);
1217 #endif /* def FEATURE_TRUST */
1219 /*********************************************************************
1221 * Function : execute_single_pcrs_command
1223 * Description : Apply single pcrs command to the subject.
1224 * The subject itself is left untouched, memory for the result
1225 * is malloc()ed and it is the caller's responsibility to free
1226 * the result when it's no longer needed.
1229 * 1 : subject = the subject (== original) string
1230 * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
1231 * 3 : hits = int* for returning the number of modifications
1233 * Returns : NULL in case of errors, otherwise the
1234 * result of the pcrs command.
1236 *********************************************************************/
1237 char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
1241 char *result = NULL;
1245 assert(pcrs_command);
1248 size = strlen(subject);
1250 if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
1252 log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
1253 pcrs_command, error);
1255 else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
1257 log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
1264 job = pcrs_free_job(job);
1271 /*********************************************************************
1273 * Function : rewrite_url
1275 * Description : Rewrites a URL with a single pcrs command
1276 * and returns the result if it differs from the
1277 * original and isn't obviously invalid.
1280 * 1 : old_url = URL to rewrite.
1281 * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
1284 * Returns : NULL if the pcrs_command didn't change the url, or
1285 * the result of the modification.
1287 *********************************************************************/
1288 char *rewrite_url(char *old_url, const char *pcrs_command)
1290 char *new_url = NULL;
1294 assert(pcrs_command);
1296 new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
1300 log_error(LOG_LEVEL_REDIRECTS,
1301 "pcrs command \"%s\" didn't change \"%s\".",
1302 pcrs_command, old_url, new_url);
1305 else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
1307 log_error(LOG_LEVEL_ERROR,
1308 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
1309 "but the result doesn't look like a valid URL and will be ignored.",
1310 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1315 log_error(LOG_LEVEL_REDIRECTS,
1316 "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
1317 pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
1325 #ifdef FEATURE_FAST_REDIRECTS
1326 /*********************************************************************
1328 * Function : get_last_url
1330 * Description : Search for the last URL inside a string.
1331 * If the string already is a URL, it will
1332 * be the first URL found.
1335 * 1 : subject = the string to check
1336 * 2 : redirect_mode = +fast-redirect{} mode
1338 * Returns : NULL if no URL was found, or
1339 * the last URL found.
1341 *********************************************************************/
1342 char *get_last_url(char *subject, const char *redirect_mode)
1344 char *new_url = NULL;
1348 assert(redirect_mode);
1350 subject = strdup(subject);
1351 if (subject == NULL)
1353 log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
1357 if (0 == strcmpic(redirect_mode, "check-decoded-url"))
1359 log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
1360 new_url = url_decode(subject);
1361 if (new_url != NULL)
1368 log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
1372 log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
1375 * Find the last URL encoded in the request
1378 while ((tmp = strstr(tmp, "http://")) != NULL)
1382 tmp = (new_url != NULL) ? new_url : subject;
1383 while ((tmp = strstr(tmp, "https://")) != NULL)
1388 if ((new_url != NULL)
1389 && ( (new_url != subject)
1390 || (0 == strncmpic(subject, "http://", 7))
1391 || (0 == strncmpic(subject, "https://", 8))
1395 * Return new URL if we found a redirect
1396 * or if the subject already was a URL.
1398 * The second case makes sure that we can
1399 * chain get_last_url after another redirection check
1400 * (like rewrite_url) without losing earlier redirects.
1402 new_url = strdup(new_url);
1411 #endif /* def FEATURE_FAST_REDIRECTS */
1414 /*********************************************************************
1416 * Function : redirect_url
1418 * Description : Checks if Privoxy should answer the request with
1419 * a HTTP redirect and generates the redirect if
1423 * 1 : csp = Current client state (buffers, headers, etc...)
1425 * Returns : NULL if the request can pass, HTTP redirect otherwise.
1427 *********************************************************************/
1428 struct http_response *redirect_url(struct client_state *csp)
1430 struct http_response *rsp;
1431 #ifdef FEATURE_FAST_REDIRECTS
1433 * XXX: Do we still need FEATURE_FAST_REDIRECTS
1434 * as compile-time option? The user can easily disable
1435 * it in his action file.
1437 char * redirect_mode;
1438 #endif /* def FEATURE_FAST_REDIRECTS */
1439 char *old_url = NULL;
1440 char *new_url = NULL;
1441 char *redirection_string;
1443 if ((csp->action->flags & ACTION_REDIRECT))
1445 redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
1448 * If the redirection string begins with 's',
1449 * assume it's a pcrs command, otherwise treat it as
1450 * properly formatted URL and use it for the redirection
1453 * According to RFC 2616 section 14.30 the URL
1454 * has to be absolute and if the user tries:
1455 * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
1456 * she would get undefined results anyway.
1460 if (*redirection_string == 's')
1462 old_url = csp->http->url;
1463 new_url = rewrite_url(old_url, redirection_string);
1467 log_error(LOG_LEVEL_REDIRECTS,
1468 "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
1469 redirection_string);
1470 new_url = strdup(redirection_string);
1474 #ifdef FEATURE_FAST_REDIRECTS
1475 if ((csp->action->flags & ACTION_FAST_REDIRECTS))
1477 redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
1480 * If it exists, use the previously rewritten URL as input
1481 * otherwise just use the old path.
1483 old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
1484 new_url = get_last_url(old_url, redirect_mode);
1487 #endif /* def FEATURE_FAST_REDIRECTS */
1489 /* Did any redirect action trigger? */
1492 if (0 == strcmpic(new_url, csp->http->url))
1494 log_error(LOG_LEVEL_ERROR,
1495 "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
1496 csp->http->url, new_url);
1501 log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
1503 if (NULL == (rsp = alloc_http_response()))
1506 return cgi_error_memory();
1509 if ( enlist_unique_header(rsp->headers, "Location", new_url)
1510 || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
1513 free_http_response(rsp);
1514 return cgi_error_memory();
1516 rsp->reason = RSP_REASON_REDIRECTED;
1519 return finish_http_response(csp, rsp);
1523 /* Only reached if no redirect is required */
1529 #ifdef FEATURE_IMAGE_BLOCKING
1530 /*********************************************************************
1532 * Function : is_imageurl
1534 * Description : Given a URL, decide whether it is an image or not,
1535 * using either the info from a previous +image action
1536 * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
1537 * is MSIE and not on a Mac, tell from the browser's accept
1541 * 1 : csp = Current client state (buffers, headers, etc...)
1543 * Returns : True (nonzero) if URL is an image, false (0)
1546 *********************************************************************/
1547 int is_imageurl(const struct client_state *csp)
1549 #ifdef FEATURE_IMAGE_DETECT_MSIE
1552 tmp = get_header_value(csp->headers, "User-Agent:");
1553 if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
1555 tmp = get_header_value(csp->headers, "Accept:");
1556 if (tmp && strstr(tmp, "image/gif"))
1558 /* Client will accept HTML. If this seems counterintuitive,
1568 #endif /* def FEATURE_IMAGE_DETECT_MSIE */
1570 return ((csp->action->flags & ACTION_IMAGE) != 0);
1573 #endif /* def FEATURE_IMAGE_BLOCKING */
1576 #ifdef FEATURE_TRUST
1577 /*********************************************************************
1579 * Function : is_untrusted_url
1581 * Description : Should we "distrust" this URL (and block it)?
1583 * Yes if it matches a line in the trustfile, or if the
1584 * referrer matches a line starting with "+" in the
1589 * 1 : csp = Current client state (buffers, headers, etc...)
1591 * Returns : 0 => trusted, 1 => untrusted
1593 *********************************************************************/
1594 int is_untrusted_url(const struct client_state *csp)
1596 struct file_list *fl;
1597 struct block_spec *b;
1598 struct url_spec **trusted_url;
1599 struct http_request rhttp[1];
1600 const char * referer;
1604 * If we don't have a trustlist, we trust everybody
1606 if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
1611 memset(rhttp, '\0', sizeof(*rhttp));
1614 * Do we trust the request URL itself?
1616 for (b = b->next; b ; b = b->next)
1618 if (url_match(b->url, csp->http))
1624 if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
1626 /* no referrer was supplied */
1632 * If not, do we maybe trust its referrer?
1634 err = parse_http_url(referer, rhttp, csp);
1640 for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
1642 if (url_match(*trusted_url, rhttp))
1644 /* if the URL's referrer is from a trusted referrer, then
1645 * add the target spec to the trustfile as an unblocked
1646 * domain and return 0 (which means it's OK).
1651 if (NULL != (fp = fopen(csp->config->trustfile, "a")))
1655 char * new_entry = strdup("~");
1657 string_append(&new_entry, csp->http->hostport);
1659 path = csp->http->path;
1660 if ( (path[0] == '/')
1662 && ((path_end = strchr(path + 2, '/')) != NULL))
1664 /* since this path points into a user's home space
1665 * be sure to include this spec in the trustfile.
1667 int path_len = path_end - path; /* save offset */
1668 path = strdup(path); /* Copy string */
1671 path_end = path + path_len; /* regenerate ptr to new buffer */
1672 *(path_end + 1) = '\0'; /* Truncate path after '/' */
1674 string_join(&new_entry, path);
1678 * Give a reason for generating this entry.
1680 string_append(&new_entry, " # Trusted referrer was: ");
1681 string_append(&new_entry, referer);
1683 if (new_entry != NULL)
1685 if (-1 == fprintf(fp, "%s\n", new_entry))
1687 log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
1688 new_entry, csp->config->trustfile);
1694 /* FIXME: No way to handle out-of memory, so mostly ignoring it */
1695 log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
1702 log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
1703 csp->http->hostport, csp->config->trustfile);
1711 #endif /* def FEATURE_TRUST */
1714 /*********************************************************************
1716 * Function : pcrs_filter_response
1718 * Description : Execute all text substitutions from all applying
1719 * +filter actions on the text buffer that's been accumulated
1720 * in csp->iob->buf. If this changes the contents, set
1721 * csp->content_length to the modified size and raise the
1722 * CSP_FLAG_MODIFIED flag.
1724 * XXX: Currently pcrs_filter_response is also responsible
1725 * for dechunking and decompressing. Both should be
1726 * done in separate functions so other content modifiers
1727 * profit as well, even if pcrs filtering is disabled.
1730 * 1 : csp = Current client state (buffers, headers, etc...)
1732 * Returns : a pointer to the (newly allocated) modified buffer.
1733 * or NULL if there were no hits or something went wrong
1735 *********************************************************************/
1736 char *pcrs_filter_response(struct client_state *csp)
1739 size_t size, prev_size;
1741 char *old = csp->iob->cur, *new = NULL;
1744 struct file_list *fl;
1745 struct re_filterfile_spec *b;
1746 struct list_entry *filtername;
1748 int i, found_filters = 0;
1753 if (csp->iob->cur >= csp->iob->eod)
1757 size = (size_t)(csp->iob->eod - csp->iob->cur);
1760 * Need to check the set of re_filterfiles...
1762 for (i = 0; i < MAX_AF_FILES; i++)
1775 if (0 == found_filters)
1777 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1782 * If the body has a "chunked" transfer-encoding,
1783 * get rid of it first, adjusting size and iob->eod
1785 if (csp->flags & CSP_FLAG_CHUNKED)
1787 log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
1788 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1792 csp->iob->eod = csp->iob->cur + size;
1793 csp->flags |= CSP_FLAG_MODIFIED;
1798 * If the body has a compressed transfer-encoding,
1799 * uncompress it first, adjusting size and iob->eod.
1800 * Note that decompression occurs after de-chunking.
1802 if (csp->content_type & (CT_GZIP | CT_DEFLATE))
1804 /* Notice that we at least tried to decompress. */
1805 if (JB_ERR_OK != decompress_iob(csp))
1808 * We failed to decompress the data; there's no point
1809 * in continuing since we can't filter.
1811 * XXX: Actually the Accept-Encoding header may
1812 * just be incorrect in which case we could continue
1815 * Unset CT_GZIP and CT_DEFLATE to remember not
1816 * to modify the Content-Encoding header later.
1818 csp->content_type &= ~CT_GZIP;
1819 csp->content_type &= ~CT_DEFLATE;
1822 log_error(LOG_LEVEL_RE_FILTER,
1823 "Decompression successful. Old size: %d, new size: %d.",
1824 size, csp->iob->eod - csp->iob->cur);
1827 * Decompression gives us a completely new iob,
1828 * so we need to update.
1830 size = (size_t)(csp->iob->eod - csp->iob->cur);
1831 old = csp->iob->cur;
1833 csp->flags |= CSP_FLAG_MODIFIED;
1834 csp->content_type &= ~CT_TABOO;
1838 for (i = 0; i < MAX_AF_FILES; i++)
1841 if ((NULL == fl) || (NULL == fl->f))
1844 * Either there are no filter files
1845 * left, or this filter file just
1846 * contains no valid filters.
1848 * Continue to be sure we don't miss
1849 * valid filter files that are chained
1850 * after empty or invalid ones.
1855 * For all applying +filter actions, look if a filter by that
1856 * name exists and if yes, execute it's pcrs_joblist on the
1859 for (b = fl->f; b; b = b->next)
1861 for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
1862 filtername ; filtername = filtername->next)
1864 if (strcmp(b->name, filtername->str) == 0)
1866 int current_hits = 0; /* Number of hits caused by this filter */
1867 int job_number = 0; /* Which job we're currently executing */
1868 int job_hits = 0; /* How many hits the current job caused */
1870 if ( NULL == b->joblist )
1872 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1877 /* Apply all jobs from the joblist */
1878 for (job = b->joblist; NULL != job; job = job->next)
1881 job_hits = pcrs_execute(job, old, size, &new, &size);
1886 * That went well. Continue filtering
1887 * and use the result of this job as
1888 * input for the next one.
1890 current_hits += job_hits;
1891 if (old != csp->iob->cur)
1900 * This job caused an unexpected error. Inform the user
1901 * and skip the rest of the jobs in this filter. We could
1902 * continue with the next job, but usually the jobs
1903 * depend on each other or are similar enough to
1904 * fail for the same reason.
1906 * At the moment our pcrs expects the error codes of pcre 3.4,
1907 * but newer pcre versions can return additional error codes.
1908 * As a result pcrs_strerror()'s error message might be
1909 * "Unknown error ...", therefore we print the numerical value
1912 * XXX: Is this important enough for LOG_LEVEL_ERROR or
1913 * should we use LOG_LEVEL_RE_FILTER instead?
1915 log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
1916 b->name, job_number, pcrs_strerror(job_hits), job_hits);
1921 log_error(LOG_LEVEL_RE_FILTER,
1922 "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
1923 csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
1925 hits += current_hits;
1932 * If there were no hits, destroy our copy and let
1933 * chat() use the original in csp->iob
1941 csp->flags |= CSP_FLAG_MODIFIED;
1942 csp->content_length = size;
1950 /*********************************************************************
1952 * Function : gif_deanimate_response
1954 * Description : Deanimate the GIF image that has been accumulated in
1955 * csp->iob->buf, set csp->content_length to the modified
1956 * size and raise the CSP_FLAG_MODIFIED flag.
1959 * 1 : csp = Current client state (buffers, headers, etc...)
1961 * Returns : a pointer to the (newly allocated) modified buffer.
1962 * or NULL in case something went wrong.
1964 *********************************************************************/
1965 char *gif_deanimate_response(struct client_state *csp)
1967 struct binbuffer *in, *out;
1971 size = (size_t)(csp->iob->eod - csp->iob->cur);
1974 * If the body has a "chunked" transfer-encoding,
1975 * get rid of it first, adjusting size and iob->eod
1977 if (csp->flags & CSP_FLAG_CHUNKED)
1979 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
1980 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
1984 csp->iob->eod = csp->iob->cur + size;
1985 csp->flags |= CSP_FLAG_MODIFIED;
1988 if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
1989 || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
1991 log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
1995 in->buffer = csp->iob->cur;
1998 if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
2000 log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
2007 if ((int)size == out->offset)
2009 log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
2013 log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
2015 csp->content_length = out->offset;
2016 csp->flags |= CSP_FLAG_MODIFIED;
2026 /*********************************************************************
2028 * Function : jpeg_inspect_response
2033 * 1 : csp = Current client state (buffers, headers, etc...)
2035 * Returns : a pointer to the (newly allocated) modified buffer
2036 * or NULL in case something went wrong.
2038 *********************************************************************/
2039 char *jpeg_inspect_response(struct client_state *csp)
2041 struct binbuffer *in = NULL;
2042 struct binbuffer *out = NULL;
2046 size = (size_t)(csp->iob->eod - csp->iob->cur);
2049 * If the body has a "chunked" transfer-encoding,
2050 * get rid of it first, adjusting size and iob->eod
2052 if (csp->flags & CSP_FLAG_CHUNKED)
2054 log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
2055 if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
2059 csp->iob->eod = csp->iob->cur + size;
2060 csp->flags |= CSP_FLAG_MODIFIED;
2063 if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
2065 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
2069 if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
2071 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
2075 in->buffer = csp->iob->cur;
2079 * Calling jpeg_inspect has the side-effect of creating and
2080 * modifying the image buffer of "out" directly.
2082 if (jpeg_inspect(in, out))
2084 log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
2092 csp->content_length = out->offset;
2093 csp->flags |= CSP_FLAG_MODIFIED;
2103 /*********************************************************************
2105 * Function : remove_chunked_transfer_coding
2107 * Description : In-situ remove the "chunked" transfer coding as defined
2108 * in rfc2616 from a buffer.
2111 * 1 : buffer = Pointer to the text buffer
2112 * 2 : size = Number of bytes to be processed
2114 * Returns : The new size, i.e. the number of bytes from buffer which
2115 * are occupied by the stripped body, or 0 in case something
2118 *********************************************************************/
2119 size_t remove_chunked_transfer_coding(char *buffer, const size_t size)
2122 unsigned int chunksize = 0;
2123 char *from_p, *to_p;
2126 from_p = to_p = buffer;
2128 if (sscanf(buffer, "%x", &chunksize) != 1)
2130 log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
2134 while (chunksize > 0)
2136 if (NULL == (from_p = strstr(from_p, "\r\n")))
2138 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2142 if ((newsize += chunksize) >= size)
2144 log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
2149 memmove(to_p, from_p, (size_t) chunksize);
2150 to_p = buffer + newsize;
2151 from_p += chunksize + 2;
2153 if (sscanf(from_p, "%x", &chunksize) != 1)
2155 log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
2160 /* FIXME: Should this get its own loglevel? */
2161 log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
2167 /*********************************************************************
2169 * Function : url_actions
2171 * Description : Gets the actions for this URL.
2174 * 1 : http = http_request request for blocked URLs
2175 * 2 : csp = Current client state (buffers, headers, etc...)
2179 *********************************************************************/
2180 void url_actions(struct http_request *http,
2181 struct client_state *csp)
2183 struct file_list *fl;
2184 struct url_actions *b;
2187 init_current_action(csp->action);
2189 for (i = 0; i < MAX_AF_FILES; i++)
2191 if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
2196 apply_url_actions(csp->action, http, b);
2203 /*********************************************************************
2205 * Function : apply_url_actions
2207 * Description : Applies a list of URL actions.
2210 * 1 : action = Destination.
2211 * 2 : http = Current URL
2212 * 3 : b = list of URL actions to apply
2216 *********************************************************************/
2217 void apply_url_actions(struct current_action_spec *action,
2218 struct http_request *http,
2219 struct url_actions *b)
2223 /* Should never happen */
2227 for (b = b->next; NULL != b; b = b->next)
2229 if (url_match(b->url, http))
2231 merge_current_action(action, b->action);
2237 /*********************************************************************
2239 * Function : forward_url
2241 * Description : Should we forward this to another proxy?
2244 * 1 : http = http_request request for current URL
2245 * 2 : csp = Current client state (buffers, headers, etc...)
2247 * Returns : Pointer to forwarding information.
2249 *********************************************************************/
2250 const struct forward_spec * forward_url(struct http_request *http,
2251 struct client_state *csp)
2253 static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
2254 struct forward_spec *fwd = csp->config->forward;
2263 if (url_match(fwd->url, http))
2274 /*********************************************************************
2276 * Function : direct_response
2278 * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
2279 * request and if so, return a HTTP 501 to the client.
2281 * FIXME: I have a stupid name and I should handle the
2282 * requests properly. Still, what we do here is rfc-
2283 * compliant, whereas ignoring or forwarding are not.
2286 * 1 : csp = Current client state (buffers, headers, etc...)
2288 * Returns : http_response if , NULL if nonmatch or handler fail
2290 *********************************************************************/
2291 struct http_response *direct_response(struct client_state *csp)
2293 struct http_response *rsp;
2294 struct list_entry *p;
2296 if ((0 == strcmpic(csp->http->gpc, "trace"))
2297 || (0 == strcmpic(csp->http->gpc, "options")))
2299 for (p = csp->headers->first; (p != NULL) ; p = p->next)
2301 if (!strncmpic("Max-Forwards:", p->str, 13))
2303 unsigned int max_forwards;
2306 * If it's a Max-Forwards value of zero,
2307 * we have to intercept the request.
2309 if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
2312 * FIXME: We could handle at least TRACE here,
2313 * but that would require a verbatim copy of
2314 * the request which we don't have anymore
2316 log_error(LOG_LEVEL_HEADER,
2317 "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
2320 /* Get mem for response or fail*/
2321 if (NULL == (rsp = alloc_http_response()))
2323 return cgi_error_memory();
2326 if (NULL == (rsp->status = strdup("501 Not Implemented")))
2328 free_http_response(rsp);
2329 return cgi_error_memory();
2333 rsp->reason = RSP_REASON_UNSUPPORTED;
2335 return(finish_http_response(csp, rsp));