1 const char parsers_rcs[] = "$Id: parsers.c,v 1.118 2007/12/28 16:56:35 fabiankeil Exp $";
2 /*********************************************************************
4 * File : $Source: /cvsroot/ijbswa/current/parsers.c,v $
6 * Purpose : Declares functions to parse/crunch headers and pages.
7 * Functions declared include:
8 * `add_to_iob', `client_cookie_adder', `client_from',
9 * `client_referrer', `client_send_cookie', `client_ua',
10 * `client_uagent', `client_x_forwarded',
11 * `client_x_forwarded_adder', `client_xtra_adder',
12 * `content_type', `crumble', `destroy_list', `enlist',
13 * `flush_socket', ``get_header', `sed', `filter_header'
14 * `server_content_encoding', `server_content_disposition',
15 * `server_last_modified', `client_accept_language',
16 * `crunch_client_header', `client_if_modified_since',
17 * `client_if_none_match', `get_destination_from_headers',
18 * `parse_header_time', `decompress_iob' and `server_set_cookie'.
20 * Copyright : Written by and Copyright (C) 2001-2007 the SourceForge
21 * Privoxy team. http://www.privoxy.org/
23 * Based on the Internet Junkbuster originally written
24 * by and Copyright (C) 1997 Anonymous Coders and
25 * Junkbusters Corporation. http://www.junkbusters.com
27 * This program is free software; you can redistribute it
28 * and/or modify it under the terms of the GNU General
29 * Public License as published by the Free Software
30 * Foundation; either version 2 of the License, or (at
31 * your option) any later version.
33 * This program is distributed in the hope that it will
34 * be useful, but WITHOUT ANY WARRANTY; without even the
35 * implied warranty of MERCHANTABILITY or FITNESS FOR A
36 * PARTICULAR PURPOSE. See the GNU General Public
37 * License for more details.
39 * The GNU General Public License should be included with
40 * this file. If not, you can view it at
41 * http://www.gnu.org/copyleft/gpl.html
42 * or write to the Free Software Foundation, Inc., 59
43 * Temple Place - Suite 330, Boston, MA 02111-1307, USA.
47 * Revision 1.118 2007/12/28 16:56:35 fabiankeil
48 * Minor server_content_disposition() changes:
49 * - Don't regenerate the header name all lower-case.
50 * - Some white space fixes.
51 * - Remove useless log message in case of ENOMEM.
53 * Revision 1.117 2007/12/06 18:11:50 fabiankeil
54 * Garbage-collect the code to add a X-Forwarded-For
55 * header as it seems to be mostly used by accident.
57 * Revision 1.116 2007/12/01 13:04:22 fabiankeil
58 * Fix a crash on mingw32 with some Last Modified times in the future.
60 * Revision 1.115 2007/11/02 16:52:50 fabiankeil
61 * Remove a "can't happen" error block which, over
62 * time, mutated into a "guaranteed to happen" block.
64 * Revision 1.114 2007/10/19 16:56:26 fabiankeil
65 * - Downgrade "Buffer limit reached" message to LOG_LEVEL_INFO.
66 * - Use shiny new content_filters_enabled() in client_range().
68 * Revision 1.113 2007/10/10 17:29:57 fabiankeil
69 * I forgot about Poland.
71 * Revision 1.112 2007/10/09 16:38:40 fabiankeil
72 * Remove Range and If-Range headers if content filtering is enabled.
74 * Revision 1.111 2007/10/04 18:07:00 fabiankeil
75 * Move ACTION_VANILLA_WAFER handling from jcc's chat() into
76 * client_cookie_adder() to make sure send-vanilla-wafer can be
77 * controlled through tags (and thus regression-tested).
79 * Revision 1.110 2007/09/29 10:42:37 fabiankeil
80 * - Remove "scanning headers for" log message again.
81 * - Some more whitespace fixes.
83 * Revision 1.109 2007/09/08 14:25:48 fabiankeil
84 * Refactor client_referrer() and add conditional-forge parameter.
86 * Revision 1.108 2007/08/28 18:21:03 fabiankeil
87 * A bunch of whitespace fixes, pointy hat to me.
89 * Revision 1.107 2007/08/28 18:16:32 fabiankeil
90 * Fix possible memory corruption in server_http, make sure it's not
91 * executed for ordinary server headers and mark some problems for later.
93 * Revision 1.106 2007/08/18 14:30:32 fabiankeil
94 * Let content-type-overwrite{} honour force-text-mode again.
96 * Revision 1.105 2007/08/11 14:49:49 fabiankeil
97 * - Add prototpyes for the header parsers and make them static.
98 * - Comment out client_accept_encoding_adder() which isn't used right now.
100 * Revision 1.104 2007/07/14 07:38:19 fabiankeil
101 * Move the ACTION_FORCE_TEXT_MODE check out of
102 * server_content_type(). Signal other functions
103 * whether or not a content type has been declared.
104 * Part of the fix for BR#1750917.
106 * Revision 1.103 2007/06/01 16:31:54 fabiankeil
107 * Change sed() to return a jb_err in preparation for forward-override{}.
109 * Revision 1.102 2007/05/27 12:39:32 fabiankeil
110 * Adjust "X-Filter: No" to disable dedicated header filters.
112 * Revision 1.101 2007/05/14 10:16:41 fabiankeil
113 * Streamline client_cookie_adder().
115 * Revision 1.100 2007/04/30 15:53:11 fabiankeil
116 * Make sure filters with dynamic jobs actually use them.
118 * Revision 1.99 2007/04/30 15:06:26 fabiankeil
119 * - Introduce dynamic pcrs jobs that can resolve variables.
120 * - Remove unnecessary update_action_bits_for_all_tags() call.
122 * Revision 1.98 2007/04/17 18:32:10 fabiankeil
123 * - Make tagging based on tags set by earlier taggers
124 * of the same kind possible.
125 * - Log whether or not new tags cause action bits updates
126 * (in which case a matching tag-pattern section exists).
127 * - Log if the user tries to set a tag that is already set.
129 * Revision 1.97 2007/04/15 16:39:21 fabiankeil
130 * Introduce tags as alternative way to specify which
131 * actions apply to a request. At the moment tags can be
132 * created based on client and server headers.
134 * Revision 1.96 2007/04/12 12:53:58 fabiankeil
135 * Log a warning if the content is compressed, filtering is
136 * enabled and Privoxy was compiled without zlib support.
139 * Revision 1.95 2007/03/25 14:26:40 fabiankeil
140 * - Fix warnings when compiled with glibc.
141 * - Don't use crumble() for cookie crunching.
142 * - Move cookie time parsing into parse_header_time().
143 * - Let parse_header_time() return a jb_err code
144 * instead of a pointer that can only be used to
145 * check for NULL anyway.
147 * Revision 1.94 2007/03/21 12:23:53 fabiankeil
148 * - Add better protection against malicious gzip headers.
149 * - Stop logging the first hundred bytes of decompressed content.
150 * It looks like it's working and there is always debug 16.
151 * - Log the content size after decompression in decompress_iob()
152 * instead of pcrs_filter_response().
154 * Revision 1.93 2007/03/20 15:21:44 fabiankeil
155 * - Use dedicated header filter actions instead of abusing "filter".
156 * Replace "filter-client-headers" and "filter-client-headers"
157 * with "server-header-filter" and "client-header-filter".
158 * - Remove filter_client_header() and filter_client_header(),
159 * filter_header() now checks the shiny new
160 * CSP_FLAG_CLIENT_HEADER_PARSING_DONE flag instead.
162 * Revision 1.92 2007/03/05 13:25:32 fabiankeil
163 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
164 * - Handle "Cookie:" and "Connection:" headers a bit smarter
165 * (don't crunch them just to recreate them later on).
166 * - Add another non-standard time format for the cookie
167 * expiration date detection.
168 * - Fix a valgrind warning.
170 * Revision 1.91 2007/02/24 12:27:32 fabiankeil
171 * Improve cookie expiration date detection.
173 * Revision 1.90 2007/02/08 19:12:35 fabiankeil
174 * Don't run server_content_length() the first time
175 * sed() parses server headers; only adjust the
176 * Content-Length header if the page was modified.
178 * Revision 1.89 2007/02/07 16:52:11 fabiankeil
179 * Fix log messages regarding the cookie time format
180 * (cookie and request URL were mixed up).
182 * Revision 1.88 2007/02/07 11:27:12 fabiankeil
183 * - Let decompress_iob()
184 * - not corrupt the content if decompression fails
185 * early. (the first byte(s) were lost).
186 * - use pointer arithmetics with defined outcome for
188 * - Use a different kludge to remember a failed decompression.
190 * Revision 1.87 2007/01/31 16:21:38 fabiankeil
191 * Search for Max-Forwards headers case-insensitive,
192 * don't generate the "501 unsupported" message for invalid
193 * Max-Forwards values and don't increase negative ones.
195 * Revision 1.86 2007/01/30 13:05:26 fabiankeil
196 * - Let server_set_cookie() check the expiration date
197 * of cookies and don't touch the ones that are already
198 * expired. Fixes problems with low quality web applications
199 * as described in BR 932612.
201 * - Adjust comment in client_max_forwards to reality;
202 * remove invalid Max-Forwards headers.
204 * Revision 1.85 2007/01/26 15:33:46 fabiankeil
205 * Stop filter_header() from unintentionally removing
206 * empty header lines that were enlisted by the continue
209 * Revision 1.84 2007/01/24 12:56:52 fabiankeil
210 * - Repeat the request URL before logging any headers.
211 * Makes reading the log easier in case of simultaneous requests.
212 * - If there are more than one Content-Type headers in one request,
213 * use the first one and remove the others.
214 * - Remove "newval" variable in server_content_type().
215 * It's only used once.
217 * Revision 1.83 2007/01/12 15:03:02 fabiankeil
218 * Correct a cast, check inflateEnd() exit code
219 * to see if we have to, replace sprintf calls
222 * Revision 1.82 2007/01/01 19:36:37 fabiankeil
223 * Integrate a modified version of Wil Mahan's
224 * zlib patch (PR #895531).
226 * Revision 1.81 2006/12/31 22:21:33 fabiankeil
227 * Skip empty filter files in filter_header()
228 * but don't ignore the ones that come afterwards.
229 * Fixes BR 1619208, this time for real.
231 * Revision 1.80 2006/12/29 19:08:22 fabiankeil
232 * Reverted parts of my last commit
233 * to keep error handling working.
235 * Revision 1.79 2006/12/29 18:04:40 fabiankeil
236 * Fixed gcc43 conversion warnings.
238 * Revision 1.78 2006/12/26 17:19:20 fabiankeil
239 * Bringing back the "useless" localtime() call
240 * I removed in revision 1.67. On some platforms
241 * it's necessary to prevent time zone offsets.
243 * Revision 1.77 2006/12/07 18:44:26 fabiankeil
244 * Rebuild request URL in get_destination_from_headers()
245 * to make sure redirect{pcrs command} works as expected
246 * for intercepted requests.
248 * Revision 1.76 2006/12/06 19:52:25 fabiankeil
249 * Added get_destination_from_headers().
251 * Revision 1.75 2006/11/13 19:05:51 fabiankeil
252 * Make pthread mutex locking more generic. Instead of
253 * checking for OSX and OpenBSD, check for FEATURE_PTHREAD
254 * and use mutex locking unless there is an _r function
255 * available. Better safe than sorry.
257 * Fixes "./configure --disable-pthread" and should result
258 * in less threading-related problems on pthread-using platforms,
259 * but it still doesn't fix BR#1122404.
261 * Revision 1.74 2006/10/02 16:59:12 fabiankeil
262 * The special header "X-Filter: No" now disables
263 * header filtering as well.
265 * Revision 1.73 2006/09/23 13:26:38 roro
266 * Replace TABs by spaces in source code.
268 * Revision 1.72 2006/09/23 12:37:21 fabiankeil
269 * Don't print a log message every time filter_headers is
270 * entered or left. It only creates noise without any real
273 * Revision 1.71 2006/09/21 19:55:17 fabiankeil
274 * Fix +hide-if-modified-since{-n}.
276 * Revision 1.70 2006/09/08 12:06:34 fabiankeil
277 * Have hide-if-modified-since interpret the random
278 * range value as minutes instead of hours. Allows
279 * more fine-grained configuration.
281 * Revision 1.69 2006/09/06 16:25:51 fabiankeil
282 * Always have parse_header_time return a pointer
283 * that actual makes sense, even though we currently
284 * only need it to detect problems.
286 * Revision 1.68 2006/09/06 10:43:32 fabiankeil
287 * Added config option enable-remote-http-toggle
288 * to specify if Privoxy should recognize special
289 * headers (currently only X-Filter) to change its
290 * behaviour. Disabled by default.
292 * Revision 1.67 2006/09/04 11:01:26 fabiankeil
293 * After filtering de-chunked instances, remove
294 * "Transfer-Encoding" header entirely instead of changing
295 * it to "Transfer-Encoding: identity", which is invalid.
296 * Thanks Michael Shields <shields@msrl.com>. Fixes PR 1318658.
298 * Don't use localtime in parse_header_time. An empty time struct
299 * is good enough, it gets overwritten by strptime anyway.
301 * Revision 1.66 2006/09/03 19:38:28 fabiankeil
302 * Use gmtime_r if available, fallback to gmtime with mutex
303 * protection for MacOSX and use vanilla gmtime for the rest.
305 * Revision 1.65 2006/08/22 10:55:56 fabiankeil
306 * Changed client_referrer to use the right type (size_t) for
307 * hostlenght and to shorten the temporary referrer string with
308 * '\0' instead of adding a useless line break.
310 * Revision 1.64 2006/08/17 17:15:10 fabiankeil
311 * - Back to timegm() using GnuPG's replacement if necessary.
312 * Using mktime() and localtime() could add a on hour offset if
313 * the randomize factor was big enough to lead to a summer/wintertime
316 * - Removed now-useless Privoxy 3.0.3 compatibility glue.
318 * - Moved randomization code into pick_from_range().
320 * - Changed parse_header_time definition.
321 * time_t isn't guaranteed to be signed and
322 * if it isn't, -1 isn't available as error code.
323 * Changed some variable types in client_if_modified_since()
324 * because of the same reason.
326 * Revision 1.63 2006/08/14 13:18:08 david__schmidt
327 * OS/2 compilation compatibility fixups
329 * Revision 1.62 2006/08/14 08:58:42 fabiankeil
330 * Changed include from strptime.c to strptime.h
332 * Revision 1.61 2006/08/14 08:25:19 fabiankeil
333 * Split filter-headers{} into filter-client-headers{}
334 * and filter-server-headers{}.
335 * Added parse_header_time() to share some code.
336 * Replaced timegm() with mktime().
338 * Revision 1.60 2006/08/12 03:54:37 david__schmidt
339 * Windows service integration
341 * Revision 1.59 2006/08/03 02:46:41 david__schmidt
342 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
344 * Revision 1.58 2006/07/18 14:48:47 david__schmidt
345 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
346 * with what was really the latest development (the v_3_0_branch branch)
348 * Revision 1.56.2.10 2006/01/21 16:16:08 david__schmidt
349 * Thanks to Edward Carrel for his patch to modernize OSX's
\rpthreads support. See bug #1409623.
351 * Revision 1.56.2.9 2004/10/03 12:53:45 david__schmidt
352 * Add the ability to check jpeg images for invalid
353 * lengths of comment blocks. Defensive strategy
354 * against the exploit:
355 * Microsoft Security Bulletin MS04-028
356 * Buffer Overrun in JPEG Processing (GDI+) Could
357 * Allow Code Execution (833987)
358 * Enabled with +inspect-jpegs in actions files.
360 * Revision 1.56.2.8 2003/07/11 13:21:25 oes
361 * Excluded text/plain objects from filtering. This fixes a
362 * couple of client-crashing, download corruption and
363 * Privoxy performance issues, whose root cause lies in
364 * web servers labelling content of unknown type as text/plain.
366 * Revision 1.56.2.7 2003/05/06 12:07:26 oes
367 * Fixed bug #729900: Suspicious HOST: headers are now killed and regenerated if necessary
369 * Revision 1.56.2.6 2003/04/14 21:28:30 oes
370 * Completing the previous change
372 * Revision 1.56.2.5 2003/04/14 12:08:16 oes
373 * Added temporary workaround for bug in PHP < 4.2.3
375 * Revision 1.56.2.4 2003/03/07 03:41:05 david__schmidt
376 * Wrapping all *_r functions (the non-_r versions of them) with mutex semaphores for OSX. Hopefully this will take care of all of those pesky crash reports.
378 * Revision 1.56.2.3 2002/11/10 04:20:02 hal9
379 * Fix typo: supressed -> suppressed
381 * Revision 1.56.2.2 2002/09/25 14:59:53 oes
382 * Improved cookie logging
384 * Revision 1.56.2.1 2002/09/25 14:52:45 oes
385 * Added basic support for OPTIONS and TRACE HTTP methods:
386 * - New parser function client_max_forwards which decrements
387 * the Max-Forwards HTTP header field of OPTIONS and TRACE
388 * requests by one before forwarding
389 * - New parser function client_host which extracts the host
390 * and port information from the HTTP header field if the
391 * request URI was not absolute
392 * - Don't crumble and re-add the Host: header, but only generate
393 * and append if missing
395 * Revision 1.56 2002/05/12 15:34:22 jongfoster
396 * Fixing typo in a comment
398 * Revision 1.55 2002/05/08 16:01:07 oes
399 * Optimized add_to_iob:
400 * - Use realloc instead of malloc(), memcpy(), free()
401 * - Expand to powers of two if possible, to get
402 * O(log n) reallocs instead of O(n).
403 * - Moved check for buffer limit here from chat
404 * - Report failure via returncode
406 * Revision 1.54 2002/04/02 15:03:16 oes
407 * Tiny code cosmetics
409 * Revision 1.53 2002/03/26 22:29:55 swa
410 * we have a new homepage!
412 * Revision 1.52 2002/03/24 13:25:43 swa
413 * name change related issues
415 * Revision 1.51 2002/03/13 00:27:05 jongfoster
418 * Revision 1.50 2002/03/12 01:45:35 oes
419 * More verbose logging
421 * Revision 1.49 2002/03/09 20:03:52 jongfoster
422 * - Making various functions return int rather than size_t.
423 * (Undoing a recent change). Since size_t is unsigned on
424 * Windows, functions like read_socket that return -1 on
425 * error cannot return a size_t.
427 * THIS WAS A MAJOR BUG - it caused frequent, unpredictable
428 * crashes, and also frequently caused JB to jump to 100%
429 * CPU and stay there. (Because it thought it had just
430 * read ((unsigned)-1) == 4Gb of data...)
432 * - The signature of write_socket has changed, it now simply
433 * returns success=0/failure=nonzero.
435 * - Trying to get rid of a few warnings --with-debug on
436 * Windows, I've introduced a new type "jb_socket". This is
437 * used for the socket file descriptors. On Windows, this
438 * is SOCKET (a typedef for unsigned). Everywhere else, it's
439 * an int. The error value can't be -1 any more, so it's
440 * now JB_INVALID_SOCKET (which is -1 on UNIX, and in
441 * Windows it maps to the #define INVALID_SOCKET.)
443 * - The signature of bind_port has changed.
445 * Revision 1.48 2002/03/07 03:46:53 oes
446 * Fixed compiler warnings etc
448 * Revision 1.47 2002/02/20 23:15:13 jongfoster
449 * Parsing functions now handle out-of-memory gracefully by returning
452 * Revision 1.46 2002/01/17 21:03:47 jongfoster
453 * Moving all our URL and URL pattern parsing code to urlmatch.c.
455 * Revision 1.45 2002/01/09 14:33:03 oes
456 * Added support for localtime_r.
458 * Revision 1.44 2001/12/14 01:22:54 steudten
459 * Remove 'user:pass@' from 'proto://user:pass@host' for the
460 * new added header 'Host: ..'. (See Req ID 491818)
462 * Revision 1.43 2001/11/23 00:26:38 jongfoster
463 * Fixing two really stupid errors in my previous commit
465 * Revision 1.42 2001/11/22 21:59:30 jongfoster
466 * Adding code to handle +no-cookies-keep
468 * Revision 1.41 2001/11/05 23:43:05 steudten
469 * Add time+date to log files.
471 * Revision 1.40 2001/10/26 20:13:09 jongfoster
472 * ctype.h is needed in Windows, too.
474 * Revision 1.39 2001/10/26 17:40:04 oes
475 * Introduced get_header_value()
476 * Removed http->user_agent, csp->referrer and csp->accept_types
477 * Removed client_accept()
479 * Revision 1.38 2001/10/25 03:40:48 david__schmidt
480 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
481 * threads to call select() simultaneously. So, it's time to do a real, live,
482 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
483 * (native). Both versions will work, but using __OS2__ offers multi-threading.
485 * Revision 1.37 2001/10/23 21:36:02 jongfoster
486 * Documenting sed()'s error behaviou (doc change only)
488 * Revision 1.36 2001/10/13 12:51:51 joergs
489 * Removed client_host, (was only required for the old 2.0.2-11 http://noijb.
490 * force-load), instead crumble Host: and add it (again) in client_host_adder
491 * (in case we get a HTTP/1.0 request without Host: header and forward it to
492 * a HTTP/1.1 server/proxy).
494 * Revision 1.35 2001/10/09 22:39:21 jongfoster
495 * assert.h is also required under Win32, so moving out of #ifndef _WIN32
498 * Revision 1.34 2001/10/07 18:50:55 oes
499 * Added server_content_encoding, renamed server_transfer_encoding
501 * Revision 1.33 2001/10/07 18:04:49 oes
502 * Changed server_http11 to server_http and its pattern to "HTTP".
503 * Additional functionality: it now saves the HTTP status into
504 * csp->http->status and sets CT_TABOO for Status 206 (partial range)
506 * Revision 1.32 2001/10/07 15:43:28 oes
507 * Removed FEATURE_DENY_GZIP and replaced it with client_accept_encoding,
508 * client_te and client_accept_encoding_adder, triggered by the new
509 * +no-compression action. For HTTP/1.1 the Accept-Encoding header is
510 * changed to allow only identity and chunked, and the TE header is
511 * crunched. For HTTP/1.0, Accept-Encoding is crunched.
513 * parse_http_request no longer does anything than parsing. The rewriting
514 * of http->cmd and version mangling are gone. It now also recognizes
515 * the put and delete methods and saves the url in http->url. Removed
518 * renamed content_type and content_length to have the server_ prefix
520 * server_content_type now only works if csp->content_type != CT_TABOO
522 * added server_transfer_encoding, which
523 * - Sets CT_TABOO to prohibit filtering if encoding compresses
524 * - Raises the CSP_FLAG_CHUNKED flag if Encoding is "chunked"
525 * - Change from "chunked" to "identity" if body was chunked
526 * but has been de-chunked for filtering.
528 * added server_content_md5 which crunches any Content-MD5 headers
529 * if the body was modified.
531 * made server_http11 conditional on +downgrade action
533 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
535 * Revision 1.31 2001/10/05 14:25:02 oes
536 * Crumble Keep-Alive from Server
538 * Revision 1.30 2001/09/29 12:56:03 joergs
539 * IJB now changes HTTP/1.1 to HTTP/1.0 in requests and answers.
541 * Revision 1.29 2001/09/24 21:09:24 jongfoster
542 * Fixing 2 memory leaks that Guy spotted, where the paramater to
543 * enlist() was not being free()d.
545 * Revision 1.28 2001/09/22 16:32:28 jongfoster
546 * Removing unused #includes.
548 * Revision 1.27 2001/09/20 15:45:25 steudten
550 * add casting from size_t to int for printf()
551 * remove local variable shadow s2
553 * Revision 1.26 2001/09/16 17:05:14 jongfoster
554 * Removing unused #include showarg.h
556 * Revision 1.25 2001/09/16 13:21:27 jongfoster
557 * Changes to use new list functions.
559 * Revision 1.24 2001/09/13 23:05:50 jongfoster
560 * Changing the string paramater to the header parsers a "const".
562 * Revision 1.23 2001/09/12 18:08:19 steudten
564 * In parse_http_request() header rewriting miss the host value, so
565 * from http://www.mydomain.com the result was just " / " not
566 * http://www.mydomain.com/ in case we forward.
568 * Revision 1.22 2001/09/10 10:58:53 oes
569 * Silenced compiler warnings
571 * Revision 1.21 2001/07/31 14:46:00 oes
572 * - Persistant connections now suppressed
573 * - sed() no longer appends empty header to csp->headers
575 * Revision 1.20 2001/07/30 22:08:36 jongfoster
576 * Tidying up #defines:
577 * - All feature #defines are now of the form FEATURE_xxx
578 * - Permanently turned off WIN_GUI_EDIT
579 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
581 * Revision 1.19 2001/07/25 17:21:54 oes
582 * client_uagent now saves copy of User-Agent: header value
584 * Revision 1.18 2001/07/13 14:02:46 oes
585 * - Included fix to repair broken HTTP requests that
586 * don't contain a path, not even '/'.
587 * - Removed all #ifdef PCRS
588 * - content_type now always inspected and classified as
589 * text, gif or other.
590 * - formatting / comments
592 * Revision 1.17 2001/06/29 21:45:41 oes
593 * Indentation, CRLF->LF, Tab-> Space
595 * Revision 1.16 2001/06/29 13:32:42 oes
597 * - Adapted free_http_request
598 * - Removed logentry from cancelled commit
600 * Revision 1.15 2001/06/03 19:12:38 oes
601 * deleted const struct interceptors
603 * Revision 1.14 2001/06/01 18:49:17 jongfoster
604 * Replaced "list_share" with "list" - the tiny memory gain was not
605 * worth the extra complexity.
607 * Revision 1.13 2001/05/31 21:30:33 jongfoster
608 * Removed list code - it's now in list.[ch]
609 * Renamed "permission" to "action", and changed many features
610 * to use the actions file rather than the global config.
612 * Revision 1.12 2001/05/31 17:33:13 oes
616 * Revision 1.11 2001/05/29 20:11:19 joergs
617 * '/ * inside comment' warning removed.
619 * Revision 1.10 2001/05/29 09:50:24 jongfoster
620 * Unified blocklist/imagelist/permissionslist.
621 * File format is still under discussion, but the internal changes
624 * Also modified interceptor behaviour:
625 * - We now intercept all URLs beginning with one of the following
626 * prefixes (and *only* these prefixes):
628 * * http://ijbswa.sf.net/config/
629 * * http://ijbswa.sourceforge.net/config/
630 * - New interceptors "home page" - go to http://i.j.b/ to see it.
631 * - Internal changes so that intercepted and fast redirect pages
632 * are not replaced with an image.
633 * - Interceptors now have the option to send a binary page direct
634 * to the client. (i.e. ijb-send-banner uses this)
635 * - Implemented show-url-info interceptor. (Which is why I needed
636 * the above interceptors changes - a typical URL is
637 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
638 * The previous mechanism would not have intercepted that, and
639 * if it had been intercepted then it then it would have replaced
642 * Revision 1.9 2001/05/28 17:26:33 jongfoster
643 * Fixing segfault if last header was crunched.
644 * Fixing Windows build (snprintf() is _snprintf() under Win32, but we
645 * can use the cross-platform sprintf() instead.)
647 * Revision 1.8 2001/05/27 22:17:04 oes
649 * - re_process_buffer no longer writes the modified buffer
650 * to the client, which was very ugly. It now returns the
651 * buffer, which it is then written by chat.
653 * - content_length now adjusts the Content-Length: header
654 * for modified documents rather than crunch()ing it.
655 * (Length info in csp->content_length, which is 0 for
656 * unmodified documents)
658 * - For this to work, sed() is called twice when filtering.
660 * Revision 1.7 2001/05/27 13:19:06 oes
661 * Patched Joergs solution for the content-length in.
663 * Revision 1.6 2001/05/26 13:39:32 jongfoster
664 * Only crunches Content-Length header if applying RE filtering.
665 * Without this fix, Microsoft Windows Update wouldn't work.
667 * Revision 1.5 2001/05/26 00:28:36 jongfoster
668 * Automatic reloading of config file.
669 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
670 * Most of the global variables have been moved to a new
671 * struct configuration_spec, accessed through csp->config->globalname
672 * Most of the globals remaining are used by the Win32 GUI.
674 * Revision 1.4 2001/05/22 18:46:04 oes
676 * - Enabled filtering banners by size rather than URL
677 * by adding patterns that replace all standard banner
678 * sizes with the "Junkbuster" gif to the re_filterfile
680 * - Enabled filtering WebBugs by providing a pattern
681 * which kills all 1x1 images
683 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
684 * which is selected by the (nonstandard and therefore
685 * capital) letter 'U' in the option string.
686 * It causes the quantifiers to be ungreedy by default.
687 * Appending a ? turns back to greedy (!).
689 * - Added a new interceptor ijb-send-banner, which
690 * sends back the "Junkbuster" gif. Without imagelist or
691 * MSIE detection support, or if tinygif = 1, or the
692 * URL isn't recognized as an imageurl, a lame HTML
693 * explanation is sent instead.
695 * - Added new feature, which permits blocking remote
696 * script redirects and firing back a local redirect
698 * The feature is conditionally compiled, i.e. it
699 * can be disabled with --disable-fast-redirects,
700 * plus it must be activated by a "fast-redirects"
701 * line in the config file, has its own log level
702 * and of course wants to be displayed by show-proxy-args
703 * Note: Boy, all the #ifdefs in 1001 locations and
704 * all the fumbling with configure.in and acconfig.h
705 * were *way* more work than the feature itself :-(
707 * - Because a generic redirect template was needed for
708 * this, tinygif = 3 now uses the same.
710 * - Moved GIFs, and other static HTTP response templates
715 * - Removed some >400 CRs again (Jon, you really worked
718 * Revision 1.3 2001/05/20 01:21:20 jongfoster
719 * Version 2.9.4 checkin.
720 * - Merged popupfile and cookiefile, and added control over PCRS
721 * filtering, in new "permissionsfile".
722 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
723 * file error you now get a message box (in the Win32 GUI) rather
724 * than the program exiting with no explanation.
725 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
727 * - Removed tabs from "config"
728 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
729 * - Bumped up version number.
731 * Revision 1.2 2001/05/17 23:02:36 oes
732 * - Made referrer option accept 'L' as a substitute for '§'
734 * Revision 1.1.1.1 2001/05/15 13:59:01 oes
735 * Initial import of version 2.9.3 source tree
738 *********************************************************************/
745 #include <sys/types.h>
755 * Convince GNU's libc to provide a strptime prototype.
758 #endif /*__GLIBC__ */
765 #if !defined(_WIN32) && !defined(__OS2__)
771 #ifdef FEATURE_PTHREAD
773 /* jcc.h is for mutex semapores only */
774 #endif /* def FEATURE_PTHREAD */
780 #include "jbsockets.h"
781 #include "miscutil.h"
786 #ifndef HAVE_STRPTIME
787 #include "strptime.h"
790 const char parsers_h_rcs[] = PARSERS_H_VERSION;
792 /* Fix a problem with Solaris. There should be no effect on other
794 * Solaris's isspace() is a macro which uses its argument directly
795 * as an array index. Therefore we need to make sure that high-bit
796 * characters generate +ve values, and ideally we also want to make
797 * the argument match the declared parameter type of "int".
799 * Why did they write a character function that can't take a simple
800 * "char" argument? Doh!
802 #define ijb_isupper(__X) isupper((int)(unsigned char)(__X))
803 #define ijb_tolower(__X) tolower((int)(unsigned char)(__X))
805 static jb_err scan_headers(struct client_state *csp);
806 static jb_err header_tagger(struct client_state *csp, char *header);
807 static jb_err parse_header_time(const char *header_time, time_t *result);
809 static jb_err crumble (struct client_state *csp, char **header);
810 static jb_err connection (struct client_state *csp, char **header);
811 static jb_err filter_header (struct client_state *csp, char **header);
812 static jb_err client_referrer (struct client_state *csp, char **header);
813 static jb_err client_uagent (struct client_state *csp, char **header);
814 static jb_err client_ua (struct client_state *csp, char **header);
815 static jb_err client_from (struct client_state *csp, char **header);
816 static jb_err client_send_cookie (struct client_state *csp, char **header);
817 static jb_err client_x_forwarded (struct client_state *csp, char **header);
818 static jb_err client_accept_encoding (struct client_state *csp, char **header);
819 static jb_err client_te (struct client_state *csp, char **header);
820 static jb_err client_max_forwards (struct client_state *csp, char **header);
821 static jb_err client_host (struct client_state *csp, char **header);
822 static jb_err client_if_modified_since (struct client_state *csp, char **header);
823 static jb_err client_accept_language (struct client_state *csp, char **header);
824 static jb_err client_if_none_match (struct client_state *csp, char **header);
825 static jb_err crunch_client_header (struct client_state *csp, char **header);
826 static jb_err client_x_filter (struct client_state *csp, char **header);
827 static jb_err client_range (struct client_state *csp, char **header);
828 static jb_err server_set_cookie (struct client_state *csp, char **header);
829 static jb_err server_content_type (struct client_state *csp, char **header);
830 static jb_err server_content_length (struct client_state *csp, char **header);
831 static jb_err server_content_md5 (struct client_state *csp, char **header);
832 static jb_err server_content_encoding (struct client_state *csp, char **header);
833 static jb_err server_transfer_coding (struct client_state *csp, char **header);
834 static jb_err server_http (struct client_state *csp, char **header);
835 static jb_err crunch_server_header (struct client_state *csp, char **header);
836 static jb_err server_last_modified (struct client_state *csp, char **header);
837 static jb_err server_content_disposition(struct client_state *csp, char **header);
839 static jb_err client_host_adder (struct client_state *csp);
840 static jb_err client_cookie_adder (struct client_state *csp);
841 static jb_err client_xtra_adder (struct client_state *csp);
842 static jb_err connection_close_adder (struct client_state *csp);
844 static jb_err create_forged_referrer(char **header, const char *hostport);
845 static jb_err create_fake_referrer(char **header, const char *fake_referrer);
846 static jb_err handle_conditional_hide_referrer_parameter(char **header,
847 const char *host, const int parameter_conditional_block);
849 const struct parsers client_patterns[] = {
850 { "referer:", 8, client_referrer },
851 { "user-agent:", 11, client_uagent },
852 { "ua-", 3, client_ua },
853 { "from:", 5, client_from },
854 { "cookie:", 7, client_send_cookie },
855 { "x-forwarded-for:", 16, client_x_forwarded },
856 { "Accept-Encoding:", 16, client_accept_encoding },
857 { "TE:", 3, client_te },
858 { "Host:", 5, client_host },
859 { "if-modified-since:", 18, client_if_modified_since },
860 { "Keep-Alive:", 11, crumble },
861 { "connection:", 11, connection },
862 { "proxy-connection:", 17, crumble },
863 { "max-forwards:", 13, client_max_forwards },
864 { "Accept-Language:", 16, client_accept_language },
865 { "if-none-match:", 14, client_if_none_match },
866 { "Range:", 6, client_range },
867 { "If-Range:", 9, client_range },
868 { "X-Filter:", 9, client_x_filter },
869 { "*", 0, crunch_client_header },
870 { "*", 0, filter_header },
874 const struct parsers server_patterns[] = {
875 { "HTTP/", 5, server_http },
876 { "set-cookie:", 11, server_set_cookie },
877 { "connection:", 11, connection },
878 { "Content-Type:", 13, server_content_type },
879 { "Content-MD5:", 12, server_content_md5 },
880 { "Content-Encoding:", 17, server_content_encoding },
881 { "Transfer-Encoding:", 18, server_transfer_coding },
882 { "Keep-Alive:", 11, crumble },
883 { "content-disposition:", 20, server_content_disposition },
884 { "Last-Modified:", 14, server_last_modified },
885 { "*", 0, crunch_server_header },
886 { "*", 0, filter_header },
890 const struct parsers server_patterns_light[] = {
891 { "Content-Length:", 15, server_content_length },
892 { "Transfer-Encoding:", 18, server_transfer_coding },
894 { "Content-Encoding:", 17, server_content_encoding },
895 #endif /* def FEATURE_ZLIB */
899 const add_header_func_ptr add_client_headers[] = {
903 /* Temporarily disabled: client_accept_encoding_adder, */
904 connection_close_adder,
908 const add_header_func_ptr add_server_headers[] = {
909 connection_close_adder,
913 /* The vanilla wafer. */
914 static const char VANILLA_WAFER[] =
915 "NOTICE=TO_WHOM_IT_MAY_CONCERN_"
916 "Do_not_send_me_any_copyrighted_information_other_than_the_"
917 "document_that_I_am_requesting_or_any_of_its_necessary_components._"
918 "In_particular_do_not_send_me_any_cookies_that_"
919 "are_subject_to_a_claim_of_copyright_by_anybody._"
920 "Take_notice_that_I_refuse_to_be_bound_by_any_license_condition_"
921 "(copyright_or_otherwise)_applying_to_any_cookie._";
923 /*********************************************************************
925 * Function : flush_socket
927 * Description : Write any pending "buffered" content.
930 * 1 : fd = file descriptor of the socket to read
931 * 2 : csp = Current client state (buffers, headers, etc...)
933 * Returns : On success, the number of bytes written are returned (zero
934 * indicates nothing was written). On error, -1 is returned,
935 * and errno is set appropriately. If count is zero and the
936 * file descriptor refers to a regular file, 0 will be
937 * returned without causing any other effect. For a special
938 * file, the results are not portable.
940 *********************************************************************/
941 int flush_socket(jb_socket fd, struct client_state *csp)
943 struct iob *iob = csp->iob;
944 int len = iob->eod - iob->cur;
951 if (write_socket(fd, iob->cur, (size_t)len))
955 iob->eod = iob->cur = iob->buf;
961 /*********************************************************************
963 * Function : add_to_iob
965 * Description : Add content to the buffered page, expanding the
966 * buffer if necessary.
969 * 1 : csp = Current client state (buffers, headers, etc...)
970 * 2 : buf = holds the content to be added to the page
971 * 3 : n = number of bytes to be added
973 * Returns : JB_ERR_OK on success, JB_ERR_MEMORY if out-of-memory
974 * or buffer limit reached.
976 *********************************************************************/
977 jb_err add_to_iob(struct client_state *csp, char *buf, int n)
979 struct iob *iob = csp->iob;
980 size_t used, offset, need, want;
983 if (n <= 0) return JB_ERR_OK;
985 used = (size_t)(iob->eod - iob->buf);
986 offset = (size_t)(iob->cur - iob->buf);
987 need = used + (size_t)n + 1;
990 * If the buffer can't hold the new data, extend it first.
991 * Use the next power of two if possible, else use the actual need.
993 if (need > csp->config->buffer_limit)
995 log_error(LOG_LEVEL_INFO, "Buffer limit reached while extending the buffer (iob)");
996 return JB_ERR_MEMORY;
999 if (need > iob->size)
1001 for (want = csp->iob->size ? csp->iob->size : 512; want <= need;) want *= 2;
1003 if (want <= csp->config->buffer_limit && NULL != (p = (char *)realloc(iob->buf, want)))
1007 else if (NULL != (p = (char *)realloc(iob->buf, need)))
1013 log_error(LOG_LEVEL_ERROR, "Extending the buffer (iob) failed: %E");
1014 return JB_ERR_MEMORY;
1017 /* Update the iob pointers */
1018 iob->cur = p + offset;
1019 iob->eod = p + used;
1023 /* copy the new data into the iob buffer */
1024 memcpy(iob->eod, buf, (size_t)n);
1026 /* point to the end of the data */
1029 /* null terminate == cheap insurance */
1038 /*********************************************************************
1040 * Function : decompress_iob
1042 * Description : Decompress buffered page, expanding the
1043 * buffer as necessary. csp->iob->cur
1044 * should point to the the beginning of the
1045 * compressed data block.
1048 * 1 : csp = Current client state (buffers, headers, etc...)
1050 * Returns : JB_ERR_OK on success,
1051 * JB_ERR_MEMORY if out-of-memory limit reached, and
1052 * JB_ERR_COMPRESS if error decompressing buffer.
1054 *********************************************************************/
1055 jb_err decompress_iob(struct client_state *csp)
1057 char *buf; /* new, uncompressed buffer */
1058 char *cur; /* Current iob position (to keep the original
1059 * iob->cur unmodified if we return early) */
1060 size_t bufsize; /* allocated size of the new buffer */
1061 size_t old_size; /* Content size before decompression */
1062 size_t skip_size; /* Number of bytes at the beginning of the iob
1063 that we should NOT decompress. */
1064 int status; /* return status of the inflate() call */
1065 z_stream zstr; /* used by calls to zlib */
1067 assert(csp->iob->cur - csp->iob->buf > 0);
1068 assert(csp->iob->eod - csp->iob->cur > 0);
1070 bufsize = csp->iob->size;
1071 skip_size = (size_t)(csp->iob->cur - csp->iob->buf);
1072 old_size = (size_t)(csp->iob->eod - csp->iob->cur);
1074 cur = csp->iob->cur;
1079 * This is to protect the parsing of gzipped data,
1080 * but it should(?) be valid for deflated data also.
1082 log_error(LOG_LEVEL_ERROR, "Buffer too small decompressing iob");
1083 return JB_ERR_COMPRESS;
1086 if (csp->content_type & CT_GZIP)
1089 * Our task is slightly complicated by the facts that data
1090 * compressed by gzip does not include a zlib header, and
1091 * that there is no easily accessible interface in zlib to
1092 * handle a gzip header. We strip off the gzip header by
1093 * hand, and later inform zlib not to expect a header.
1097 * Strip off the gzip header. Please see RFC 1952 for more
1098 * explanation of the appropriate fields.
1100 if ((*cur++ != (char)0x1f)
1101 || (*cur++ != (char)0x8b)
1102 || (*cur++ != Z_DEFLATED))
1104 log_error(LOG_LEVEL_ERROR, "Invalid gzip header when decompressing");
1105 return JB_ERR_COMPRESS;
1111 * XXX: These magic numbers should be replaced
1112 * with macros to give a better idea what they do.
1116 /* The gzip header has reserved bits set; bail out. */
1117 log_error(LOG_LEVEL_ERROR, "Invalid gzip header flags when decompressing");
1118 return JB_ERR_COMPRESS;
1122 /* Skip extra fields if necessary. */
1126 * Skip a given number of bytes, specified
1127 * as a 16-bit little-endian value.
1130 * XXX: This code used to be:
1132 * csp->iob->cur += *csp->iob->cur++ + (*csp->iob->cur++ << 8);
1134 * which I had to change into:
1136 * cur += *cur++ + (*cur++ << 8);
1138 * at which point gcc43 finally noticed that the value
1139 * of cur is undefined (it depends on which of the
1140 * summands is evaluated first).
1142 * I haven't come across a site where this
1143 * code is actually executed yet, but I hope
1147 skip_bytes = *cur++;
1148 skip_bytes = *cur++ << 8;
1150 assert(skip_bytes == *csp->iob->cur - 2 + ((*csp->iob->cur - 1) << 8));
1153 * The number of bytes to skip should be positive
1154 * and we'd like to stay in the buffer.
1156 if ((skip_bytes < 0) || (skip_bytes >= (csp->iob->eod - cur)))
1158 log_error(LOG_LEVEL_ERROR,
1159 "Unreasonable amount of bytes to skip (%d). Stopping decompression",
1161 return JB_ERR_COMPRESS;
1163 log_error(LOG_LEVEL_INFO,
1164 "Skipping %d bytes for gzip compression. Does this sound right?",
1169 /* Skip the filename if necessary. */
1172 /* A null-terminated string is supposed to follow. */
1173 while (*cur++ && (cur < csp->iob->eod));
1177 /* Skip the comment if necessary. */
1180 /* A null-terminated string is supposed to follow. */
1181 while (*cur++ && (cur < csp->iob->eod));
1184 /* Skip the CRC if necessary. */
1190 if (cur >= csp->iob->eod)
1193 * If the current position pointer reached or passed
1194 * the buffer end, we were obviously tricked to skip
1197 log_error(LOG_LEVEL_ERROR,
1198 "Malformed gzip header detected. Aborting decompression.");
1199 return JB_ERR_COMPRESS;
1203 else if (csp->content_type & CT_DEFLATE)
1206 * XXX: The debug level should be lowered
1207 * before the next stable release.
1209 log_error(LOG_LEVEL_INFO, "Decompressing deflated iob: %d", *cur);
1211 * In theory (that is, according to RFC 1950), deflate-compressed
1212 * data should begin with a two-byte zlib header and have an
1213 * adler32 checksum at the end. It seems that in practice only
1214 * the raw compressed data is sent. Note that this means that
1215 * we are not RFC 1950-compliant here, but the advantage is that
1216 * this actually works. :)
1218 * We add a dummy null byte to tell zlib where the data ends,
1219 * and later inform it not to expect a header.
1221 * Fortunately, add_to_iob() has thoughtfully null-terminated
1222 * the buffer; we can just increment the end pointer to include
1229 log_error(LOG_LEVEL_ERROR,
1230 "Unable to determine compression format for decompression");
1231 return JB_ERR_COMPRESS;
1234 /* Set up the fields required by zlib. */
1235 zstr.next_in = (Bytef *)cur;
1236 zstr.avail_in = (unsigned int)(csp->iob->eod - cur);
1237 zstr.zalloc = Z_NULL;
1238 zstr.zfree = Z_NULL;
1239 zstr.opaque = Z_NULL;
1242 * Passing -MAX_WBITS to inflateInit2 tells the library
1243 * that there is no zlib header.
1245 if (inflateInit2 (&zstr, -MAX_WBITS) != Z_OK)
1247 log_error(LOG_LEVEL_ERROR, "Error initializing decompression");
1248 return JB_ERR_COMPRESS;
1252 * Next, we allocate new storage for the inflated data.
1253 * We don't modify the existing iob yet, so in case there
1254 * is error in decompression we can recover gracefully.
1256 buf = zalloc(bufsize);
1259 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
1260 return JB_ERR_MEMORY;
1263 assert(bufsize >= skip_size);
1264 memcpy(buf, csp->iob->buf, skip_size);
1265 zstr.avail_out = bufsize - skip_size;
1266 zstr.next_out = (Bytef *)buf + skip_size;
1268 /* Try to decompress the whole stream in one shot. */
1269 while (Z_BUF_ERROR == (status = inflate(&zstr, Z_FINISH)))
1271 /* We need to allocate more memory for the output buffer. */
1273 char *tmpbuf; /* used for realloc'ing the buffer */
1274 size_t oldbufsize = bufsize; /* keep track of the old bufsize */
1277 * If zlib wants more data then there's a problem, because
1278 * the complete compressed file should have been buffered.
1280 if (0 == zstr.avail_in)
1282 log_error(LOG_LEVEL_ERROR, "Unexpected end of compressed iob");
1283 return JB_ERR_COMPRESS;
1287 * If we tried the limit and still didn't have enough
1288 * memory, just give up.
1290 if (bufsize == csp->config->buffer_limit)
1292 log_error(LOG_LEVEL_ERROR, "Buffer limit reached while decompressing iob");
1293 return JB_ERR_MEMORY;
1296 /* Try doubling the buffer size each time. */
1299 /* Don't exceed the buffer limit. */
1300 if (bufsize > csp->config->buffer_limit)
1302 bufsize = csp->config->buffer_limit;
1305 /* Try to allocate the new buffer. */
1306 tmpbuf = realloc(buf, bufsize);
1309 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
1311 return JB_ERR_MEMORY;
1315 char *oldnext_out = (char *)zstr.next_out;
1318 * Update the fields for inflate() to use the new
1319 * buffer, which may be in a location different from
1322 zstr.avail_out += bufsize - oldbufsize;
1323 zstr.next_out = (Bytef *)tmpbuf + bufsize - zstr.avail_out;
1326 * Compare with an uglier method of calculating these values
1327 * that doesn't require the extra oldbufsize variable.
1329 assert(zstr.avail_out == tmpbuf + bufsize - (char *)zstr.next_out);
1330 assert((char *)zstr.next_out == tmpbuf + ((char *)oldnext_out - buf));
1331 assert(zstr.avail_out > 0);
1337 if (Z_STREAM_ERROR == inflateEnd(&zstr))
1339 log_error(LOG_LEVEL_ERROR,
1340 "Inconsistent stream state after decompression: %s", zstr.msg);
1342 * XXX: Intentionally no return.
1344 * According to zlib.h, Z_STREAM_ERROR is returned
1345 * "if the stream state was inconsistent".
1347 * I assume in this case inflate()'s status
1348 * would also be something different than Z_STREAM_END
1349 * so this check should be redundant, but lets see.
1353 if (status != Z_STREAM_END)
1355 /* We failed to decompress the stream. */
1356 log_error(LOG_LEVEL_ERROR,
1357 "Error in decompressing to the buffer (iob): %s", zstr.msg);
1358 return JB_ERR_COMPRESS;
1362 * Finally, we can actually update the iob, since the
1363 * decompression was successful. First, free the old
1366 freez(csp->iob->buf);
1368 /* Now, update the iob to use the new buffer. */
1369 csp->iob->buf = buf;
1370 csp->iob->cur = csp->iob->buf + skip_size;
1371 csp->iob->eod = (char *)zstr.next_out;
1372 csp->iob->size = bufsize;
1375 * Make sure the new uncompressed iob obeys some minimal
1376 * consistency conditions.
1378 if ((csp->iob->buf < csp->iob->cur)
1379 && (csp->iob->cur <= csp->iob->eod)
1380 && (csp->iob->eod <= csp->iob->buf + csp->iob->size))
1382 const size_t new_size = (size_t)(csp->iob->eod - csp->iob->cur);
1385 log_error(LOG_LEVEL_RE_FILTER,
1386 "Decompression successful. Old size: %d, new size: %d.",
1387 old_size, new_size);
1391 /* zlib thinks this is OK, so lets do the same. */
1392 log_error(LOG_LEVEL_INFO, "Decompression didn't result in any content.");
1397 /* It seems that zlib did something weird. */
1398 log_error(LOG_LEVEL_ERROR,
1399 "Unexpected error decompressing the buffer (iob): %d==%d, %d>%d, %d<%d",
1400 csp->iob->cur, csp->iob->buf + skip_size, csp->iob->eod, csp->iob->buf,
1401 csp->iob->eod, csp->iob->buf + csp->iob->size);
1402 return JB_ERR_COMPRESS;
1408 #endif /* defined(FEATURE_ZLIB) */
1411 /*********************************************************************
1413 * Function : get_header
1415 * Description : This (odd) routine will parse the csp->iob
1418 * 1 : csp = Current client state (buffers, headers, etc...)
1420 * Returns : Any one of the following:
1422 * 1) a pointer to a dynamically allocated string that contains a header line
1423 * 2) NULL indicating that the end of the header was reached
1424 * 3) "" indicating that the end of the iob was reached before finding
1425 * a complete header line.
1427 *********************************************************************/
1428 char *get_header(struct client_state *csp)
1434 if ((iob->cur == NULL)
1435 || ((p = strchr(iob->cur, '\n')) == NULL))
1437 return(""); /* couldn't find a complete header */
1442 ret = strdup(iob->cur);
1445 /* FIXME No way to handle error properly */
1446 log_error(LOG_LEVEL_FATAL, "Out of memory in get_header()");
1451 if ((q = strchr(ret, '\r')) != NULL) *q = '\0';
1453 /* is this a blank line (i.e. the end of the header) ? */
1465 /*********************************************************************
1467 * Function : get_header_value
1469 * Description : Get the value of a given header from a chained list
1470 * of header lines or return NULL if no such header is
1471 * present in the list.
1474 * 1 : header_list = pointer to list
1475 * 2 : header_name = string with name of header to look for.
1476 * Trailing colon required, capitalization
1479 * Returns : NULL if not found, else value of header
1481 *********************************************************************/
1482 char *get_header_value(const struct list *header_list, const char *header_name)
1484 struct list_entry *cur_entry;
1488 assert(header_list);
1489 assert(header_name);
1490 length = strlen(header_name);
1492 for (cur_entry = header_list->first; cur_entry ; cur_entry = cur_entry->next)
1496 if (!strncmpic(cur_entry->str, header_name, length))
1499 * Found: return pointer to start of value
1501 ret = (char *) (cur_entry->str + length);
1502 while (*ret && ijb_isspace(*ret)) ret++;
1516 /*********************************************************************
1518 * Function : scan_headers
1520 * Description : Scans headers, applies tags and updates action bits.
1523 * 1 : csp = Current client state (buffers, headers, etc...)
1525 * Returns : JB_ERR_OK
1527 *********************************************************************/
1528 static jb_err scan_headers(struct client_state *csp)
1530 struct list_entry *h; /* Header */
1531 jb_err err = JB_ERR_OK;
1533 for (h = csp->headers->first; (err == JB_ERR_OK) && (h != NULL) ; h = h->next)
1535 /* Header crunch()ed in previous run? -> ignore */
1536 if (h->str == NULL) continue;
1537 log_error(LOG_LEVEL_HEADER, "scan: %s", h->str);
1538 err = header_tagger(csp, h->str);
1545 /*********************************************************************
1549 * Description : add, delete or modify lines in the HTTP header streams.
1550 * On entry, it receives a linked list of headers space
1551 * that was allocated dynamically (both the list nodes
1552 * and the header contents).
1554 * As a side effect it frees the space used by the original
1557 * XXX: should be split to remove the first_run hack.
1560 * 1 : pats = list of patterns to match against headers
1561 * 2 : more_headers = list of functions to add more
1562 * headers (client or server)
1563 * 3 : csp = Current client state (buffers, headers, etc...)
1565 * Returns : JB_ERR_OK in case off success, or
1566 * JB_ERR_MEMORY on out-of-memory error.
1568 *********************************************************************/
1569 jb_err sed(const struct parsers pats[],
1570 const add_header_func_ptr more_headers[],
1571 struct client_state *csp)
1573 struct list_entry *p;
1574 const struct parsers *v;
1575 const add_header_func_ptr *f;
1576 jb_err err = JB_ERR_OK;
1580 * If filtering is enabled, sed is run twice,
1581 * but most of the work needs to be done only once.
1583 first_run = (more_headers != NULL ) ? 1 : 0;
1585 if (first_run) /* Parse and print */
1589 for (v = pats; (err == JB_ERR_OK) && (v->str != NULL) ; v++)
1591 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL) ; p = p->next)
1593 /* Header crunch()ed in previous run? -> ignore */
1594 if (p->str == NULL) continue;
1596 /* Does the current parser handle this header? */
1597 if ((strncmpic(p->str, v->str, v->len) == 0) || (v->len == CHECK_EVERY_HEADER_REMAINING))
1599 err = v->parser(csp, (char **)&(p->str));
1603 /* place any additional headers on the csp->headers list */
1604 for (f = more_headers; (err == JB_ERR_OK) && (*f) ; f++)
1609 else /* Parse only */
1612 * The second run is only needed if the body was modified
1613 * and the content-lenght has changed.
1615 if (strncmpic(csp->http->cmd, "HEAD", 4))
1617 /*XXX: Code duplication */
1618 for (v = pats; (err == JB_ERR_OK) && (v->str != NULL) ; v++)
1620 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL) ; p = p->next)
1622 /* Header crunch()ed in previous run? -> ignore */
1623 if (p->str == NULL) continue;
1625 /* Does the current parser handle this header? */
1626 if (strncmpic(p->str, v->str, v->len) == 0)
1628 err = v->parser(csp, (char **)&(p->str));
1640 /*********************************************************************
1642 * Function : header_tagger
1644 * Description : Executes all text substitutions from applying
1645 * tag actions and saves the result as tag.
1647 * XXX: Shares enough code with filter_header() and
1648 * pcrs_filter_response() to warrant some helper functions.
1651 * 1 : csp = Current client state (buffers, headers, etc...)
1652 * 2 : header = Header that is used as tagger input
1654 * Returns : JB_ERR_OK on success and always succeeds
1656 *********************************************************************/
1657 static jb_err header_tagger(struct client_state *csp, char *header)
1659 int wanted_filter_type;
1660 int multi_action_index;
1664 struct file_list *fl;
1665 struct re_filterfile_spec *b;
1666 struct list_entry *tag_name;
1668 int found_filters = 0;
1669 const size_t header_length = strlen(header);
1671 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1673 wanted_filter_type = FT_SERVER_HEADER_TAGGER;
1674 multi_action_index = ACTION_MULTI_SERVER_HEADER_TAGGER;
1678 wanted_filter_type = FT_CLIENT_HEADER_TAGGER;
1679 multi_action_index = ACTION_MULTI_CLIENT_HEADER_TAGGER;
1682 /* Check if there are any filters */
1683 for (i = 0; i < MAX_AF_FILES; i++)
1696 if (0 == found_filters)
1698 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regex tagging.");
1702 for (i = 0; i < MAX_AF_FILES; i++)
1705 if ((NULL == fl) || (NULL == fl->f))
1708 * Either there are no filter files
1709 * left, or this filter file just
1710 * contains no valid filters.
1712 * Continue to be sure we don't miss
1713 * valid filter files that are chained
1714 * after empty or invalid ones.
1719 /* For all filters, */
1720 for (b = fl->f; b; b = b->next)
1722 if (b->type != wanted_filter_type)
1724 /* skip the ones we don't care about, */
1727 /* leaving only taggers that could apply, of which we use the ones, */
1728 for (tag_name = csp->action->multi[multi_action_index]->first;
1729 NULL != tag_name; tag_name = tag_name->next)
1731 /* that do apply, and */
1732 if (strcmp(b->name, tag_name->str) == 0)
1734 char *modified_tag = NULL;
1736 size_t size = header_length;
1737 pcrs_job *joblist = b->joblist;
1739 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1741 if (NULL == joblist)
1743 log_error(LOG_LEVEL_RE_FILTER,
1744 "Tagger %s has empty joblist. Nothing to do.", b->name);
1748 /* execute their pcrs_joblist on the header. */
1749 for (job = joblist; NULL != job; job = job->next)
1751 const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
1755 /* Success, continue with the modified version. */
1764 /* Tagger doesn't match */
1767 /* Regex failure, log it but continue anyway. */
1768 log_error(LOG_LEVEL_ERROR,
1769 "Problems with tagger \'%s\' and header \'%s\': %s",
1770 b->name, *header, pcrs_strerror(hits));
1772 freez(modified_tag);
1776 if (b->dynamic) pcrs_free_joblist(joblist);
1778 /* If this tagger matched */
1784 * There is to technical limitation which makes
1785 * it impossible to use empty tags, but I assume
1786 * no one would do it intentionally.
1789 log_error(LOG_LEVEL_INFO,
1790 "Tagger \'%s\' created an empty tag. Ignored.",
1795 if (!list_contains_item(csp->tags, tag))
1797 if (JB_ERR_OK != enlist(csp->tags, tag))
1799 log_error(LOG_LEVEL_ERROR,
1800 "Insufficient memory to add tag \'%s\', "
1801 "based on tagger \'%s\' and header \'%s\'",
1802 tag, b->name, *header);
1806 char *action_message;
1808 * update the action bits right away, to make
1809 * tagging based on tags set by earlier taggers
1810 * of the same kind possible.
1812 if (update_action_bits_for_tag(csp, tag))
1814 action_message = "Action bits updated accordingly.";
1818 action_message = "No action bits update necessary.";
1821 log_error(LOG_LEVEL_HEADER,
1822 "Tagger \'%s\' added tag \'%s\'. %s",
1823 b->name, tag, action_message);
1828 /* XXX: Is this log-worthy? */
1829 log_error(LOG_LEVEL_HEADER,
1830 "Tagger \'%s\' didn't add tag \'%s\'. "
1831 "Tag already present", b->name, tag);
1834 } /* if the tagger matched */
1835 } /* if the tagger applies */
1836 } /* for every tagger that could apply */
1837 } /* for all filters */
1838 } /* for all filter files */
1843 /* here begins the family of parser functions that reformat header lines */
1845 /*********************************************************************
1847 * Function : filter_header
1849 * Description : Executes all text substitutions from all applying
1850 * +(server|client)-header-filter actions on the header.
1851 * Most of the code was copied from pcrs_filter_response,
1852 * including the rather short variable names
1855 * 1 : csp = Current client state (buffers, headers, etc...)
1856 * 2 : header = On input, pointer to header to modify.
1857 * On output, pointer to the modified header, or NULL
1858 * to remove the header. This function frees the
1859 * original string if necessary.
1861 * Returns : JB_ERR_OK on success and always succeeds
1863 *********************************************************************/
1864 static jb_err filter_header(struct client_state *csp, char **header)
1868 size_t size = strlen(*header);
1870 char *newheader = NULL;
1873 struct file_list *fl;
1874 struct re_filterfile_spec *b;
1875 struct list_entry *filtername;
1877 int i, found_filters = 0;
1878 int wanted_filter_type;
1879 int multi_action_index;
1881 if (csp->flags & CSP_FLAG_NO_FILTERING)
1886 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1888 wanted_filter_type = FT_SERVER_HEADER_FILTER;
1889 multi_action_index = ACTION_MULTI_SERVER_HEADER_FILTER;
1893 wanted_filter_type = FT_CLIENT_HEADER_FILTER;
1894 multi_action_index = ACTION_MULTI_CLIENT_HEADER_FILTER;
1898 * Need to check the set of re_filterfiles...
1900 for (i = 0; i < MAX_AF_FILES; i++)
1913 if (0 == found_filters)
1915 log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
1919 for (i = 0; i < MAX_AF_FILES; i++)
1922 if ((NULL == fl) || (NULL == fl->f))
1925 * Either there are no filter files
1926 * left, or this filter file just
1927 * contains no valid filters.
1929 * Continue to be sure we don't miss
1930 * valid filter files that are chained
1931 * after empty or invalid ones.
1936 * For all applying +filter actions, look if a filter by that
1937 * name exists and if yes, execute its pcrs_joblist on the
1940 for (b = fl->f; b; b = b->next)
1942 if (b->type != wanted_filter_type)
1944 /* Skip other filter types */
1948 for (filtername = csp->action->multi[multi_action_index]->first;
1949 filtername ; filtername = filtername->next)
1951 if (strcmp(b->name, filtername->str) == 0)
1953 int current_hits = 0;
1954 pcrs_job *joblist = b->joblist;
1956 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1958 if (NULL == joblist)
1960 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1964 log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
1965 *header, size, b->name);
1967 /* Apply all jobs from the joblist */
1968 for (job = joblist; NULL != job; job = job->next)
1970 matches = pcrs_execute(job, *header, size, &newheader, &size);
1973 current_hits += matches;
1974 log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
1976 *header = newheader;
1978 else if ( 0 == matches )
1980 /* Filter doesn't change header */
1986 log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s",
1987 *header, b->name, pcrs_strerror(matches));
1988 if (newheader != NULL)
1990 log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader);
1996 if (b->dynamic) pcrs_free_joblist(joblist);
1998 log_error(LOG_LEVEL_RE_FILTER, "... produced %d hits (new size %d).", current_hits, size);
1999 hits += current_hits;
2006 * Additionally checking for hits is important because if
2007 * the continue hack is triggered, server headers can
2008 * arrive empty to separate multiple heads from each other.
2010 if ((0 == size) && hits)
2012 log_error(LOG_LEVEL_HEADER, "Removing empty header %s", *header);
2020 /*********************************************************************
2022 * Function : connection
2024 * Description : Makes sure that the value of the Connection: header
2025 * is "close" and signals connection_close_adder
2029 * 1 : csp = Current client state (buffers, headers, etc...)
2030 * 2 : header = On input, pointer to header to modify.
2031 * On output, pointer to the modified header, or NULL
2032 * to remove the header. This function frees the
2033 * original string if necessary.
2035 * Returns : JB_ERR_OK on success, or
2036 * JB_ERR_MEMORY on out-of-memory error.
2038 *********************************************************************/
2039 static jb_err connection(struct client_state *csp, char **header)
2041 char *old_header = *header;
2043 /* Do we have a 'Connection: close' header? */
2044 if (strcmpic(*header, "Connection: close"))
2046 /* No, create one */
2047 *header = strdup("Connection: close");
2050 return JB_ERR_MEMORY;
2052 log_error(LOG_LEVEL_HEADER, "Replaced: \'%s\' with \'%s\'", old_header, *header);
2056 /* Signal connection_close_adder() to return early. */
2057 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
2059 csp->flags |= CSP_FLAG_SERVER_CONNECTION_CLOSE_SET;
2063 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_CLOSE_SET;
2070 /*********************************************************************
2072 * Function : crumble
2074 * Description : This is called if a header matches a pattern to "crunch"
2077 * 1 : csp = Current client state (buffers, headers, etc...)
2078 * 2 : header = On input, pointer to header to modify.
2079 * On output, pointer to the modified header, or NULL
2080 * to remove the header. This function frees the
2081 * original string if necessary.
2083 * Returns : JB_ERR_OK on success, or
2084 * JB_ERR_MEMORY on out-of-memory error.
2086 *********************************************************************/
2087 static jb_err crumble(struct client_state *csp, char **header)
2089 log_error(LOG_LEVEL_HEADER, "crumble crunched: %s!", *header);
2095 /*********************************************************************
2097 * Function : crunch_server_header
2099 * Description : Crunch server header if it matches a string supplied by the
2100 * user. Called from `sed'.
2103 * 1 : csp = Current client state (buffers, headers, etc...)
2104 * 2 : header = On input, pointer to header to modify.
2105 * On output, pointer to the modified header, or NULL
2106 * to remove the header. This function frees the
2107 * original string if necessary.
2109 * Returns : JB_ERR_OK on success and always succeeds
2111 *********************************************************************/
2112 static jb_err crunch_server_header(struct client_state *csp, char **header)
2114 const char *crunch_pattern;
2116 /* Do we feel like crunching? */
2117 if ((csp->action->flags & ACTION_CRUNCH_SERVER_HEADER))
2119 crunch_pattern = csp->action->string[ACTION_STRING_SERVER_HEADER];
2121 /* Is the current header the lucky one? */
2122 if (strstr(*header, crunch_pattern))
2124 log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern);
2133 /*********************************************************************
2135 * Function : server_content_type
2137 * Description : Set the content-type for filterable types (text/.*,
2138 * .*xml.*, javascript and image/gif) unless filtering has been
2139 * forbidden (CT_TABOO) while parsing earlier headers.
2140 * NOTE: Since text/plain is commonly used by web servers
2141 * for files whose correct type is unknown, we don't
2142 * set CT_TEXT for it.
2145 * 1 : csp = Current client state (buffers, headers, etc...)
2146 * 2 : header = On input, pointer to header to modify.
2147 * On output, pointer to the modified header, or NULL
2148 * to remove the header. This function frees the
2149 * original string if necessary.
2151 * Returns : JB_ERR_OK on success, or
2152 * JB_ERR_MEMORY on out-of-memory error.
2154 *********************************************************************/
2155 static jb_err server_content_type(struct client_state *csp, char **header)
2157 /* Remove header if it isn't the first Content-Type header */
2158 if ((csp->content_type & CT_DECLARED))
2161 * Another, slightly slower, way to see if
2162 * we already parsed another Content-Type header.
2164 assert(NULL != get_header_value(csp->headers, "Content-Type:"));
2166 log_error(LOG_LEVEL_ERROR,
2167 "Multiple Content-Type headers. Removing and ignoring: \'%s\'",
2175 * Signal that the Content-Type has been set.
2177 csp->content_type |= CT_DECLARED;
2179 if (!(csp->content_type & CT_TABOO))
2182 * XXX: The assumption that text/plain is a sign of
2183 * binary data seems to be somewhat unreasonable nowadays
2184 * and should be dropped after 3.0.8 is out.
2186 if ((strstr(*header, "text/") && !strstr(*header, "plain"))
2187 || strstr(*header, "xml")
2188 || strstr(*header, "application/x-javascript"))
2190 csp->content_type |= CT_TEXT;
2192 else if (strstr(*header, "image/gif"))
2194 csp->content_type |= CT_GIF;
2196 else if (strstr(*header, "image/jpeg"))
2198 csp->content_type |= CT_JPEG;
2203 * Are we messing with the content type?
2205 if (csp->action->flags & ACTION_CONTENT_TYPE_OVERWRITE)
2208 * Make sure the user doesn't accidently
2209 * change the content type of binary documents.
2211 if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
2214 *header = strdup("Content-Type: ");
2215 string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
2219 log_error(LOG_LEVEL_HEADER, "Insufficient memory to replace Content-Type!");
2220 return JB_ERR_MEMORY;
2222 log_error(LOG_LEVEL_HEADER, "Modified: %s!", *header);
2226 log_error(LOG_LEVEL_HEADER, "%s not replaced. "
2227 "It doesn't look like a content type that should be filtered. "
2228 "Enable force-text-mode if you know what you're doing.", *header);
2236 /*********************************************************************
2238 * Function : server_transfer_coding
2240 * Description : - Prohibit filtering (CT_TABOO) if transfer coding compresses
2241 * - Raise the CSP_FLAG_CHUNKED flag if coding is "chunked"
2242 * - Remove header if body was chunked but has been
2243 * de-chunked for filtering.
2246 * 1 : csp = Current client state (buffers, headers, etc...)
2247 * 2 : header = On input, pointer to header to modify.
2248 * On output, pointer to the modified header, or NULL
2249 * to remove the header. This function frees the
2250 * original string if necessary.
2252 * Returns : JB_ERR_OK on success, or
2253 * JB_ERR_MEMORY on out-of-memory error.
2255 *********************************************************************/
2256 static jb_err server_transfer_coding(struct client_state *csp, char **header)
2259 * Turn off pcrs and gif filtering if body compressed
2261 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2265 * XXX: Added to test if we could use CT_GZIP and CT_DEFLATE here.
2267 log_error(LOG_LEVEL_INFO, "Marking content type for %s as CT_TABOO because of %s.",
2268 csp->http->cmd, *header);
2269 #endif /* def FEATURE_ZLIB */
2270 csp->content_type = CT_TABOO;
2274 * Raise flag if body chunked
2276 if (strstr(*header, "chunked"))
2278 csp->flags |= CSP_FLAG_CHUNKED;
2281 * If the body was modified, it has been de-chunked first
2282 * and the header must be removed.
2284 * FIXME: If there is more than one transfer encoding,
2285 * only the "chunked" part should be removed here.
2287 if (csp->flags & CSP_FLAG_MODIFIED)
2289 log_error(LOG_LEVEL_HEADER, "Removing: %s", *header);
2298 /*********************************************************************
2300 * Function : server_content_encoding
2302 * Description : This function is run twice for each request,
2303 * unless FEATURE_ZLIB and filtering are disabled.
2305 * The first run is used to check if the content
2306 * is compressed, if FEATURE_ZLIB is disabled
2307 * filtering is then disabled as well, if FEATURE_ZLIB
2308 * is enabled the content is marked for decompression.
2310 * The second run is used to remove the Content-Encoding
2311 * header if the decompression was successful.
2314 * 1 : csp = Current client state (buffers, headers, etc...)
2315 * 2 : header = On input, pointer to header to modify.
2316 * On output, pointer to the modified header, or NULL
2317 * to remove the header. This function frees the
2318 * original string if necessary.
2320 * Returns : JB_ERR_OK on success, or
2321 * JB_ERR_MEMORY on out-of-memory error.
2323 *********************************************************************/
2324 static jb_err server_content_encoding(struct client_state *csp, char **header)
2327 if ((csp->flags & CSP_FLAG_MODIFIED)
2328 && (csp->content_type & (CT_GZIP | CT_DEFLATE)))
2331 * We successfully decompressed the content,
2332 * and have to clean the header now, so the
2333 * client no longer expects compressed data..
2335 * XXX: There is a difference between cleaning
2336 * and removing it completely.
2338 log_error(LOG_LEVEL_HEADER, "Crunching: %s", *header);
2341 else if (strstr(*header, "gzip"))
2343 /* Mark for gzip decompression */
2344 csp->content_type |= CT_GZIP;
2346 else if (strstr(*header, "deflate"))
2348 /* Mark for zlib decompression */
2349 csp->content_type |= CT_DEFLATE;
2351 else if (strstr(*header, "compress"))
2354 * We can't decompress this; therefore we can't filter
2357 csp->content_type |= CT_TABOO;
2359 #else /* !defined(FEATURE_ZLIB) */
2360 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2363 * Body is compressed, turn off pcrs and gif filtering.
2365 csp->content_type |= CT_TABOO;
2368 * Log a warning if the user expects the content to be filtered.
2370 if ((csp->rlist != NULL) &&
2371 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2373 log_error(LOG_LEVEL_INFO,
2374 "Compressed content detected, content filtering disabled. "
2375 "Consider recompiling Privoxy with zlib support or "
2376 "enable the prevent-compression action.");
2379 #endif /* defined(FEATURE_ZLIB) */
2386 /*********************************************************************
2388 * Function : server_content_length
2390 * Description : Adjust Content-Length header if we modified
2394 * 1 : csp = Current client state (buffers, headers, etc...)
2395 * 2 : header = On input, pointer to header to modify.
2396 * On output, pointer to the modified header, or NULL
2397 * to remove the header. This function frees the
2398 * original string if necessary.
2400 * Returns : JB_ERR_OK on success, or
2401 * JB_ERR_MEMORY on out-of-memory error.
2403 *********************************************************************/
2404 static jb_err server_content_length(struct client_state *csp, char **header)
2406 const size_t max_header_length = 80;
2408 /* Regenerate header if the content was modified. */
2409 if (csp->flags & CSP_FLAG_MODIFIED)
2412 *header = (char *) zalloc(max_header_length);
2413 if (*header == NULL)
2415 return JB_ERR_MEMORY;
2418 snprintf(*header, max_header_length, "Content-Length: %d",
2419 (int)csp->content_length);
2420 log_error(LOG_LEVEL_HEADER, "Adjusted Content-Length to %d",
2421 (int)csp->content_length);
2428 /*********************************************************************
2430 * Function : server_content_md5
2432 * Description : Crumble any Content-MD5 headers if the document was
2433 * modified. FIXME: Should we re-compute instead?
2436 * 1 : csp = Current client state (buffers, headers, etc...)
2437 * 2 : header = On input, pointer to header to modify.
2438 * On output, pointer to the modified header, or NULL
2439 * to remove the header. This function frees the
2440 * original string if necessary.
2442 * Returns : JB_ERR_OK on success, or
2443 * JB_ERR_MEMORY on out-of-memory error.
2445 *********************************************************************/
2446 static jb_err server_content_md5(struct client_state *csp, char **header)
2448 if (csp->flags & CSP_FLAG_MODIFIED)
2450 log_error(LOG_LEVEL_HEADER, "Crunching Content-MD5");
2458 /*********************************************************************
2460 * Function : server_content_disposition
2462 * Description : If enabled, blocks or modifies the "Content-Disposition" header.
2463 * Called from `sed'.
2466 * 1 : csp = Current client state (buffers, headers, etc...)
2467 * 2 : header = On input, pointer to header to modify.
2468 * On output, pointer to the modified header, or NULL
2469 * to remove the header. This function frees the
2470 * original string if necessary.
2472 * Returns : JB_ERR_OK on success, or
2473 * JB_ERR_MEMORY on out-of-memory error.
2475 *********************************************************************/
2476 static jb_err server_content_disposition(struct client_state *csp, char **header)
2481 * Are we messing with the Content-Disposition header?
2483 if ((csp->action->flags & ACTION_HIDE_CONTENT_DISPOSITION) == 0)
2489 newval = csp->action->string[ACTION_STRING_CONTENT_DISPOSITION];
2491 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2494 * Blocking content-disposition header
2496 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2503 * Replacing Content-Disposition header
2506 *header = strdup("Content-Disposition: ");
2507 string_append(header, newval);
2509 if (*header != NULL)
2511 log_error(LOG_LEVEL_HEADER,
2512 "Content-Disposition header crunched and replaced with: %s", *header);
2515 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2519 /*********************************************************************
2521 * Function : server_last_modified
2523 * Description : Changes Last-Modified header to the actual date
2524 * to help hide-if-modified-since.
2525 * Called from `sed'.
2528 * 1 : csp = Current client state (buffers, headers, etc...)
2529 * 2 : header = On input, pointer to header to modify.
2530 * On output, pointer to the modified header, or NULL
2531 * to remove the header. This function frees the
2532 * original string if necessary.
2534 * Returns : JB_ERR_OK on success, or
2535 * JB_ERR_MEMORY on out-of-memory error.
2537 *********************************************************************/
2538 static jb_err server_last_modified(struct client_state *csp, char **header)
2541 char buf[BUFFER_SIZE];
2544 #ifdef HAVE_GMTIME_R
2547 struct tm *timeptr = NULL;
2548 time_t now, last_modified;
2550 long int days, hours, minutes, seconds;
2553 * Are we messing with the Last-Modified header?
2555 if ((csp->action->flags & ACTION_OVERWRITE_LAST_MODIFIED) == 0)
2561 newval = csp->action->string[ACTION_STRING_LAST_MODIFIED];
2563 if (0 == strcmpic(newval, "block") )
2566 * Blocking Last-Modified header. Useless but why not.
2568 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2572 else if (0 == strcmpic(newval, "reset-to-request-time"))
2575 * Setting Last-Modified Header to now.
2577 get_http_time(0, buf);
2579 *header = strdup("Last-Modified: ");
2580 string_append(header, buf);
2582 if (*header == NULL)
2584 log_error(LOG_LEVEL_HEADER, "Insufficent memory. Last-Modified header got lost, boohoo.");
2588 log_error(LOG_LEVEL_HEADER, "Reset to present time: %s", *header);
2591 else if (0 == strcmpic(newval, "randomize"))
2593 const char *header_time = *header + sizeof("Last-Modified:");
2595 log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
2597 #ifdef HAVE_GMTIME_R
2598 timeptr = gmtime_r(&now, &gmt);
2599 #elif FEATURE_PTHREAD
2600 pthread_mutex_lock(&gmtime_mutex);
2601 timeptr = gmtime(&now);
2602 pthread_mutex_unlock(&gmtime_mutex);
2604 timeptr = gmtime(&now);
2606 if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
2608 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
2613 rtime = (long int)difftime(now, last_modified);
2622 log_error(LOG_LEVEL_HEADER, "Server time in the future.");
2624 rtime = pick_from_range(rtime);
2625 if (negative) rtime *= -1;
2626 last_modified += rtime;
2627 #ifdef HAVE_GMTIME_R
2628 timeptr = gmtime_r(&last_modified, &gmt);
2629 #elif FEATURE_PTHREAD
2630 pthread_mutex_lock(&gmtime_mutex);
2631 timeptr = gmtime(&last_modified);
2632 pthread_mutex_unlock(&gmtime_mutex);
2634 timeptr = gmtime(&last_modified);
2636 strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr);
2638 *header = strdup("Last-Modified: ");
2639 string_append(header, newheader);
2641 if (*header == NULL)
2643 log_error(LOG_LEVEL_ERROR, "Insufficent memory, header crunched without replacement.");
2644 return JB_ERR_MEMORY;
2647 if (LOG_LEVEL_HEADER & debug) /* Save cycles if the user isn't interested. */
2649 days = rtime / (3600 * 24);
2650 hours = rtime / 3600 % 24;
2651 minutes = rtime / 60 % 60;
2652 seconds = rtime % 60;
2654 log_error(LOG_LEVEL_HEADER, "Randomized: %s (added %d da%s %d hou%s %d minut%s %d second%s",
2655 *header, days, (days == 1) ? "y" : "ys", hours, (hours == 1) ? "r" : "rs",
2656 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
2661 log_error(LOG_LEVEL_HEADER, "Randomized ... or not. No time difference to work with.");
2670 /*********************************************************************
2672 * Function : client_accept_encoding
2674 * Description : Rewrite the client's Accept-Encoding header so that
2675 * if doesn't allow compression, if the action applies.
2676 * Note: For HTTP/1.0 the absence of the header is enough.
2679 * 1 : csp = Current client state (buffers, headers, etc...)
2680 * 2 : header = On input, pointer to header to modify.
2681 * On output, pointer to the modified header, or NULL
2682 * to remove the header. This function frees the
2683 * original string if necessary.
2685 * Returns : JB_ERR_OK on success, or
2686 * JB_ERR_MEMORY on out-of-memory error.
2688 *********************************************************************/
2689 static jb_err client_accept_encoding(struct client_state *csp, char **header)
2691 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2693 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress content");
2697 /* Temporarily disable the correct behaviour to
2698 * work around a PHP bug.
2700 * if (!strcmpic(csp->http->ver, "HTTP/1.1"))
2702 * *header = strdup("Accept-Encoding: identity;q=1.0, *;q=0");
2703 * if (*header == NULL)
2705 * return JB_ERR_MEMORY;
2716 /*********************************************************************
2718 * Function : client_te
2720 * Description : Rewrite the client's TE header so that
2721 * if doesn't allow compression, if the action applies.
2724 * 1 : csp = Current client state (buffers, headers, etc...)
2725 * 2 : header = On input, pointer to header to modify.
2726 * On output, pointer to the modified header, or NULL
2727 * to remove the header. This function frees the
2728 * original string if necessary.
2730 * Returns : JB_ERR_OK on success, or
2731 * JB_ERR_MEMORY on out-of-memory error.
2733 *********************************************************************/
2734 static jb_err client_te(struct client_state *csp, char **header)
2736 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2739 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress transfer");
2746 /*********************************************************************
2748 * Function : client_referrer
2750 * Description : Handle the "referer" config setting properly.
2751 * Called from `sed'.
2754 * 1 : csp = Current client state (buffers, headers, etc...)
2755 * 2 : header = On input, pointer to header to modify.
2756 * On output, pointer to the modified header, or NULL
2757 * to remove the header. This function frees the
2758 * original string if necessary.
2760 * Returns : JB_ERR_OK on success, or
2761 * JB_ERR_MEMORY on out-of-memory error.
2763 *********************************************************************/
2764 static jb_err client_referrer(struct client_state *csp, char **header)
2766 const char *parameter;
2767 /* booleans for parameters we have to check multiple times */
2768 int parameter_conditional_block;
2769 int parameter_conditional_forge;
2771 #ifdef FEATURE_FORCE_LOAD
2773 * Since the referrer can include the prefix even
2774 * if the request itself is non-forced, we must
2775 * clean it unconditionally.
2777 * XXX: strclean is too broad
2779 strclean(*header, FORCE_PREFIX);
2780 #endif /* def FEATURE_FORCE_LOAD */
2782 if ((csp->action->flags & ACTION_HIDE_REFERER) == 0)
2784 /* Nothing left to do */
2788 parameter = csp->action->string[ACTION_STRING_REFERER];
2789 assert(parameter != NULL);
2790 parameter_conditional_block = (0 == strcmpic(parameter, "conditional-block"));
2791 parameter_conditional_forge = (0 == strcmpic(parameter, "conditional-forge"));
2793 if (!parameter_conditional_block && !parameter_conditional_forge)
2796 * As conditional-block and conditional-forge are the only
2797 * parameters that rely on the original referrer, we can
2798 * remove it now for all the others.
2803 if (0 == strcmpic(parameter, "block"))
2805 log_error(LOG_LEVEL_HEADER, "Referer crunched!");
2808 else if (parameter_conditional_block || parameter_conditional_forge)
2810 return handle_conditional_hide_referrer_parameter(header,
2811 csp->http->hostport, parameter_conditional_block);
2813 else if (0 == strcmpic(parameter, "forge"))
2815 return create_forged_referrer(header, csp->http->hostport);
2819 /* interpret parameter as user-supplied referer to fake */
2820 return create_fake_referrer(header, parameter);
2825 /*********************************************************************
2827 * Function : client_accept_language
2829 * Description : Handle the "Accept-Language" config setting properly.
2830 * Called from `sed'.
2833 * 1 : csp = Current client state (buffers, headers, etc...)
2834 * 2 : header = On input, pointer to header to modify.
2835 * On output, pointer to the modified header, or NULL
2836 * to remove the header. This function frees the
2837 * original string if necessary.
2839 * Returns : JB_ERR_OK on success, or
2840 * JB_ERR_MEMORY on out-of-memory error.
2842 *********************************************************************/
2843 static jb_err client_accept_language(struct client_state *csp, char **header)
2848 * Are we messing with the Accept-Language?
2850 if ((csp->action->flags & ACTION_HIDE_ACCEPT_LANGUAGE) == 0)
2852 /*I don't think so*/
2856 newval = csp->action->string[ACTION_STRING_LANGUAGE];
2858 if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
2861 * Blocking Accept-Language header
2863 log_error(LOG_LEVEL_HEADER, "Crunching Accept-Language!");
2870 * Replacing Accept-Language header
2873 *header = strdup("Accept-Language: ");
2874 string_append(header, newval);
2876 if (*header == NULL)
2878 log_error(LOG_LEVEL_ERROR,
2879 "Insufficent memory. Accept-Language header crunched without replacement.");
2883 log_error(LOG_LEVEL_HEADER,
2884 "Accept-Language header crunched and replaced with: %s", *header);
2887 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2891 /*********************************************************************
2893 * Function : crunch_client_header
2895 * Description : Crunch client header if it matches a string supplied by the
2896 * user. Called from `sed'.
2899 * 1 : csp = Current client state (buffers, headers, etc...)
2900 * 2 : header = On input, pointer to header to modify.
2901 * On output, pointer to the modified header, or NULL
2902 * to remove the header. This function frees the
2903 * original string if necessary.
2905 * Returns : JB_ERR_OK on success and always succeeds
2907 *********************************************************************/
2908 static jb_err crunch_client_header(struct client_state *csp, char **header)
2910 const char *crunch_pattern;
2912 /* Do we feel like crunching? */
2913 if ((csp->action->flags & ACTION_CRUNCH_CLIENT_HEADER))
2915 crunch_pattern = csp->action->string[ACTION_STRING_CLIENT_HEADER];
2917 /* Is the current header the lucky one? */
2918 if (strstr(*header, crunch_pattern))
2920 log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern);
2928 /*********************************************************************
2930 * Function : client_uagent
2932 * Description : Handle the "user-agent" config setting properly
2933 * and remember its original value to enable browser
2934 * bug workarounds. Called from `sed'.
2937 * 1 : csp = Current client state (buffers, headers, etc...)
2938 * 2 : header = On input, pointer to header to modify.
2939 * On output, pointer to the modified header, or NULL
2940 * to remove the header. This function frees the
2941 * original string if necessary.
2943 * Returns : JB_ERR_OK on success, or
2944 * JB_ERR_MEMORY on out-of-memory error.
2946 *********************************************************************/
2947 static jb_err client_uagent(struct client_state *csp, char **header)
2951 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0)
2956 newval = csp->action->string[ACTION_STRING_USER_AGENT];
2963 *header = strdup("User-Agent: ");
2964 string_append(header, newval);
2966 log_error(LOG_LEVEL_HEADER, "Modified: %s", *header);
2968 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2972 /*********************************************************************
2974 * Function : client_ua
2976 * Description : Handle "ua-" headers properly. Called from `sed'.
2979 * 1 : csp = Current client state (buffers, headers, etc...)
2980 * 2 : header = On input, pointer to header to modify.
2981 * On output, pointer to the modified header, or NULL
2982 * to remove the header. This function frees the
2983 * original string if necessary.
2985 * Returns : JB_ERR_OK on success, or
2986 * JB_ERR_MEMORY on out-of-memory error.
2988 *********************************************************************/
2989 static jb_err client_ua(struct client_state *csp, char **header)
2991 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) != 0)
2993 log_error(LOG_LEVEL_HEADER, "crunched User-Agent!");
3001 /*********************************************************************
3003 * Function : client_from
3005 * Description : Handle the "from" config setting properly.
3006 * Called from `sed'.
3009 * 1 : csp = Current client state (buffers, headers, etc...)
3010 * 2 : header = On input, pointer to header to modify.
3011 * On output, pointer to the modified header, or NULL
3012 * to remove the header. This function frees the
3013 * original string if necessary.
3015 * Returns : JB_ERR_OK on success, or
3016 * JB_ERR_MEMORY on out-of-memory error.
3018 *********************************************************************/
3019 static jb_err client_from(struct client_state *csp, char **header)
3023 if ((csp->action->flags & ACTION_HIDE_FROM) == 0)
3030 newval = csp->action->string[ACTION_STRING_FROM];
3033 * Are we blocking the e-mail address?
3035 if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
3037 log_error(LOG_LEVEL_HEADER, "crunched From!");
3041 log_error(LOG_LEVEL_HEADER, " modified");
3043 *header = strdup("From: ");
3044 string_append(header, newval);
3046 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3050 /*********************************************************************
3052 * Function : client_send_cookie
3054 * Description : Crunches the "cookie" header if necessary.
3055 * Called from `sed'.
3057 * XXX: Stupid name, doesn't send squat.
3060 * 1 : csp = Current client state (buffers, headers, etc...)
3061 * 2 : header = On input, pointer to header to modify.
3062 * On output, pointer to the modified header, or NULL
3063 * to remove the header. This function frees the
3064 * original string if necessary.
3066 * Returns : JB_ERR_OK on success, or
3067 * JB_ERR_MEMORY on out-of-memory error.
3069 *********************************************************************/
3070 static jb_err client_send_cookie(struct client_state *csp, char **header)
3072 if (csp->action->flags & ACTION_NO_COOKIE_READ)
3074 log_error(LOG_LEVEL_HEADER, "Crunched outgoing cookie: %s", *header);
3082 /*********************************************************************
3084 * Function : client_x_forwarded
3086 * Description : Handle the "x-forwarded-for" config setting properly,
3087 * also used in the add_client_headers list. Called from `sed'.
3090 * 1 : csp = Current client state (buffers, headers, etc...)
3091 * 2 : header = On input, pointer to header to modify.
3092 * On output, pointer to the modified header, or NULL
3093 * to remove the header. This function frees the
3094 * original string if necessary.
3096 * Returns : JB_ERR_OK on success, or
3097 * JB_ERR_MEMORY on out-of-memory error.
3099 *********************************************************************/
3100 jb_err client_x_forwarded(struct client_state *csp, char **header)
3102 if ((csp->action->flags & ACTION_HIDE_FORWARDED) != 0)
3105 log_error(LOG_LEVEL_HEADER, "crunched x-forwarded-for!");
3112 /*********************************************************************
3114 * Function : client_max_forwards
3116 * Description : If the HTTP method is OPTIONS or TRACE, subtract one
3117 * from the value of the Max-Forwards header field.
3120 * 1 : csp = Current client state (buffers, headers, etc...)
3121 * 2 : header = On input, pointer to header to modify.
3122 * On output, pointer to the modified header, or NULL
3123 * to remove the header. This function frees the
3124 * original string if necessary.
3126 * Returns : JB_ERR_OK on success, or
3127 * JB_ERR_MEMORY on out-of-memory error.
3129 *********************************************************************/
3130 static jb_err client_max_forwards(struct client_state *csp, char **header)
3134 if ((0 == strcmpic(csp->http->gpc, "trace")) ||
3135 (0 == strcmpic(csp->http->gpc, "options")))
3137 assert(*(*header+12) == ':');
3138 if (1 == sscanf(*header+12, ": %u", &max_forwards))
3140 if (max_forwards > 0)
3142 snprintf(*header, strlen(*header)+1, "Max-Forwards: %u", --max_forwards);
3143 log_error(LOG_LEVEL_HEADER, "Max-Forwards value for %s request reduced to %u.",
3144 csp->http->gpc, max_forwards);
3146 else if (max_forwards < 0)
3148 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3154 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3163 /*********************************************************************
3165 * Function : client_host
3167 * Description : If the request URI did not contain host and
3168 * port information, parse and evaluate the Host
3171 * Also, kill ill-formed HOST: headers as sent by
3172 * Apple's iTunes software when used with a proxy.
3175 * 1 : csp = Current client state (buffers, headers, etc...)
3176 * 2 : header = On input, pointer to header to modify.
3177 * On output, pointer to the modified header, or NULL
3178 * to remove the header. This function frees the
3179 * original string if necessary.
3181 * Returns : JB_ERR_OK on success, or
3182 * JB_ERR_MEMORY on out-of-memory error.
3184 *********************************************************************/
3185 static jb_err client_host(struct client_state *csp, char **header)
3190 * If the header field name is all upper-case, chances are that it's
3191 * an ill-formed one from iTunes. BTW, killing innocent headers here is
3192 * not a problem -- they are regenerated later.
3194 if ((*header)[1] == 'O')
3196 log_error(LOG_LEVEL_HEADER, "Killed all-caps Host header line: %s", *header);
3201 if (!csp->http->hostport || (*csp->http->hostport == '*') ||
3202 *csp->http->hostport == ' ' || *csp->http->hostport == '\0')
3205 if (NULL == (p = strdup((*header)+6)))
3207 return JB_ERR_MEMORY;
3210 if (NULL == (q = strdup(p)))
3213 return JB_ERR_MEMORY;
3216 freez(csp->http->hostport);
3217 csp->http->hostport = p;
3218 freez(csp->http->host);
3219 csp->http->host = q;
3220 q = strchr(csp->http->host, ':');
3223 /* Terminate hostname and evaluate port string */
3225 csp->http->port = atoi(q);
3229 csp->http->port = csp->http->ssl ? 443 : 80;
3232 log_error(LOG_LEVEL_HEADER, "New host and port from Host field: %s = %s:%d",
3233 csp->http->hostport, csp->http->host, csp->http->port);
3236 /* Signal client_host_adder() to return right away */
3237 csp->flags |= CSP_FLAG_HOST_HEADER_IS_SET;
3243 /*********************************************************************
3245 * Function : client_if_modified_since
3247 * Description : Remove or modify the If-Modified-Since header.
3250 * 1 : csp = Current client state (buffers, headers, etc...)
3251 * 2 : header = On input, pointer to header to modify.
3252 * On output, pointer to the modified header, or NULL
3253 * to remove the header. This function frees the
3254 * original string if necessary.
3256 * Returns : JB_ERR_OK on success, or
3257 * JB_ERR_MEMORY on out-of-memory error.
3259 *********************************************************************/
3260 static jb_err client_if_modified_since(struct client_state *csp, char **header)
3263 #ifdef HAVE_GMTIME_R
3266 struct tm *timeptr = NULL;
3270 long int hours, minutes, seconds;
3274 if ( 0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
3277 * The client got an error message because of a temporary problem,
3278 * the problem is gone and the client now tries to revalidate our
3279 * error message on the real server. The revalidation would always
3280 * end with the transmission of the whole document and there is
3281 * no need to expose the bogus If-Modified-Since header.
3283 log_error(LOG_LEVEL_HEADER, "Crunching useless If-Modified-Since header.");
3286 else if (csp->action->flags & ACTION_HIDE_IF_MODIFIED_SINCE)
3288 newval = csp->action->string[ACTION_STRING_IF_MODIFIED_SINCE];
3290 if ((0 == strcmpic(newval, "block")))
3292 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3295 else /* add random value */
3297 const char *header_time = *header + sizeof("If-Modified-Since:");
3299 if (JB_ERR_OK != parse_header_time(header_time, &tm))
3301 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
3306 rtime = strtol(newval, &endptr, 0);
3309 log_error(LOG_LEVEL_HEADER, "Randomizing: %s (random range: %d minut%s)",
3310 *header, rtime, (rtime == 1 || rtime == -1) ? "e": "es");
3317 rtime = pick_from_range(rtime);
3321 log_error(LOG_LEVEL_ERROR, "Random range is 0. Assuming time transformation test.",
3324 tm += rtime * (negative ? -1 : 1);
3325 #ifdef HAVE_GMTIME_R
3326 timeptr = gmtime_r(&tm, &gmt);
3327 #elif FEATURE_PTHREAD
3328 pthread_mutex_lock(&gmtime_mutex);
3329 timeptr = gmtime(&tm);
3330 pthread_mutex_unlock(&gmtime_mutex);
3332 timeptr = gmtime(&tm);
3334 strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr);
3337 *header = strdup("If-Modified-Since: ");
3338 string_append(header, newheader);
3340 if (*header == NULL)
3342 log_error(LOG_LEVEL_HEADER, "Insufficent memory, header crunched without replacement.");
3343 return JB_ERR_MEMORY;
3346 if (LOG_LEVEL_HEADER & debug) /* Save cycles if the user isn't interested. */
3348 hours = rtime / 3600;
3349 minutes = rtime / 60 % 60;
3350 seconds = rtime % 60;
3352 log_error(LOG_LEVEL_HEADER, "Randomized: %s (%s %d hou%s %d minut%s %d second%s",
3353 *header, (negative) ? "subtracted" : "added", hours, (hours == 1) ? "r" : "rs",
3354 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
3364 /*********************************************************************
3366 * Function : client_if_none_match
3368 * Description : Remove the If-None-Match header.
3371 * 1 : csp = Current client state (buffers, headers, etc...)
3372 * 2 : header = On input, pointer to header to modify.
3373 * On output, pointer to the modified header, or NULL
3374 * to remove the header. This function frees the
3375 * original string if necessary.
3377 * Returns : JB_ERR_OK on success, or
3378 * JB_ERR_MEMORY on out-of-memory error.
3380 *********************************************************************/
3381 static jb_err client_if_none_match(struct client_state *csp, char **header)
3383 if (csp->action->flags & ACTION_CRUNCH_IF_NONE_MATCH)
3385 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3393 /*********************************************************************
3395 * Function : client_x_filter
3397 * Description : Disables filtering if the client set "X-Filter: No".
3398 * Called from `sed'.
3401 * 1 : csp = Current client state (buffers, headers, etc...)
3402 * 2 : header = On input, pointer to header to modify.
3403 * On output, pointer to the modified header, or NULL
3404 * to remove the header. This function frees the
3405 * original string if necessary.
3407 * Returns : JB_ERR_OK on success
3409 *********************************************************************/
3410 jb_err client_x_filter(struct client_state *csp, char **header)
3412 if ( 0 == strcmpic(*header, "X-Filter: No"))
3414 if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE))
3416 log_error(LOG_LEVEL_INFO, "Ignored the client's request to fetch without filtering.");
3420 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
3422 log_error(LOG_LEVEL_HEADER,
3423 "force-text-mode overruled the client's request to fetch without filtering!");
3427 csp->content_type = CT_TABOO; /* XXX: This hack shouldn't be necessary */
3428 csp->flags |= CSP_FLAG_NO_FILTERING;
3429 log_error(LOG_LEVEL_HEADER, "Accepted the client's request to fetch without filtering.");
3431 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3439 /*********************************************************************
3441 * Function : client_range
3443 * Description : Removes Range and If-Range headers if content
3444 * filtering is enabled. If the client's version of
3445 * the document has been altered by Privoxy, the server
3446 * could interpret the range differently than the client
3447 * intended in which case the user could end up with
3448 * corrupted content.
3451 * 1 : csp = Current client state (buffers, headers, etc...)
3452 * 2 : header = On input, pointer to header to modify.
3453 * On output, pointer to the modified header, or NULL
3454 * to remove the header. This function frees the
3455 * original string if necessary.
3457 * Returns : JB_ERR_OK
3459 *********************************************************************/
3460 static jb_err client_range(struct client_state *csp, char **header)
3462 if (content_filters_enabled(csp))
3464 log_error(LOG_LEVEL_HEADER, "Content filtering is enabled."
3465 " Crunching: \'%s\' to prevent range-mismatch problems.", *header);
3472 /* the following functions add headers directly to the header list */
3474 /*********************************************************************
3476 * Function : client_host_adder
3478 * Description : Adds the Host: header field if it is missing.
3479 * Called from `sed'.
3482 * 1 : csp = Current client state (buffers, headers, etc...)
3484 * Returns : JB_ERR_OK on success, or
3485 * JB_ERR_MEMORY on out-of-memory error.
3487 *********************************************************************/
3488 static jb_err client_host_adder(struct client_state *csp)
3493 if (csp->flags & CSP_FLAG_HOST_HEADER_IS_SET)
3495 /* Header already set by the client, nothing to do. */
3499 if ( !csp->http->hostport || !*(csp->http->hostport))
3501 /* XXX: When does this happen and why is it OK? */
3502 log_error(LOG_LEVEL_INFO, "Weirdness in client_host_adder detected and ignored.");
3507 * remove 'user:pass@' from 'proto://user:pass@host'
3509 if ( (p = strchr( csp->http->hostport, '@')) != NULL )
3515 p = csp->http->hostport;
3518 /* XXX: Just add it, we already made sure that it will be unique */
3519 log_error(LOG_LEVEL_HEADER, "addh-unique: Host: %s", p);
3520 err = enlist_unique_header(csp->headers, "Host", p);
3526 /*********************************************************************
3528 * Function : client_cookie_adder
3530 * Description : Used in the add_client_headers list to add "wafers".
3531 * Called from `sed'.
3534 * 1 : csp = Current client state (buffers, headers, etc...)
3536 * Returns : JB_ERR_OK on success, or
3537 * JB_ERR_MEMORY on out-of-memory error.
3539 *********************************************************************/
3540 jb_err client_cookie_adder(struct client_state *csp)
3543 struct list_entry *wafer;
3544 struct list_entry *wafer_list;
3548 * If the user has not supplied any wafers, and the user has not
3549 * told us to suppress the vanilla wafer, then send the vanilla wafer.
3551 if ((0 != (csp->action->flags & ACTION_VANILLA_WAFER))
3552 && list_is_empty(csp->action->multi[ACTION_MULTI_WAFER]))
3554 enlist(csp->action->multi[ACTION_MULTI_WAFER], VANILLA_WAFER);
3557 wafer_list = csp->action->multi[ACTION_MULTI_WAFER]->first;
3559 if (NULL == wafer_list)
3565 tmp = strdup("Cookie: ");
3567 for (wafer = wafer_list; (NULL != tmp) && (NULL != wafer); wafer = wafer->next)
3569 if (wafer != wafer_list)
3571 /* As this isn't the first wafer, we need a delimiter. */
3572 string_append(&tmp, "; ");
3574 string_join(&tmp, cookie_encode(wafer->str));
3579 return JB_ERR_MEMORY;
3582 log_error(LOG_LEVEL_HEADER, "addh: %s", tmp);
3583 err = enlist(csp->headers, tmp);
3590 /*********************************************************************
3592 * Function : client_accept_encoding_adder
3594 * Description : Add an Accept-Encoding header to the client's request
3595 * that disables compression if the action applies, and
3596 * the header is not already there. Called from `sed'.
3597 * Note: For HTTP/1.0, the absence of the header is enough.
3600 * 1 : csp = Current client state (buffers, headers, etc...)
3602 * Returns : JB_ERR_OK on success, or
3603 * JB_ERR_MEMORY on out-of-memory error.
3605 *********************************************************************/
3606 static jb_err client_accept_encoding_adder(struct client_state *csp)
3608 if ( ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
3609 && (!strcmpic(csp->http->ver, "HTTP/1.1")) )
3611 return enlist_unique(csp->headers, "Accept-Encoding: identity;q=1.0, *;q=0", 16);
3619 /*********************************************************************
3621 * Function : client_xtra_adder
3623 * Description : Used in the add_client_headers list. Called from `sed'.
3626 * 1 : csp = Current client state (buffers, headers, etc...)
3628 * Returns : JB_ERR_OK on success, or
3629 * JB_ERR_MEMORY on out-of-memory error.
3631 *********************************************************************/
3632 static jb_err client_xtra_adder(struct client_state *csp)
3634 struct list_entry *lst;
3637 for (lst = csp->action->multi[ACTION_MULTI_ADD_HEADER]->first;
3638 lst ; lst = lst->next)
3640 log_error(LOG_LEVEL_HEADER, "addh: %s", lst->str);
3641 err = enlist(csp->headers, lst->str);
3653 /*********************************************************************
3655 * Function : connection_close_adder
3657 * Description : "Temporary" fix for the needed but missing HTTP/1.1
3658 * support. Adds a "Connection: close" header to csp->headers
3659 * unless the header was already present. Called from `sed'.
3661 * FIXME: This whole function shouldn't be neccessary!
3664 * 1 : csp = Current client state (buffers, headers, etc...)
3666 * Returns : JB_ERR_OK on success, or
3667 * JB_ERR_MEMORY on out-of-memory error.
3669 *********************************************************************/
3670 static jb_err connection_close_adder(struct client_state *csp)
3672 const unsigned int flags = csp->flags;
3675 * Return right away if
3677 * - we're parsing server headers and the server header
3678 * "Connection: close" is already set, or if
3680 * - we're parsing client headers and the client header
3681 * "Connection: close" is already set.
3683 if ((flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE
3684 && flags & CSP_FLAG_SERVER_CONNECTION_CLOSE_SET)
3685 ||(!(flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3686 && flags & CSP_FLAG_CLIENT_CONNECTION_CLOSE_SET))
3691 log_error(LOG_LEVEL_HEADER, "Adding: Connection: close");
3693 return enlist(csp->headers, "Connection: close");
3697 /*********************************************************************
3699 * Function : server_http
3701 * Description : - Save the HTTP Status into csp->http->status
3702 * - Set CT_TABOO to prevent filtering if the answer
3703 * is a partial range (HTTP status 206)
3704 * - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade
3708 * 1 : csp = Current client state (buffers, headers, etc...)
3709 * 2 : header = On input, pointer to header to modify.
3710 * On output, pointer to the modified header, or NULL
3711 * to remove the header. This function frees the
3712 * original string if necessary.
3714 * Returns : JB_ERR_OK on success, or
3715 * JB_ERR_MEMORY on out-of-memory error.
3717 *********************************************************************/
3718 static jb_err server_http(struct client_state *csp, char **header)
3720 sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status));
3721 if (csp->http->status == 206)
3723 csp->content_type = CT_TABOO;
3726 if ((csp->action->flags & ACTION_DOWNGRADE) != 0)
3728 /* XXX: Should we do a real validity check here? */
3729 if (strlen(*header) > 8)
3732 log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0");
3737 * XXX: Should we block the request or
3738 * enlist a valid status code line here?
3740 log_error(LOG_LEVEL_INFO, "Malformed server response detected. "
3741 "Downgrading to HTTP/1.0 impossible.");
3749 /*********************************************************************
3751 * Function : server_set_cookie
3753 * Description : Handle the server "cookie" header properly.
3754 * Log cookie to the jar file. Then "crunch",
3755 * accept or rewrite it to a session cookie.
3756 * Called from `sed'.
3758 * TODO: Allow the user to specify a new expiration
3759 * time to cause the cookie to expire even before the
3760 * browser is closed.
3763 * 1 : csp = Current client state (buffers, headers, etc...)
3764 * 2 : header = On input, pointer to header to modify.
3765 * On output, pointer to the modified header, or NULL
3766 * to remove the header. This function frees the
3767 * original string if necessary.
3769 * Returns : JB_ERR_OK on success, or
3770 * JB_ERR_MEMORY on out-of-memory error.
3772 *********************************************************************/
3773 static jb_err server_set_cookie(struct client_state *csp, char **header)
3780 #ifdef FEATURE_COOKIE_JAR
3781 if (csp->config->jar)
3784 * Write timestamp into outbuf.
3786 * Complex because not all OSs have tm_gmtoff or
3787 * the %z field in strftime()
3789 char tempbuf[ BUFFER_SIZE ];
3791 #ifdef HAVE_LOCALTIME_R
3792 tm_now = *localtime_r(&now, &tm_now);
3793 #elif FEATURE_PTHREAD
3794 pthread_mutex_lock(&localtime_mutex);
3795 tm_now = *localtime (&now);
3796 pthread_mutex_unlock(&localtime_mutex);
3798 tm_now = *localtime (&now);
3800 strftime(tempbuf, BUFFER_SIZE-6, "%b %d %H:%M:%S ", &tm_now);
3802 /* strlen("set-cookie: ") = 12 */
3803 fprintf(csp->config->jar, "%s %s\t%s\n", tempbuf, csp->http->host, *header + 12);
3805 #endif /* def FEATURE_COOKIE_JAR */
3807 if ((csp->action->flags & ACTION_NO_COOKIE_SET) != 0)
3809 log_error(LOG_LEVEL_HEADER, "Crunching incoming cookie: %s", *header);
3812 else if ((csp->action->flags & ACTION_NO_COOKIE_KEEP) != 0)
3814 /* Flag whether or not to log a message */
3817 /* A variable to store the tag we're working on */
3820 /* Skip "Set-Cookie:" (11 characters) in header */
3821 cur_tag = *header + 11;
3823 /* skip whitespace between "Set-Cookie:" and value */
3824 while (*cur_tag && ijb_isspace(*cur_tag))
3829 /* Loop through each tag in the cookie */
3833 char *next_tag = strchr(cur_tag, ';');
3834 if (next_tag != NULL)
3836 /* Skip the ';' character itself */
3839 /* skip whitespace ";" and start of tag */
3840 while (*next_tag && ijb_isspace(*next_tag))
3847 /* "Next tag" is the end of the string */
3848 next_tag = cur_tag + strlen(cur_tag);
3852 * Check the expiration date to see
3853 * if the cookie is still valid, if yes,
3854 * rewrite it to a session cookie.
3856 if ((strncmpic(cur_tag, "expires=", 8) == 0) && *(cur_tag + 8))
3858 char *expiration_date = cur_tag + 8; /* Skip "[Ee]xpires=" */
3860 /* Did we detect the date properly? */
3861 if (JB_ERR_OK != parse_header_time(expiration_date, &cookie_time))
3864 * Nope, treat it as if it was still valid.
3866 * XXX: Should we remove the whole cookie instead?
3868 log_error(LOG_LEVEL_ERROR,
3869 "Can't parse \'%s\', send by %s. Unsupported time format?", cur_tag, csp->http->url);
3870 memmove(cur_tag, next_tag, strlen(next_tag) + 1);
3876 * Yes. Check if the cookie is still valid.
3878 * If the cookie is already expired it's probably
3879 * a delete cookie and even if it isn't, the browser
3880 * will discard it anyway.
3884 * XXX: timegm() isn't available on some AmigaOS
3885 * versions and our replacement doesn't work.
3887 * Our options are to either:
3889 * - disable session-cookies-only completely if timegm
3892 * - to simply remove all expired tags, like it has
3893 * been done until Privoxy 3.0.6 and to live with
3894 * the consequence that it can cause login/logout
3895 * problems on servers that don't validate their
3896 * input properly, or
3898 * - to replace it with mktime in which
3899 * case there is a slight chance of valid cookies
3900 * passing as already expired.
3902 * This is the way it's currently done and it's not
3903 * as bad as it sounds. If the missing GMT offset is
3904 * enough to change the result of the expiration check
3905 * the cookie will be only valid for a few hours
3906 * anyway, which in many cases will be shorter
3907 * than a browser session.
3909 if (cookie_time - now < 0)
3911 log_error(LOG_LEVEL_HEADER,
3912 "Cookie \'%s\' is already expired and can pass unmodified.", *header);
3913 /* Just in case some clown sets more then one expiration date */
3919 * Still valid, delete expiration date by copying
3920 * the rest of the string over it.
3922 * (Note that we cannot just use "strcpy(cur_tag, next_tag)",
3923 * since the behaviour of strcpy is undefined for overlapping
3926 memmove(cur_tag, next_tag, strlen(next_tag) + 1);
3928 /* That changed the header, need to issue a log message */
3932 * Note that the next tag has now been moved to *cur_tag,
3933 * so we do not need to update the cur_tag pointer.
3941 /* Move on to next cookie tag */
3948 assert(NULL != *header);
3949 log_error(LOG_LEVEL_HEADER, "Cookie rewritten to a temporary one: %s",
3958 #ifdef FEATURE_FORCE_LOAD
3959 /*********************************************************************
3961 * Function : strclean
3963 * Description : In-Situ-Eliminate all occurances of substring in
3967 * 1 : string = string to clean
3968 * 2 : substring = substring to eliminate
3970 * Returns : Number of eliminations
3972 *********************************************************************/
3973 int strclean(const char *string, const char *substring)
3979 len = strlen(substring);
3981 while((pos = strstr(string, substring)) != NULL)
3988 while (*p++ != '\0');
3995 #endif /* def FEATURE_FORCE_LOAD */
3998 /*********************************************************************
4000 * Function : parse_header_time
4002 * Description : Parses time formats used in HTTP header strings
4003 * to get the numerical respresentation.
4006 * 1 : header_time = HTTP header time as string.
4007 * 2 : result = storage for header_time in seconds
4009 * Returns : JB_ERR_OK if the time format was recognized, or
4010 * JB_ERR_PARSE otherwise.
4012 *********************************************************************/
4013 static jb_err parse_header_time(const char *header_time, time_t *result)
4018 * Zero out gmt to prevent time zone offsets.
4020 * While this is only necessary on some platforms
4021 * (mingw32 for example), I don't know how to
4022 * detect these automatically and doing it everywhere
4025 memset(&gmt, 0, sizeof(gmt));
4027 /* Tue, 02 Jun 2037 20:00:00 */
4028 if ((NULL == strptime(header_time, "%a, %d %b %Y %H:%M:%S", &gmt))
4029 /* Tue, 02-Jun-2037 20:00:00 */
4030 && (NULL == strptime(header_time, "%a, %d-%b-%Y %H:%M:%S", &gmt))
4031 /* Tue, 02-Jun-37 20:00:00 */
4032 && (NULL == strptime(header_time, "%a, %d-%b-%y %H:%M:%S", &gmt))
4033 /* Tuesday, 02-Jun-2037 20:00:00 */
4034 && (NULL == strptime(header_time, "%A, %d-%b-%Y %H:%M:%S", &gmt))
4035 /* Tuesday Jun 02 20:00:00 2037 */
4036 && (NULL == strptime(header_time, "%A %b %d %H:%M:%S %Y", &gmt)))
4038 return JB_ERR_PARSE;
4041 *result = timegm(&gmt);
4048 /*********************************************************************
4050 * Function : get_destination_from_headers
4052 * Description : Parse the "Host:" header to get the request's destination.
4053 * Only needed if the client's request was forcefully
4054 * redirected into Privoxy.
4056 * Code mainly copied from client_host() which is currently
4057 * run too late for this purpose.
4060 * 1 : headers = List of headers (one of them hopefully being
4061 * the "Host:" header)
4062 * 2 : http = storage for the result (host, port and hostport).
4064 * Returns : JB_ERR_MEMORY in case of memory problems,
4065 * JB_ERR_PARSE if the host header couldn't be found,
4066 * JB_ERR_OK otherwise.
4068 *********************************************************************/
4069 jb_err get_destination_from_headers(const struct list *headers, struct http_request *http)
4075 host = get_header_value(headers, "Host:");
4079 log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found.");
4080 return JB_ERR_PARSE;
4083 if (NULL == (p = strdup((host))))
4085 log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
4086 return JB_ERR_MEMORY;
4089 if (NULL == (q = strdup(p)))
4092 log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
4093 return JB_ERR_MEMORY;
4096 freez(http->hostport);
4100 q = strchr(http->host, ':');
4103 /* Terminate hostname and evaluate port string */
4105 http->port = atoi(q);
4109 http->port = http->ssl ? 443 : 80;
4112 /* Rebuild request URL */
4114 http->url = strdup(http->ssl ? "https://" : "http://");
4115 string_append(&http->url, http->hostport);
4116 string_append(&http->url, http->path);
4117 if (http->url == NULL)
4119 return JB_ERR_MEMORY;
4122 log_error(LOG_LEVEL_HEADER, "Destination extracted from \"Host:\" header. New request URL: %s",
4130 /*********************************************************************
4132 * Function : create_forged_referrer
4134 * Description : Helper for client_referrer to forge a referer as
4135 * 'http://[hostname:port/' to fool stupid
4136 * checks for in-site links
4139 * 1 : header = Pointer to header pointer
4140 * 2 : hostport = Host and optionally port as string
4142 * Returns : JB_ERR_OK in case of success, or
4143 * JB_ERR_MEMORY in case of memory problems.
4145 *********************************************************************/
4146 static jb_err create_forged_referrer(char **header, const char *hostport)
4148 assert(NULL == *header);
4150 *header = strdup("Referer: http://");
4151 string_append(header, hostport);
4152 string_append(header, "/");
4154 if (NULL == *header)
4156 return JB_ERR_MEMORY;
4159 log_error(LOG_LEVEL_HEADER, "Referer forged to: %s", *header);
4166 /*********************************************************************
4168 * Function : create_fake_referrer
4170 * Description : Helper for client_referrer to create a fake referrer
4171 * based on a string supplied by the user.
4174 * 1 : header = Pointer to header pointer
4175 * 2 : hosthost = Referrer to fake
4177 * Returns : JB_ERR_OK in case of success, or
4178 * JB_ERR_MEMORY in case of memory problems.
4180 *********************************************************************/
4181 static jb_err create_fake_referrer(char **header, const char *fake_referrer)
4183 assert(NULL == *header);
4185 if ((0 != strncmpic(fake_referrer, "http://", 7)) && (0 != strncmpic(fake_referrer, "https://", 8)))
4187 log_error(LOG_LEVEL_HEADER,
4188 "Parameter: +hide-referrer{%s} is a bad idea, but I don't care.", fake_referrer);
4190 *header = strdup("Referer: ");
4191 string_append(header, fake_referrer);
4193 if (NULL == *header)
4195 return JB_ERR_MEMORY;
4198 log_error(LOG_LEVEL_HEADER, "Referer replaced with: %s", *header);
4205 /*********************************************************************
4207 * Function : handle_conditional_hide_referrer_parameter
4209 * Description : Helper for client_referrer to crunch or forge
4210 * the referrer header if the host has changed.
4213 * 1 : header = Pointer to header pointer
4214 * 2 : host = The target host (may include the port)
4215 * 3 : parameter_conditional_block = Boolean to signal
4216 * if we're in conditional-block mode. If not set,
4217 * we're in conditional-forge mode.
4219 * Returns : JB_ERR_OK in case of success, or
4220 * JB_ERR_MEMORY in case of memory problems.
4222 *********************************************************************/
4223 static jb_err handle_conditional_hide_referrer_parameter(char **header,
4224 const char *host, const int parameter_conditional_block)
4226 char *referer = strdup(*header);
4227 const size_t hostlenght = strlen(host);
4229 if (NULL == referer)
4232 return JB_ERR_MEMORY;
4235 /* referer begins with 'Referer: http[s]://' */
4236 if (hostlenght < (strlen(referer)-17))
4239 * Shorten referer to make sure the referer is blocked
4240 * if www.example.org/www.example.com-shall-see-the-referer/
4241 * links to www.example.com/
4243 referer[hostlenght+17] = '\0';
4245 if (NULL == strstr(referer, host))
4247 /* Host has changed */
4248 if (parameter_conditional_block)
4250 log_error(LOG_LEVEL_HEADER, "New host is: %s. Crunching %s!", host, *header);
4257 return create_forged_referrer(header, host);