1 const char parsers_rcs[] = "$Id: parsers.c,v 1.120 2008/01/04 17:43:45 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 1.120 2008/01/04 17:43:45 fabiankeil
48 * Improve the warning messages that get logged if the action files
49 * "enable" filters but no filters of that type have been loaded.
51 * Revision 1.119 2007/12/28 18:32:51 fabiankeil
52 * In server_content_type():
53 * - Don't require leading white space when detecting image content types.
54 * - Change '... not replaced ...' message to sound less crazy if the text
55 * type actually is 'text/plain'.
56 * - Mark the 'text/plain == binary data' assumption for removal.
57 * - Remove a bunch of trailing white space.
59 * Revision 1.118 2007/12/28 16:56:35 fabiankeil
60 * Minor server_content_disposition() changes:
61 * - Don't regenerate the header name all lower-case.
62 * - Some white space fixes.
63 * - Remove useless log message in case of ENOMEM.
65 * Revision 1.117 2007/12/06 18:11:50 fabiankeil
66 * Garbage-collect the code to add a X-Forwarded-For
67 * header as it seems to be mostly used by accident.
69 * Revision 1.116 2007/12/01 13:04:22 fabiankeil
70 * Fix a crash on mingw32 with some Last Modified times in the future.
72 * Revision 1.115 2007/11/02 16:52:50 fabiankeil
73 * Remove a "can't happen" error block which, over
74 * time, mutated into a "guaranteed to happen" block.
76 * Revision 1.114 2007/10/19 16:56:26 fabiankeil
77 * - Downgrade "Buffer limit reached" message to LOG_LEVEL_INFO.
78 * - Use shiny new content_filters_enabled() in client_range().
80 * Revision 1.113 2007/10/10 17:29:57 fabiankeil
81 * I forgot about Poland.
83 * Revision 1.112 2007/10/09 16:38:40 fabiankeil
84 * Remove Range and If-Range headers if content filtering is enabled.
86 * Revision 1.111 2007/10/04 18:07:00 fabiankeil
87 * Move ACTION_VANILLA_WAFER handling from jcc's chat() into
88 * client_cookie_adder() to make sure send-vanilla-wafer can be
89 * controlled through tags (and thus regression-tested).
91 * Revision 1.110 2007/09/29 10:42:37 fabiankeil
92 * - Remove "scanning headers for" log message again.
93 * - Some more whitespace fixes.
95 * Revision 1.109 2007/09/08 14:25:48 fabiankeil
96 * Refactor client_referrer() and add conditional-forge parameter.
98 * Revision 1.108 2007/08/28 18:21:03 fabiankeil
99 * A bunch of whitespace fixes, pointy hat to me.
101 * Revision 1.107 2007/08/28 18:16:32 fabiankeil
102 * Fix possible memory corruption in server_http, make sure it's not
103 * executed for ordinary server headers and mark some problems for later.
105 * Revision 1.106 2007/08/18 14:30:32 fabiankeil
106 * Let content-type-overwrite{} honour force-text-mode again.
108 * Revision 1.105 2007/08/11 14:49:49 fabiankeil
109 * - Add prototpyes for the header parsers and make them static.
110 * - Comment out client_accept_encoding_adder() which isn't used right now.
112 * Revision 1.104 2007/07/14 07:38:19 fabiankeil
113 * Move the ACTION_FORCE_TEXT_MODE check out of
114 * server_content_type(). Signal other functions
115 * whether or not a content type has been declared.
116 * Part of the fix for BR#1750917.
118 * Revision 1.103 2007/06/01 16:31:54 fabiankeil
119 * Change sed() to return a jb_err in preparation for forward-override{}.
121 * Revision 1.102 2007/05/27 12:39:32 fabiankeil
122 * Adjust "X-Filter: No" to disable dedicated header filters.
124 * Revision 1.101 2007/05/14 10:16:41 fabiankeil
125 * Streamline client_cookie_adder().
127 * Revision 1.100 2007/04/30 15:53:11 fabiankeil
128 * Make sure filters with dynamic jobs actually use them.
130 * Revision 1.99 2007/04/30 15:06:26 fabiankeil
131 * - Introduce dynamic pcrs jobs that can resolve variables.
132 * - Remove unnecessary update_action_bits_for_all_tags() call.
134 * Revision 1.98 2007/04/17 18:32:10 fabiankeil
135 * - Make tagging based on tags set by earlier taggers
136 * of the same kind possible.
137 * - Log whether or not new tags cause action bits updates
138 * (in which case a matching tag-pattern section exists).
139 * - Log if the user tries to set a tag that is already set.
141 * Revision 1.97 2007/04/15 16:39:21 fabiankeil
142 * Introduce tags as alternative way to specify which
143 * actions apply to a request. At the moment tags can be
144 * created based on client and server headers.
146 * Revision 1.96 2007/04/12 12:53:58 fabiankeil
147 * Log a warning if the content is compressed, filtering is
148 * enabled and Privoxy was compiled without zlib support.
151 * Revision 1.95 2007/03/25 14:26:40 fabiankeil
152 * - Fix warnings when compiled with glibc.
153 * - Don't use crumble() for cookie crunching.
154 * - Move cookie time parsing into parse_header_time().
155 * - Let parse_header_time() return a jb_err code
156 * instead of a pointer that can only be used to
157 * check for NULL anyway.
159 * Revision 1.94 2007/03/21 12:23:53 fabiankeil
160 * - Add better protection against malicious gzip headers.
161 * - Stop logging the first hundred bytes of decompressed content.
162 * It looks like it's working and there is always debug 16.
163 * - Log the content size after decompression in decompress_iob()
164 * instead of pcrs_filter_response().
166 * Revision 1.93 2007/03/20 15:21:44 fabiankeil
167 * - Use dedicated header filter actions instead of abusing "filter".
168 * Replace "filter-client-headers" and "filter-client-headers"
169 * with "server-header-filter" and "client-header-filter".
170 * - Remove filter_client_header() and filter_client_header(),
171 * filter_header() now checks the shiny new
172 * CSP_FLAG_CLIENT_HEADER_PARSING_DONE flag instead.
174 * Revision 1.92 2007/03/05 13:25:32 fabiankeil
175 * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
176 * - Handle "Cookie:" and "Connection:" headers a bit smarter
177 * (don't crunch them just to recreate them later on).
178 * - Add another non-standard time format for the cookie
179 * expiration date detection.
180 * - Fix a valgrind warning.
182 * Revision 1.91 2007/02/24 12:27:32 fabiankeil
183 * Improve cookie expiration date detection.
185 * Revision 1.90 2007/02/08 19:12:35 fabiankeil
186 * Don't run server_content_length() the first time
187 * sed() parses server headers; only adjust the
188 * Content-Length header if the page was modified.
190 * Revision 1.89 2007/02/07 16:52:11 fabiankeil
191 * Fix log messages regarding the cookie time format
192 * (cookie and request URL were mixed up).
194 * Revision 1.88 2007/02/07 11:27:12 fabiankeil
195 * - Let decompress_iob()
196 * - not corrupt the content if decompression fails
197 * early. (the first byte(s) were lost).
198 * - use pointer arithmetics with defined outcome for
200 * - Use a different kludge to remember a failed decompression.
202 * Revision 1.87 2007/01/31 16:21:38 fabiankeil
203 * Search for Max-Forwards headers case-insensitive,
204 * don't generate the "501 unsupported" message for invalid
205 * Max-Forwards values and don't increase negative ones.
207 * Revision 1.86 2007/01/30 13:05:26 fabiankeil
208 * - Let server_set_cookie() check the expiration date
209 * of cookies and don't touch the ones that are already
210 * expired. Fixes problems with low quality web applications
211 * as described in BR 932612.
213 * - Adjust comment in client_max_forwards to reality;
214 * remove invalid Max-Forwards headers.
216 * Revision 1.85 2007/01/26 15:33:46 fabiankeil
217 * Stop filter_header() from unintentionally removing
218 * empty header lines that were enlisted by the continue
221 * Revision 1.84 2007/01/24 12:56:52 fabiankeil
222 * - Repeat the request URL before logging any headers.
223 * Makes reading the log easier in case of simultaneous requests.
224 * - If there are more than one Content-Type headers in one request,
225 * use the first one and remove the others.
226 * - Remove "newval" variable in server_content_type().
227 * It's only used once.
229 * Revision 1.83 2007/01/12 15:03:02 fabiankeil
230 * Correct a cast, check inflateEnd() exit code
231 * to see if we have to, replace sprintf calls
234 * Revision 1.82 2007/01/01 19:36:37 fabiankeil
235 * Integrate a modified version of Wil Mahan's
236 * zlib patch (PR #895531).
238 * Revision 1.81 2006/12/31 22:21:33 fabiankeil
239 * Skip empty filter files in filter_header()
240 * but don't ignore the ones that come afterwards.
241 * Fixes BR 1619208, this time for real.
243 * Revision 1.80 2006/12/29 19:08:22 fabiankeil
244 * Reverted parts of my last commit
245 * to keep error handling working.
247 * Revision 1.79 2006/12/29 18:04:40 fabiankeil
248 * Fixed gcc43 conversion warnings.
250 * Revision 1.78 2006/12/26 17:19:20 fabiankeil
251 * Bringing back the "useless" localtime() call
252 * I removed in revision 1.67. On some platforms
253 * it's necessary to prevent time zone offsets.
255 * Revision 1.77 2006/12/07 18:44:26 fabiankeil
256 * Rebuild request URL in get_destination_from_headers()
257 * to make sure redirect{pcrs command} works as expected
258 * for intercepted requests.
260 * Revision 1.76 2006/12/06 19:52:25 fabiankeil
261 * Added get_destination_from_headers().
263 * Revision 1.75 2006/11/13 19:05:51 fabiankeil
264 * Make pthread mutex locking more generic. Instead of
265 * checking for OSX and OpenBSD, check for FEATURE_PTHREAD
266 * and use mutex locking unless there is an _r function
267 * available. Better safe than sorry.
269 * Fixes "./configure --disable-pthread" and should result
270 * in less threading-related problems on pthread-using platforms,
271 * but it still doesn't fix BR#1122404.
273 * Revision 1.74 2006/10/02 16:59:12 fabiankeil
274 * The special header "X-Filter: No" now disables
275 * header filtering as well.
277 * Revision 1.73 2006/09/23 13:26:38 roro
278 * Replace TABs by spaces in source code.
280 * Revision 1.72 2006/09/23 12:37:21 fabiankeil
281 * Don't print a log message every time filter_headers is
282 * entered or left. It only creates noise without any real
285 * Revision 1.71 2006/09/21 19:55:17 fabiankeil
286 * Fix +hide-if-modified-since{-n}.
288 * Revision 1.70 2006/09/08 12:06:34 fabiankeil
289 * Have hide-if-modified-since interpret the random
290 * range value as minutes instead of hours. Allows
291 * more fine-grained configuration.
293 * Revision 1.69 2006/09/06 16:25:51 fabiankeil
294 * Always have parse_header_time return a pointer
295 * that actual makes sense, even though we currently
296 * only need it to detect problems.
298 * Revision 1.68 2006/09/06 10:43:32 fabiankeil
299 * Added config option enable-remote-http-toggle
300 * to specify if Privoxy should recognize special
301 * headers (currently only X-Filter) to change its
302 * behaviour. Disabled by default.
304 * Revision 1.67 2006/09/04 11:01:26 fabiankeil
305 * After filtering de-chunked instances, remove
306 * "Transfer-Encoding" header entirely instead of changing
307 * it to "Transfer-Encoding: identity", which is invalid.
308 * Thanks Michael Shields <shields@msrl.com>. Fixes PR 1318658.
310 * Don't use localtime in parse_header_time. An empty time struct
311 * is good enough, it gets overwritten by strptime anyway.
313 * Revision 1.66 2006/09/03 19:38:28 fabiankeil
314 * Use gmtime_r if available, fallback to gmtime with mutex
315 * protection for MacOSX and use vanilla gmtime for the rest.
317 * Revision 1.65 2006/08/22 10:55:56 fabiankeil
318 * Changed client_referrer to use the right type (size_t) for
319 * hostlenght and to shorten the temporary referrer string with
320 * '\0' instead of adding a useless line break.
322 * Revision 1.64 2006/08/17 17:15:10 fabiankeil
323 * - Back to timegm() using GnuPG's replacement if necessary.
324 * Using mktime() and localtime() could add a on hour offset if
325 * the randomize factor was big enough to lead to a summer/wintertime
328 * - Removed now-useless Privoxy 3.0.3 compatibility glue.
330 * - Moved randomization code into pick_from_range().
332 * - Changed parse_header_time definition.
333 * time_t isn't guaranteed to be signed and
334 * if it isn't, -1 isn't available as error code.
335 * Changed some variable types in client_if_modified_since()
336 * because of the same reason.
338 * Revision 1.63 2006/08/14 13:18:08 david__schmidt
339 * OS/2 compilation compatibility fixups
341 * Revision 1.62 2006/08/14 08:58:42 fabiankeil
342 * Changed include from strptime.c to strptime.h
344 * Revision 1.61 2006/08/14 08:25:19 fabiankeil
345 * Split filter-headers{} into filter-client-headers{}
346 * and filter-server-headers{}.
347 * Added parse_header_time() to share some code.
348 * Replaced timegm() with mktime().
350 * Revision 1.60 2006/08/12 03:54:37 david__schmidt
351 * Windows service integration
353 * Revision 1.59 2006/08/03 02:46:41 david__schmidt
354 * Incorporate Fabian Keil's patch work:
\rhttp://www.fabiankeil.de/sourcecode/privoxy/
356 * Revision 1.58 2006/07/18 14:48:47 david__schmidt
357 * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
358 * with what was really the latest development (the v_3_0_branch branch)
360 * Revision 1.56.2.10 2006/01/21 16:16:08 david__schmidt
361 * Thanks to Edward Carrel for his patch to modernize OSX's
\rpthreads support. See bug #1409623.
363 * Revision 1.56.2.9 2004/10/03 12:53:45 david__schmidt
364 * Add the ability to check jpeg images for invalid
365 * lengths of comment blocks. Defensive strategy
366 * against the exploit:
367 * Microsoft Security Bulletin MS04-028
368 * Buffer Overrun in JPEG Processing (GDI+) Could
369 * Allow Code Execution (833987)
370 * Enabled with +inspect-jpegs in actions files.
372 * Revision 1.56.2.8 2003/07/11 13:21:25 oes
373 * Excluded text/plain objects from filtering. This fixes a
374 * couple of client-crashing, download corruption and
375 * Privoxy performance issues, whose root cause lies in
376 * web servers labelling content of unknown type as text/plain.
378 * Revision 1.56.2.7 2003/05/06 12:07:26 oes
379 * Fixed bug #729900: Suspicious HOST: headers are now killed and regenerated if necessary
381 * Revision 1.56.2.6 2003/04/14 21:28:30 oes
382 * Completing the previous change
384 * Revision 1.56.2.5 2003/04/14 12:08:16 oes
385 * Added temporary workaround for bug in PHP < 4.2.3
387 * Revision 1.56.2.4 2003/03/07 03:41:05 david__schmidt
388 * 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.
390 * Revision 1.56.2.3 2002/11/10 04:20:02 hal9
391 * Fix typo: supressed -> suppressed
393 * Revision 1.56.2.2 2002/09/25 14:59:53 oes
394 * Improved cookie logging
396 * Revision 1.56.2.1 2002/09/25 14:52:45 oes
397 * Added basic support for OPTIONS and TRACE HTTP methods:
398 * - New parser function client_max_forwards which decrements
399 * the Max-Forwards HTTP header field of OPTIONS and TRACE
400 * requests by one before forwarding
401 * - New parser function client_host which extracts the host
402 * and port information from the HTTP header field if the
403 * request URI was not absolute
404 * - Don't crumble and re-add the Host: header, but only generate
405 * and append if missing
407 * Revision 1.56 2002/05/12 15:34:22 jongfoster
408 * Fixing typo in a comment
410 * Revision 1.55 2002/05/08 16:01:07 oes
411 * Optimized add_to_iob:
412 * - Use realloc instead of malloc(), memcpy(), free()
413 * - Expand to powers of two if possible, to get
414 * O(log n) reallocs instead of O(n).
415 * - Moved check for buffer limit here from chat
416 * - Report failure via returncode
418 * Revision 1.54 2002/04/02 15:03:16 oes
419 * Tiny code cosmetics
421 * Revision 1.53 2002/03/26 22:29:55 swa
422 * we have a new homepage!
424 * Revision 1.52 2002/03/24 13:25:43 swa
425 * name change related issues
427 * Revision 1.51 2002/03/13 00:27:05 jongfoster
430 * Revision 1.50 2002/03/12 01:45:35 oes
431 * More verbose logging
433 * Revision 1.49 2002/03/09 20:03:52 jongfoster
434 * - Making various functions return int rather than size_t.
435 * (Undoing a recent change). Since size_t is unsigned on
436 * Windows, functions like read_socket that return -1 on
437 * error cannot return a size_t.
439 * THIS WAS A MAJOR BUG - it caused frequent, unpredictable
440 * crashes, and also frequently caused JB to jump to 100%
441 * CPU and stay there. (Because it thought it had just
442 * read ((unsigned)-1) == 4Gb of data...)
444 * - The signature of write_socket has changed, it now simply
445 * returns success=0/failure=nonzero.
447 * - Trying to get rid of a few warnings --with-debug on
448 * Windows, I've introduced a new type "jb_socket". This is
449 * used for the socket file descriptors. On Windows, this
450 * is SOCKET (a typedef for unsigned). Everywhere else, it's
451 * an int. The error value can't be -1 any more, so it's
452 * now JB_INVALID_SOCKET (which is -1 on UNIX, and in
453 * Windows it maps to the #define INVALID_SOCKET.)
455 * - The signature of bind_port has changed.
457 * Revision 1.48 2002/03/07 03:46:53 oes
458 * Fixed compiler warnings etc
460 * Revision 1.47 2002/02/20 23:15:13 jongfoster
461 * Parsing functions now handle out-of-memory gracefully by returning
464 * Revision 1.46 2002/01/17 21:03:47 jongfoster
465 * Moving all our URL and URL pattern parsing code to urlmatch.c.
467 * Revision 1.45 2002/01/09 14:33:03 oes
468 * Added support for localtime_r.
470 * Revision 1.44 2001/12/14 01:22:54 steudten
471 * Remove 'user:pass@' from 'proto://user:pass@host' for the
472 * new added header 'Host: ..'. (See Req ID 491818)
474 * Revision 1.43 2001/11/23 00:26:38 jongfoster
475 * Fixing two really stupid errors in my previous commit
477 * Revision 1.42 2001/11/22 21:59:30 jongfoster
478 * Adding code to handle +no-cookies-keep
480 * Revision 1.41 2001/11/05 23:43:05 steudten
481 * Add time+date to log files.
483 * Revision 1.40 2001/10/26 20:13:09 jongfoster
484 * ctype.h is needed in Windows, too.
486 * Revision 1.39 2001/10/26 17:40:04 oes
487 * Introduced get_header_value()
488 * Removed http->user_agent, csp->referrer and csp->accept_types
489 * Removed client_accept()
491 * Revision 1.38 2001/10/25 03:40:48 david__schmidt
492 * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
493 * threads to call select() simultaneously. So, it's time to do a real, live,
494 * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
495 * (native). Both versions will work, but using __OS2__ offers multi-threading.
497 * Revision 1.37 2001/10/23 21:36:02 jongfoster
498 * Documenting sed()'s error behaviou (doc change only)
500 * Revision 1.36 2001/10/13 12:51:51 joergs
501 * Removed client_host, (was only required for the old 2.0.2-11 http://noijb.
502 * force-load), instead crumble Host: and add it (again) in client_host_adder
503 * (in case we get a HTTP/1.0 request without Host: header and forward it to
504 * a HTTP/1.1 server/proxy).
506 * Revision 1.35 2001/10/09 22:39:21 jongfoster
507 * assert.h is also required under Win32, so moving out of #ifndef _WIN32
510 * Revision 1.34 2001/10/07 18:50:55 oes
511 * Added server_content_encoding, renamed server_transfer_encoding
513 * Revision 1.33 2001/10/07 18:04:49 oes
514 * Changed server_http11 to server_http and its pattern to "HTTP".
515 * Additional functionality: it now saves the HTTP status into
516 * csp->http->status and sets CT_TABOO for Status 206 (partial range)
518 * Revision 1.32 2001/10/07 15:43:28 oes
519 * Removed FEATURE_DENY_GZIP and replaced it with client_accept_encoding,
520 * client_te and client_accept_encoding_adder, triggered by the new
521 * +no-compression action. For HTTP/1.1 the Accept-Encoding header is
522 * changed to allow only identity and chunked, and the TE header is
523 * crunched. For HTTP/1.0, Accept-Encoding is crunched.
525 * parse_http_request no longer does anything than parsing. The rewriting
526 * of http->cmd and version mangling are gone. It now also recognizes
527 * the put and delete methods and saves the url in http->url. Removed
530 * renamed content_type and content_length to have the server_ prefix
532 * server_content_type now only works if csp->content_type != CT_TABOO
534 * added server_transfer_encoding, which
535 * - Sets CT_TABOO to prohibit filtering if encoding compresses
536 * - Raises the CSP_FLAG_CHUNKED flag if Encoding is "chunked"
537 * - Change from "chunked" to "identity" if body was chunked
538 * but has been de-chunked for filtering.
540 * added server_content_md5 which crunches any Content-MD5 headers
541 * if the body was modified.
543 * made server_http11 conditional on +downgrade action
545 * Replaced 6 boolean members of csp with one bitmap (csp->flags)
547 * Revision 1.31 2001/10/05 14:25:02 oes
548 * Crumble Keep-Alive from Server
550 * Revision 1.30 2001/09/29 12:56:03 joergs
551 * IJB now changes HTTP/1.1 to HTTP/1.0 in requests and answers.
553 * Revision 1.29 2001/09/24 21:09:24 jongfoster
554 * Fixing 2 memory leaks that Guy spotted, where the paramater to
555 * enlist() was not being free()d.
557 * Revision 1.28 2001/09/22 16:32:28 jongfoster
558 * Removing unused #includes.
560 * Revision 1.27 2001/09/20 15:45:25 steudten
562 * add casting from size_t to int for printf()
563 * remove local variable shadow s2
565 * Revision 1.26 2001/09/16 17:05:14 jongfoster
566 * Removing unused #include showarg.h
568 * Revision 1.25 2001/09/16 13:21:27 jongfoster
569 * Changes to use new list functions.
571 * Revision 1.24 2001/09/13 23:05:50 jongfoster
572 * Changing the string paramater to the header parsers a "const".
574 * Revision 1.23 2001/09/12 18:08:19 steudten
576 * In parse_http_request() header rewriting miss the host value, so
577 * from http://www.mydomain.com the result was just " / " not
578 * http://www.mydomain.com/ in case we forward.
580 * Revision 1.22 2001/09/10 10:58:53 oes
581 * Silenced compiler warnings
583 * Revision 1.21 2001/07/31 14:46:00 oes
584 * - Persistant connections now suppressed
585 * - sed() no longer appends empty header to csp->headers
587 * Revision 1.20 2001/07/30 22:08:36 jongfoster
588 * Tidying up #defines:
589 * - All feature #defines are now of the form FEATURE_xxx
590 * - Permanently turned off WIN_GUI_EDIT
591 * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
593 * Revision 1.19 2001/07/25 17:21:54 oes
594 * client_uagent now saves copy of User-Agent: header value
596 * Revision 1.18 2001/07/13 14:02:46 oes
597 * - Included fix to repair broken HTTP requests that
598 * don't contain a path, not even '/'.
599 * - Removed all #ifdef PCRS
600 * - content_type now always inspected and classified as
601 * text, gif or other.
602 * - formatting / comments
604 * Revision 1.17 2001/06/29 21:45:41 oes
605 * Indentation, CRLF->LF, Tab-> Space
607 * Revision 1.16 2001/06/29 13:32:42 oes
609 * - Adapted free_http_request
610 * - Removed logentry from cancelled commit
612 * Revision 1.15 2001/06/03 19:12:38 oes
613 * deleted const struct interceptors
615 * Revision 1.14 2001/06/01 18:49:17 jongfoster
616 * Replaced "list_share" with "list" - the tiny memory gain was not
617 * worth the extra complexity.
619 * Revision 1.13 2001/05/31 21:30:33 jongfoster
620 * Removed list code - it's now in list.[ch]
621 * Renamed "permission" to "action", and changed many features
622 * to use the actions file rather than the global config.
624 * Revision 1.12 2001/05/31 17:33:13 oes
628 * Revision 1.11 2001/05/29 20:11:19 joergs
629 * '/ * inside comment' warning removed.
631 * Revision 1.10 2001/05/29 09:50:24 jongfoster
632 * Unified blocklist/imagelist/permissionslist.
633 * File format is still under discussion, but the internal changes
636 * Also modified interceptor behaviour:
637 * - We now intercept all URLs beginning with one of the following
638 * prefixes (and *only* these prefixes):
640 * * http://ijbswa.sf.net/config/
641 * * http://ijbswa.sourceforge.net/config/
642 * - New interceptors "home page" - go to http://i.j.b/ to see it.
643 * - Internal changes so that intercepted and fast redirect pages
644 * are not replaced with an image.
645 * - Interceptors now have the option to send a binary page direct
646 * to the client. (i.e. ijb-send-banner uses this)
647 * - Implemented show-url-info interceptor. (Which is why I needed
648 * the above interceptors changes - a typical URL is
649 * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
650 * The previous mechanism would not have intercepted that, and
651 * if it had been intercepted then it then it would have replaced
654 * Revision 1.9 2001/05/28 17:26:33 jongfoster
655 * Fixing segfault if last header was crunched.
656 * Fixing Windows build (snprintf() is _snprintf() under Win32, but we
657 * can use the cross-platform sprintf() instead.)
659 * Revision 1.8 2001/05/27 22:17:04 oes
661 * - re_process_buffer no longer writes the modified buffer
662 * to the client, which was very ugly. It now returns the
663 * buffer, which it is then written by chat.
665 * - content_length now adjusts the Content-Length: header
666 * for modified documents rather than crunch()ing it.
667 * (Length info in csp->content_length, which is 0 for
668 * unmodified documents)
670 * - For this to work, sed() is called twice when filtering.
672 * Revision 1.7 2001/05/27 13:19:06 oes
673 * Patched Joergs solution for the content-length in.
675 * Revision 1.6 2001/05/26 13:39:32 jongfoster
676 * Only crunches Content-Length header if applying RE filtering.
677 * Without this fix, Microsoft Windows Update wouldn't work.
679 * Revision 1.5 2001/05/26 00:28:36 jongfoster
680 * Automatic reloading of config file.
681 * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
682 * Most of the global variables have been moved to a new
683 * struct configuration_spec, accessed through csp->config->globalname
684 * Most of the globals remaining are used by the Win32 GUI.
686 * Revision 1.4 2001/05/22 18:46:04 oes
688 * - Enabled filtering banners by size rather than URL
689 * by adding patterns that replace all standard banner
690 * sizes with the "Junkbuster" gif to the re_filterfile
692 * - Enabled filtering WebBugs by providing a pattern
693 * which kills all 1x1 images
695 * - Added support for PCRE_UNGREEDY behaviour to pcrs,
696 * which is selected by the (nonstandard and therefore
697 * capital) letter 'U' in the option string.
698 * It causes the quantifiers to be ungreedy by default.
699 * Appending a ? turns back to greedy (!).
701 * - Added a new interceptor ijb-send-banner, which
702 * sends back the "Junkbuster" gif. Without imagelist or
703 * MSIE detection support, or if tinygif = 1, or the
704 * URL isn't recognized as an imageurl, a lame HTML
705 * explanation is sent instead.
707 * - Added new feature, which permits blocking remote
708 * script redirects and firing back a local redirect
710 * The feature is conditionally compiled, i.e. it
711 * can be disabled with --disable-fast-redirects,
712 * plus it must be activated by a "fast-redirects"
713 * line in the config file, has its own log level
714 * and of course wants to be displayed by show-proxy-args
715 * Note: Boy, all the #ifdefs in 1001 locations and
716 * all the fumbling with configure.in and acconfig.h
717 * were *way* more work than the feature itself :-(
719 * - Because a generic redirect template was needed for
720 * this, tinygif = 3 now uses the same.
722 * - Moved GIFs, and other static HTTP response templates
727 * - Removed some >400 CRs again (Jon, you really worked
730 * Revision 1.3 2001/05/20 01:21:20 jongfoster
731 * Version 2.9.4 checkin.
732 * - Merged popupfile and cookiefile, and added control over PCRS
733 * filtering, in new "permissionsfile".
734 * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
735 * file error you now get a message box (in the Win32 GUI) rather
736 * than the program exiting with no explanation.
737 * - Made killpopup use the PCRS MIME-type checking and HTTP-header
739 * - Removed tabs from "config"
740 * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
741 * - Bumped up version number.
743 * Revision 1.2 2001/05/17 23:02:36 oes
744 * - Made referrer option accept 'L' as a substitute for '§'
746 * Revision 1.1.1.1 2001/05/15 13:59:01 oes
747 * Initial import of version 2.9.3 source tree
750 *********************************************************************/
757 #include <sys/types.h>
767 * Convince GNU's libc to provide a strptime prototype.
770 #endif /*__GLIBC__ */
777 #if !defined(_WIN32) && !defined(__OS2__)
783 #ifdef FEATURE_PTHREAD
785 /* jcc.h is for mutex semapores only */
786 #endif /* def FEATURE_PTHREAD */
792 #include "jbsockets.h"
793 #include "miscutil.h"
798 #ifndef HAVE_STRPTIME
799 #include "strptime.h"
802 const char parsers_h_rcs[] = PARSERS_H_VERSION;
804 /* Fix a problem with Solaris. There should be no effect on other
806 * Solaris's isspace() is a macro which uses its argument directly
807 * as an array index. Therefore we need to make sure that high-bit
808 * characters generate +ve values, and ideally we also want to make
809 * the argument match the declared parameter type of "int".
811 * Why did they write a character function that can't take a simple
812 * "char" argument? Doh!
814 #define ijb_isupper(__X) isupper((int)(unsigned char)(__X))
815 #define ijb_tolower(__X) tolower((int)(unsigned char)(__X))
817 static jb_err scan_headers(struct client_state *csp);
818 static jb_err header_tagger(struct client_state *csp, char *header);
819 static jb_err parse_header_time(const char *header_time, time_t *result);
821 static jb_err crumble (struct client_state *csp, char **header);
822 static jb_err connection (struct client_state *csp, char **header);
823 static jb_err filter_header (struct client_state *csp, char **header);
824 static jb_err client_referrer (struct client_state *csp, char **header);
825 static jb_err client_uagent (struct client_state *csp, char **header);
826 static jb_err client_ua (struct client_state *csp, char **header);
827 static jb_err client_from (struct client_state *csp, char **header);
828 static jb_err client_send_cookie (struct client_state *csp, char **header);
829 static jb_err client_x_forwarded (struct client_state *csp, char **header);
830 static jb_err client_accept_encoding (struct client_state *csp, char **header);
831 static jb_err client_te (struct client_state *csp, char **header);
832 static jb_err client_max_forwards (struct client_state *csp, char **header);
833 static jb_err client_host (struct client_state *csp, char **header);
834 static jb_err client_if_modified_since (struct client_state *csp, char **header);
835 static jb_err client_accept_language (struct client_state *csp, char **header);
836 static jb_err client_if_none_match (struct client_state *csp, char **header);
837 static jb_err crunch_client_header (struct client_state *csp, char **header);
838 static jb_err client_x_filter (struct client_state *csp, char **header);
839 static jb_err client_range (struct client_state *csp, char **header);
840 static jb_err server_set_cookie (struct client_state *csp, char **header);
841 static jb_err server_content_type (struct client_state *csp, char **header);
842 static jb_err server_content_length (struct client_state *csp, char **header);
843 static jb_err server_content_md5 (struct client_state *csp, char **header);
844 static jb_err server_content_encoding (struct client_state *csp, char **header);
845 static jb_err server_transfer_coding (struct client_state *csp, char **header);
846 static jb_err server_http (struct client_state *csp, char **header);
847 static jb_err crunch_server_header (struct client_state *csp, char **header);
848 static jb_err server_last_modified (struct client_state *csp, char **header);
849 static jb_err server_content_disposition(struct client_state *csp, char **header);
851 static jb_err client_host_adder (struct client_state *csp);
852 static jb_err client_cookie_adder (struct client_state *csp);
853 static jb_err client_xtra_adder (struct client_state *csp);
854 static jb_err connection_close_adder (struct client_state *csp);
856 static jb_err create_forged_referrer(char **header, const char *hostport);
857 static jb_err create_fake_referrer(char **header, const char *fake_referrer);
858 static jb_err handle_conditional_hide_referrer_parameter(char **header,
859 const char *host, const int parameter_conditional_block);
861 const struct parsers client_patterns[] = {
862 { "referer:", 8, client_referrer },
863 { "user-agent:", 11, client_uagent },
864 { "ua-", 3, client_ua },
865 { "from:", 5, client_from },
866 { "cookie:", 7, client_send_cookie },
867 { "x-forwarded-for:", 16, client_x_forwarded },
868 { "Accept-Encoding:", 16, client_accept_encoding },
869 { "TE:", 3, client_te },
870 { "Host:", 5, client_host },
871 { "if-modified-since:", 18, client_if_modified_since },
872 { "Keep-Alive:", 11, crumble },
873 { "connection:", 11, connection },
874 { "proxy-connection:", 17, crumble },
875 { "max-forwards:", 13, client_max_forwards },
876 { "Accept-Language:", 16, client_accept_language },
877 { "if-none-match:", 14, client_if_none_match },
878 { "Range:", 6, client_range },
879 { "Request-Range:", 14, client_range },
880 { "If-Range:", 9, client_range },
881 { "X-Filter:", 9, client_x_filter },
882 { "*", 0, crunch_client_header },
883 { "*", 0, filter_header },
887 const struct parsers server_patterns[] = {
888 { "HTTP/", 5, server_http },
889 { "set-cookie:", 11, server_set_cookie },
890 { "connection:", 11, connection },
891 { "Content-Type:", 13, server_content_type },
892 { "Content-MD5:", 12, server_content_md5 },
893 { "Content-Encoding:", 17, server_content_encoding },
894 { "Transfer-Encoding:", 18, server_transfer_coding },
895 { "Keep-Alive:", 11, crumble },
896 { "content-disposition:", 20, server_content_disposition },
897 { "Last-Modified:", 14, server_last_modified },
898 { "*", 0, crunch_server_header },
899 { "*", 0, filter_header },
903 const struct parsers server_patterns_light[] = {
904 { "Content-Length:", 15, server_content_length },
905 { "Transfer-Encoding:", 18, server_transfer_coding },
907 { "Content-Encoding:", 17, server_content_encoding },
908 #endif /* def FEATURE_ZLIB */
912 const add_header_func_ptr add_client_headers[] = {
916 /* Temporarily disabled: client_accept_encoding_adder, */
917 connection_close_adder,
921 const add_header_func_ptr add_server_headers[] = {
922 connection_close_adder,
926 /* The vanilla wafer. */
927 static const char VANILLA_WAFER[] =
928 "NOTICE=TO_WHOM_IT_MAY_CONCERN_"
929 "Do_not_send_me_any_copyrighted_information_other_than_the_"
930 "document_that_I_am_requesting_or_any_of_its_necessary_components._"
931 "In_particular_do_not_send_me_any_cookies_that_"
932 "are_subject_to_a_claim_of_copyright_by_anybody._"
933 "Take_notice_that_I_refuse_to_be_bound_by_any_license_condition_"
934 "(copyright_or_otherwise)_applying_to_any_cookie._";
936 /*********************************************************************
938 * Function : flush_socket
940 * Description : Write any pending "buffered" content.
943 * 1 : fd = file descriptor of the socket to read
944 * 2 : csp = Current client state (buffers, headers, etc...)
946 * Returns : On success, the number of bytes written are returned (zero
947 * indicates nothing was written). On error, -1 is returned,
948 * and errno is set appropriately. If count is zero and the
949 * file descriptor refers to a regular file, 0 will be
950 * returned without causing any other effect. For a special
951 * file, the results are not portable.
953 *********************************************************************/
954 int flush_socket(jb_socket fd, struct client_state *csp)
956 struct iob *iob = csp->iob;
957 int len = iob->eod - iob->cur;
964 if (write_socket(fd, iob->cur, (size_t)len))
968 iob->eod = iob->cur = iob->buf;
974 /*********************************************************************
976 * Function : add_to_iob
978 * Description : Add content to the buffered page, expanding the
979 * buffer if necessary.
982 * 1 : csp = Current client state (buffers, headers, etc...)
983 * 2 : buf = holds the content to be added to the page
984 * 3 : n = number of bytes to be added
986 * Returns : JB_ERR_OK on success, JB_ERR_MEMORY if out-of-memory
987 * or buffer limit reached.
989 *********************************************************************/
990 jb_err add_to_iob(struct client_state *csp, char *buf, int n)
992 struct iob *iob = csp->iob;
993 size_t used, offset, need, want;
996 if (n <= 0) return JB_ERR_OK;
998 used = (size_t)(iob->eod - iob->buf);
999 offset = (size_t)(iob->cur - iob->buf);
1000 need = used + (size_t)n + 1;
1003 * If the buffer can't hold the new data, extend it first.
1004 * Use the next power of two if possible, else use the actual need.
1006 if (need > csp->config->buffer_limit)
1008 log_error(LOG_LEVEL_INFO, "Buffer limit reached while extending the buffer (iob)");
1009 return JB_ERR_MEMORY;
1012 if (need > iob->size)
1014 for (want = csp->iob->size ? csp->iob->size : 512; want <= need;) want *= 2;
1016 if (want <= csp->config->buffer_limit && NULL != (p = (char *)realloc(iob->buf, want)))
1020 else if (NULL != (p = (char *)realloc(iob->buf, need)))
1026 log_error(LOG_LEVEL_ERROR, "Extending the buffer (iob) failed: %E");
1027 return JB_ERR_MEMORY;
1030 /* Update the iob pointers */
1031 iob->cur = p + offset;
1032 iob->eod = p + used;
1036 /* copy the new data into the iob buffer */
1037 memcpy(iob->eod, buf, (size_t)n);
1039 /* point to the end of the data */
1042 /* null terminate == cheap insurance */
1051 /*********************************************************************
1053 * Function : decompress_iob
1055 * Description : Decompress buffered page, expanding the
1056 * buffer as necessary. csp->iob->cur
1057 * should point to the the beginning of the
1058 * compressed data block.
1061 * 1 : csp = Current client state (buffers, headers, etc...)
1063 * Returns : JB_ERR_OK on success,
1064 * JB_ERR_MEMORY if out-of-memory limit reached, and
1065 * JB_ERR_COMPRESS if error decompressing buffer.
1067 *********************************************************************/
1068 jb_err decompress_iob(struct client_state *csp)
1070 char *buf; /* new, uncompressed buffer */
1071 char *cur; /* Current iob position (to keep the original
1072 * iob->cur unmodified if we return early) */
1073 size_t bufsize; /* allocated size of the new buffer */
1074 size_t old_size; /* Content size before decompression */
1075 size_t skip_size; /* Number of bytes at the beginning of the iob
1076 that we should NOT decompress. */
1077 int status; /* return status of the inflate() call */
1078 z_stream zstr; /* used by calls to zlib */
1080 assert(csp->iob->cur - csp->iob->buf > 0);
1081 assert(csp->iob->eod - csp->iob->cur > 0);
1083 bufsize = csp->iob->size;
1084 skip_size = (size_t)(csp->iob->cur - csp->iob->buf);
1085 old_size = (size_t)(csp->iob->eod - csp->iob->cur);
1087 cur = csp->iob->cur;
1092 * This is to protect the parsing of gzipped data,
1093 * but it should(?) be valid for deflated data also.
1095 log_error(LOG_LEVEL_ERROR, "Buffer too small decompressing iob");
1096 return JB_ERR_COMPRESS;
1099 if (csp->content_type & CT_GZIP)
1102 * Our task is slightly complicated by the facts that data
1103 * compressed by gzip does not include a zlib header, and
1104 * that there is no easily accessible interface in zlib to
1105 * handle a gzip header. We strip off the gzip header by
1106 * hand, and later inform zlib not to expect a header.
1110 * Strip off the gzip header. Please see RFC 1952 for more
1111 * explanation of the appropriate fields.
1113 if ((*cur++ != (char)0x1f)
1114 || (*cur++ != (char)0x8b)
1115 || (*cur++ != Z_DEFLATED))
1117 log_error(LOG_LEVEL_ERROR, "Invalid gzip header when decompressing");
1118 return JB_ERR_COMPRESS;
1124 * XXX: These magic numbers should be replaced
1125 * with macros to give a better idea what they do.
1129 /* The gzip header has reserved bits set; bail out. */
1130 log_error(LOG_LEVEL_ERROR, "Invalid gzip header flags when decompressing");
1131 return JB_ERR_COMPRESS;
1135 /* Skip extra fields if necessary. */
1139 * Skip a given number of bytes, specified
1140 * as a 16-bit little-endian value.
1143 * XXX: This code used to be:
1145 * csp->iob->cur += *csp->iob->cur++ + (*csp->iob->cur++ << 8);
1147 * which I had to change into:
1149 * cur += *cur++ + (*cur++ << 8);
1151 * at which point gcc43 finally noticed that the value
1152 * of cur is undefined (it depends on which of the
1153 * summands is evaluated first).
1155 * I haven't come across a site where this
1156 * code is actually executed yet, but I hope
1160 skip_bytes = *cur++;
1161 skip_bytes = *cur++ << 8;
1163 assert(skip_bytes == *csp->iob->cur - 2 + ((*csp->iob->cur - 1) << 8));
1166 * The number of bytes to skip should be positive
1167 * and we'd like to stay in the buffer.
1169 if ((skip_bytes < 0) || (skip_bytes >= (csp->iob->eod - cur)))
1171 log_error(LOG_LEVEL_ERROR,
1172 "Unreasonable amount of bytes to skip (%d). Stopping decompression",
1174 return JB_ERR_COMPRESS;
1176 log_error(LOG_LEVEL_INFO,
1177 "Skipping %d bytes for gzip compression. Does this sound right?",
1182 /* Skip the filename if necessary. */
1185 /* A null-terminated string is supposed to follow. */
1186 while (*cur++ && (cur < csp->iob->eod));
1190 /* Skip the comment if necessary. */
1193 /* A null-terminated string is supposed to follow. */
1194 while (*cur++ && (cur < csp->iob->eod));
1197 /* Skip the CRC if necessary. */
1203 if (cur >= csp->iob->eod)
1206 * If the current position pointer reached or passed
1207 * the buffer end, we were obviously tricked to skip
1210 log_error(LOG_LEVEL_ERROR,
1211 "Malformed gzip header detected. Aborting decompression.");
1212 return JB_ERR_COMPRESS;
1216 else if (csp->content_type & CT_DEFLATE)
1219 * XXX: The debug level should be lowered
1220 * before the next stable release.
1222 log_error(LOG_LEVEL_INFO, "Decompressing deflated iob: %d", *cur);
1224 * In theory (that is, according to RFC 1950), deflate-compressed
1225 * data should begin with a two-byte zlib header and have an
1226 * adler32 checksum at the end. It seems that in practice only
1227 * the raw compressed data is sent. Note that this means that
1228 * we are not RFC 1950-compliant here, but the advantage is that
1229 * this actually works. :)
1231 * We add a dummy null byte to tell zlib where the data ends,
1232 * and later inform it not to expect a header.
1234 * Fortunately, add_to_iob() has thoughtfully null-terminated
1235 * the buffer; we can just increment the end pointer to include
1242 log_error(LOG_LEVEL_ERROR,
1243 "Unable to determine compression format for decompression");
1244 return JB_ERR_COMPRESS;
1247 /* Set up the fields required by zlib. */
1248 zstr.next_in = (Bytef *)cur;
1249 zstr.avail_in = (unsigned int)(csp->iob->eod - cur);
1250 zstr.zalloc = Z_NULL;
1251 zstr.zfree = Z_NULL;
1252 zstr.opaque = Z_NULL;
1255 * Passing -MAX_WBITS to inflateInit2 tells the library
1256 * that there is no zlib header.
1258 if (inflateInit2 (&zstr, -MAX_WBITS) != Z_OK)
1260 log_error(LOG_LEVEL_ERROR, "Error initializing decompression");
1261 return JB_ERR_COMPRESS;
1265 * Next, we allocate new storage for the inflated data.
1266 * We don't modify the existing iob yet, so in case there
1267 * is error in decompression we can recover gracefully.
1269 buf = zalloc(bufsize);
1272 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
1273 return JB_ERR_MEMORY;
1276 assert(bufsize >= skip_size);
1277 memcpy(buf, csp->iob->buf, skip_size);
1278 zstr.avail_out = bufsize - skip_size;
1279 zstr.next_out = (Bytef *)buf + skip_size;
1281 /* Try to decompress the whole stream in one shot. */
1282 while (Z_BUF_ERROR == (status = inflate(&zstr, Z_FINISH)))
1284 /* We need to allocate more memory for the output buffer. */
1286 char *tmpbuf; /* used for realloc'ing the buffer */
1287 size_t oldbufsize = bufsize; /* keep track of the old bufsize */
1290 * If zlib wants more data then there's a problem, because
1291 * the complete compressed file should have been buffered.
1293 if (0 == zstr.avail_in)
1295 log_error(LOG_LEVEL_ERROR, "Unexpected end of compressed iob");
1296 return JB_ERR_COMPRESS;
1300 * If we tried the limit and still didn't have enough
1301 * memory, just give up.
1303 if (bufsize == csp->config->buffer_limit)
1305 log_error(LOG_LEVEL_ERROR, "Buffer limit reached while decompressing iob");
1306 return JB_ERR_MEMORY;
1309 /* Try doubling the buffer size each time. */
1312 /* Don't exceed the buffer limit. */
1313 if (bufsize > csp->config->buffer_limit)
1315 bufsize = csp->config->buffer_limit;
1318 /* Try to allocate the new buffer. */
1319 tmpbuf = realloc(buf, bufsize);
1322 log_error(LOG_LEVEL_ERROR, "Out of memory decompressing iob");
1324 return JB_ERR_MEMORY;
1328 char *oldnext_out = (char *)zstr.next_out;
1331 * Update the fields for inflate() to use the new
1332 * buffer, which may be in a location different from
1335 zstr.avail_out += bufsize - oldbufsize;
1336 zstr.next_out = (Bytef *)tmpbuf + bufsize - zstr.avail_out;
1339 * Compare with an uglier method of calculating these values
1340 * that doesn't require the extra oldbufsize variable.
1342 assert(zstr.avail_out == tmpbuf + bufsize - (char *)zstr.next_out);
1343 assert((char *)zstr.next_out == tmpbuf + ((char *)oldnext_out - buf));
1344 assert(zstr.avail_out > 0);
1350 if (Z_STREAM_ERROR == inflateEnd(&zstr))
1352 log_error(LOG_LEVEL_ERROR,
1353 "Inconsistent stream state after decompression: %s", zstr.msg);
1355 * XXX: Intentionally no return.
1357 * According to zlib.h, Z_STREAM_ERROR is returned
1358 * "if the stream state was inconsistent".
1360 * I assume in this case inflate()'s status
1361 * would also be something different than Z_STREAM_END
1362 * so this check should be redundant, but lets see.
1366 if (status != Z_STREAM_END)
1368 /* We failed to decompress the stream. */
1369 log_error(LOG_LEVEL_ERROR,
1370 "Error in decompressing to the buffer (iob): %s", zstr.msg);
1371 return JB_ERR_COMPRESS;
1375 * Finally, we can actually update the iob, since the
1376 * decompression was successful. First, free the old
1379 freez(csp->iob->buf);
1381 /* Now, update the iob to use the new buffer. */
1382 csp->iob->buf = buf;
1383 csp->iob->cur = csp->iob->buf + skip_size;
1384 csp->iob->eod = (char *)zstr.next_out;
1385 csp->iob->size = bufsize;
1388 * Make sure the new uncompressed iob obeys some minimal
1389 * consistency conditions.
1391 if ((csp->iob->buf < csp->iob->cur)
1392 && (csp->iob->cur <= csp->iob->eod)
1393 && (csp->iob->eod <= csp->iob->buf + csp->iob->size))
1395 const size_t new_size = (size_t)(csp->iob->eod - csp->iob->cur);
1398 log_error(LOG_LEVEL_RE_FILTER,
1399 "Decompression successful. Old size: %d, new size: %d.",
1400 old_size, new_size);
1404 /* zlib thinks this is OK, so lets do the same. */
1405 log_error(LOG_LEVEL_INFO, "Decompression didn't result in any content.");
1410 /* It seems that zlib did something weird. */
1411 log_error(LOG_LEVEL_ERROR,
1412 "Unexpected error decompressing the buffer (iob): %d==%d, %d>%d, %d<%d",
1413 csp->iob->cur, csp->iob->buf + skip_size, csp->iob->eod, csp->iob->buf,
1414 csp->iob->eod, csp->iob->buf + csp->iob->size);
1415 return JB_ERR_COMPRESS;
1421 #endif /* defined(FEATURE_ZLIB) */
1424 /*********************************************************************
1426 * Function : get_header
1428 * Description : This (odd) routine will parse the csp->iob
1431 * 1 : csp = Current client state (buffers, headers, etc...)
1433 * Returns : Any one of the following:
1435 * 1) a pointer to a dynamically allocated string that contains a header line
1436 * 2) NULL indicating that the end of the header was reached
1437 * 3) "" indicating that the end of the iob was reached before finding
1438 * a complete header line.
1440 *********************************************************************/
1441 char *get_header(struct client_state *csp)
1447 if ((iob->cur == NULL)
1448 || ((p = strchr(iob->cur, '\n')) == NULL))
1450 return(""); /* couldn't find a complete header */
1455 ret = strdup(iob->cur);
1458 /* FIXME No way to handle error properly */
1459 log_error(LOG_LEVEL_FATAL, "Out of memory in get_header()");
1464 if ((q = strchr(ret, '\r')) != NULL) *q = '\0';
1466 /* is this a blank line (i.e. the end of the header) ? */
1478 /*********************************************************************
1480 * Function : get_header_value
1482 * Description : Get the value of a given header from a chained list
1483 * of header lines or return NULL if no such header is
1484 * present in the list.
1487 * 1 : header_list = pointer to list
1488 * 2 : header_name = string with name of header to look for.
1489 * Trailing colon required, capitalization
1492 * Returns : NULL if not found, else value of header
1494 *********************************************************************/
1495 char *get_header_value(const struct list *header_list, const char *header_name)
1497 struct list_entry *cur_entry;
1501 assert(header_list);
1502 assert(header_name);
1503 length = strlen(header_name);
1505 for (cur_entry = header_list->first; cur_entry ; cur_entry = cur_entry->next)
1509 if (!strncmpic(cur_entry->str, header_name, length))
1512 * Found: return pointer to start of value
1514 ret = (char *) (cur_entry->str + length);
1515 while (*ret && ijb_isspace(*ret)) ret++;
1529 /*********************************************************************
1531 * Function : scan_headers
1533 * Description : Scans headers, applies tags and updates action bits.
1536 * 1 : csp = Current client state (buffers, headers, etc...)
1538 * Returns : JB_ERR_OK
1540 *********************************************************************/
1541 static jb_err scan_headers(struct client_state *csp)
1543 struct list_entry *h; /* Header */
1544 jb_err err = JB_ERR_OK;
1546 for (h = csp->headers->first; (err == JB_ERR_OK) && (h != NULL) ; h = h->next)
1548 /* Header crunch()ed in previous run? -> ignore */
1549 if (h->str == NULL) continue;
1550 log_error(LOG_LEVEL_HEADER, "scan: %s", h->str);
1551 err = header_tagger(csp, h->str);
1558 /*********************************************************************
1562 * Description : add, delete or modify lines in the HTTP header streams.
1563 * On entry, it receives a linked list of headers space
1564 * that was allocated dynamically (both the list nodes
1565 * and the header contents).
1567 * As a side effect it frees the space used by the original
1570 * XXX: should be split to remove the first_run hack.
1573 * 1 : pats = list of patterns to match against headers
1574 * 2 : more_headers = list of functions to add more
1575 * headers (client or server)
1576 * 3 : csp = Current client state (buffers, headers, etc...)
1578 * Returns : JB_ERR_OK in case off success, or
1579 * JB_ERR_MEMORY on out-of-memory error.
1581 *********************************************************************/
1582 jb_err sed(const struct parsers pats[],
1583 const add_header_func_ptr more_headers[],
1584 struct client_state *csp)
1586 struct list_entry *p;
1587 const struct parsers *v;
1588 const add_header_func_ptr *f;
1589 jb_err err = JB_ERR_OK;
1593 * If filtering is enabled, sed is run twice,
1594 * but most of the work needs to be done only once.
1596 first_run = (more_headers != NULL ) ? 1 : 0;
1598 if (first_run) /* Parse and print */
1602 for (v = pats; (err == JB_ERR_OK) && (v->str != NULL) ; v++)
1604 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL) ; p = p->next)
1606 /* Header crunch()ed in previous run? -> ignore */
1607 if (p->str == NULL) continue;
1609 /* Does the current parser handle this header? */
1610 if ((strncmpic(p->str, v->str, v->len) == 0) || (v->len == CHECK_EVERY_HEADER_REMAINING))
1612 err = v->parser(csp, (char **)&(p->str));
1616 /* place any additional headers on the csp->headers list */
1617 for (f = more_headers; (err == JB_ERR_OK) && (*f) ; f++)
1622 else /* Parse only */
1625 * The second run is only needed if the body was modified
1626 * and the content-lenght has changed.
1628 if (strncmpic(csp->http->cmd, "HEAD", 4))
1630 /*XXX: Code duplication */
1631 for (v = pats; (err == JB_ERR_OK) && (v->str != NULL) ; v++)
1633 for (p = csp->headers->first; (err == JB_ERR_OK) && (p != NULL) ; p = p->next)
1635 /* Header crunch()ed in previous run? -> ignore */
1636 if (p->str == NULL) continue;
1638 /* Does the current parser handle this header? */
1639 if (strncmpic(p->str, v->str, v->len) == 0)
1641 err = v->parser(csp, (char **)&(p->str));
1653 /*********************************************************************
1655 * Function : header_tagger
1657 * Description : Executes all text substitutions from applying
1658 * tag actions and saves the result as tag.
1660 * XXX: Shares enough code with filter_header() and
1661 * pcrs_filter_response() to warrant some helper functions.
1664 * 1 : csp = Current client state (buffers, headers, etc...)
1665 * 2 : header = Header that is used as tagger input
1667 * Returns : JB_ERR_OK on success and always succeeds
1669 *********************************************************************/
1670 static jb_err header_tagger(struct client_state *csp, char *header)
1672 int wanted_filter_type;
1673 int multi_action_index;
1677 struct file_list *fl;
1678 struct re_filterfile_spec *b;
1679 struct list_entry *tag_name;
1681 int found_filters = 0;
1682 const size_t header_length = strlen(header);
1684 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1686 wanted_filter_type = FT_SERVER_HEADER_TAGGER;
1687 multi_action_index = ACTION_MULTI_SERVER_HEADER_TAGGER;
1691 wanted_filter_type = FT_CLIENT_HEADER_TAGGER;
1692 multi_action_index = ACTION_MULTI_CLIENT_HEADER_TAGGER;
1695 /* Check if there are any filters */
1696 for (i = 0; i < MAX_AF_FILES; i++)
1709 if (0 == found_filters)
1711 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1712 "tagging enabled, but no taggers available.");
1716 for (i = 0; i < MAX_AF_FILES; i++)
1719 if ((NULL == fl) || (NULL == fl->f))
1722 * Either there are no filter files
1723 * left, or this filter file just
1724 * contains no valid filters.
1726 * Continue to be sure we don't miss
1727 * valid filter files that are chained
1728 * after empty or invalid ones.
1733 /* For all filters, */
1734 for (b = fl->f; b; b = b->next)
1736 if (b->type != wanted_filter_type)
1738 /* skip the ones we don't care about, */
1741 /* leaving only taggers that could apply, of which we use the ones, */
1742 for (tag_name = csp->action->multi[multi_action_index]->first;
1743 NULL != tag_name; tag_name = tag_name->next)
1745 /* that do apply, and */
1746 if (strcmp(b->name, tag_name->str) == 0)
1748 char *modified_tag = NULL;
1750 size_t size = header_length;
1751 pcrs_job *joblist = b->joblist;
1753 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1755 if (NULL == joblist)
1757 log_error(LOG_LEVEL_RE_FILTER,
1758 "Tagger %s has empty joblist. Nothing to do.", b->name);
1762 /* execute their pcrs_joblist on the header. */
1763 for (job = joblist; NULL != job; job = job->next)
1765 const int hits = pcrs_execute(job, tag, size, &modified_tag, &size);
1769 /* Success, continue with the modified version. */
1778 /* Tagger doesn't match */
1781 /* Regex failure, log it but continue anyway. */
1782 log_error(LOG_LEVEL_ERROR,
1783 "Problems with tagger \'%s\' and header \'%s\': %s",
1784 b->name, *header, pcrs_strerror(hits));
1786 freez(modified_tag);
1790 if (b->dynamic) pcrs_free_joblist(joblist);
1792 /* If this tagger matched */
1798 * There is to technical limitation which makes
1799 * it impossible to use empty tags, but I assume
1800 * no one would do it intentionally.
1803 log_error(LOG_LEVEL_INFO,
1804 "Tagger \'%s\' created an empty tag. Ignored.",
1809 if (!list_contains_item(csp->tags, tag))
1811 if (JB_ERR_OK != enlist(csp->tags, tag))
1813 log_error(LOG_LEVEL_ERROR,
1814 "Insufficient memory to add tag \'%s\', "
1815 "based on tagger \'%s\' and header \'%s\'",
1816 tag, b->name, *header);
1820 char *action_message;
1822 * update the action bits right away, to make
1823 * tagging based on tags set by earlier taggers
1824 * of the same kind possible.
1826 if (update_action_bits_for_tag(csp, tag))
1828 action_message = "Action bits updated accordingly.";
1832 action_message = "No action bits update necessary.";
1835 log_error(LOG_LEVEL_HEADER,
1836 "Tagger \'%s\' added tag \'%s\'. %s",
1837 b->name, tag, action_message);
1842 /* XXX: Is this log-worthy? */
1843 log_error(LOG_LEVEL_HEADER,
1844 "Tagger \'%s\' didn't add tag \'%s\'. "
1845 "Tag already present", b->name, tag);
1848 } /* if the tagger matched */
1849 } /* if the tagger applies */
1850 } /* for every tagger that could apply */
1851 } /* for all filters */
1852 } /* for all filter files */
1857 /* here begins the family of parser functions that reformat header lines */
1859 /*********************************************************************
1861 * Function : filter_header
1863 * Description : Executes all text substitutions from all applying
1864 * +(server|client)-header-filter actions on the header.
1865 * Most of the code was copied from pcrs_filter_response,
1866 * including the rather short variable names
1869 * 1 : csp = Current client state (buffers, headers, etc...)
1870 * 2 : header = On input, pointer to header to modify.
1871 * On output, pointer to the modified header, or NULL
1872 * to remove the header. This function frees the
1873 * original string if necessary.
1875 * Returns : JB_ERR_OK on success and always succeeds
1877 *********************************************************************/
1878 static jb_err filter_header(struct client_state *csp, char **header)
1882 size_t size = strlen(*header);
1884 char *newheader = NULL;
1887 struct file_list *fl;
1888 struct re_filterfile_spec *b;
1889 struct list_entry *filtername;
1891 int i, found_filters = 0;
1892 int wanted_filter_type;
1893 int multi_action_index;
1895 if (csp->flags & CSP_FLAG_NO_FILTERING)
1900 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
1902 wanted_filter_type = FT_SERVER_HEADER_FILTER;
1903 multi_action_index = ACTION_MULTI_SERVER_HEADER_FILTER;
1907 wanted_filter_type = FT_CLIENT_HEADER_FILTER;
1908 multi_action_index = ACTION_MULTI_CLIENT_HEADER_FILTER;
1912 * Need to check the set of re_filterfiles...
1914 for (i = 0; i < MAX_AF_FILES; i++)
1927 if (0 == found_filters)
1929 log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
1930 "header filtering enabled, but no matching filters available.");
1934 for (i = 0; i < MAX_AF_FILES; i++)
1937 if ((NULL == fl) || (NULL == fl->f))
1940 * Either there are no filter files
1941 * left, or this filter file just
1942 * contains no valid filters.
1944 * Continue to be sure we don't miss
1945 * valid filter files that are chained
1946 * after empty or invalid ones.
1951 * For all applying +filter actions, look if a filter by that
1952 * name exists and if yes, execute its pcrs_joblist on the
1955 for (b = fl->f; b; b = b->next)
1957 if (b->type != wanted_filter_type)
1959 /* Skip other filter types */
1963 for (filtername = csp->action->multi[multi_action_index]->first;
1964 filtername ; filtername = filtername->next)
1966 if (strcmp(b->name, filtername->str) == 0)
1968 int current_hits = 0;
1969 pcrs_job *joblist = b->joblist;
1971 if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
1973 if (NULL == joblist)
1975 log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
1979 log_error(LOG_LEVEL_RE_FILTER, "filtering \'%s\' (size %d) with \'%s\' ...",
1980 *header, size, b->name);
1982 /* Apply all jobs from the joblist */
1983 for (job = joblist; NULL != job; job = job->next)
1985 matches = pcrs_execute(job, *header, size, &newheader, &size);
1988 current_hits += matches;
1989 log_error(LOG_LEVEL_HEADER, "Transforming \"%s\" to \"%s\"", *header, newheader);
1991 *header = newheader;
1993 else if ( 0 == matches )
1995 /* Filter doesn't change header */
2001 log_error(LOG_LEVEL_ERROR, "Filtering \'%s\' with \'%s\' didn't work out: %s",
2002 *header, b->name, pcrs_strerror(matches));
2003 if (newheader != NULL)
2005 log_error(LOG_LEVEL_ERROR, "Freeing what's left: %s", newheader);
2011 if (b->dynamic) pcrs_free_joblist(joblist);
2013 log_error(LOG_LEVEL_RE_FILTER, "... produced %d hits (new size %d).", current_hits, size);
2014 hits += current_hits;
2021 * Additionally checking for hits is important because if
2022 * the continue hack is triggered, server headers can
2023 * arrive empty to separate multiple heads from each other.
2025 if ((0 == size) && hits)
2027 log_error(LOG_LEVEL_HEADER, "Removing empty header %s", *header);
2035 /*********************************************************************
2037 * Function : connection
2039 * Description : Makes sure that the value of the Connection: header
2040 * is "close" and signals connection_close_adder
2044 * 1 : csp = Current client state (buffers, headers, etc...)
2045 * 2 : header = On input, pointer to header to modify.
2046 * On output, pointer to the modified header, or NULL
2047 * to remove the header. This function frees the
2048 * original string if necessary.
2050 * Returns : JB_ERR_OK on success, or
2051 * JB_ERR_MEMORY on out-of-memory error.
2053 *********************************************************************/
2054 static jb_err connection(struct client_state *csp, char **header)
2056 char *old_header = *header;
2058 /* Do we have a 'Connection: close' header? */
2059 if (strcmpic(*header, "Connection: close"))
2061 /* No, create one */
2062 *header = strdup("Connection: close");
2065 return JB_ERR_MEMORY;
2067 log_error(LOG_LEVEL_HEADER, "Replaced: \'%s\' with \'%s\'", old_header, *header);
2071 /* Signal connection_close_adder() to return early. */
2072 if (csp->flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
2074 csp->flags |= CSP_FLAG_SERVER_CONNECTION_CLOSE_SET;
2078 csp->flags |= CSP_FLAG_CLIENT_CONNECTION_CLOSE_SET;
2085 /*********************************************************************
2087 * Function : crumble
2089 * Description : This is called if a header matches a pattern to "crunch"
2092 * 1 : csp = Current client state (buffers, headers, etc...)
2093 * 2 : header = On input, pointer to header to modify.
2094 * On output, pointer to the modified header, or NULL
2095 * to remove the header. This function frees the
2096 * original string if necessary.
2098 * Returns : JB_ERR_OK on success, or
2099 * JB_ERR_MEMORY on out-of-memory error.
2101 *********************************************************************/
2102 static jb_err crumble(struct client_state *csp, char **header)
2104 log_error(LOG_LEVEL_HEADER, "crumble crunched: %s!", *header);
2110 /*********************************************************************
2112 * Function : crunch_server_header
2114 * Description : Crunch server header if it matches a string supplied by the
2115 * user. Called from `sed'.
2118 * 1 : csp = Current client state (buffers, headers, etc...)
2119 * 2 : header = On input, pointer to header to modify.
2120 * On output, pointer to the modified header, or NULL
2121 * to remove the header. This function frees the
2122 * original string if necessary.
2124 * Returns : JB_ERR_OK on success and always succeeds
2126 *********************************************************************/
2127 static jb_err crunch_server_header(struct client_state *csp, char **header)
2129 const char *crunch_pattern;
2131 /* Do we feel like crunching? */
2132 if ((csp->action->flags & ACTION_CRUNCH_SERVER_HEADER))
2134 crunch_pattern = csp->action->string[ACTION_STRING_SERVER_HEADER];
2136 /* Is the current header the lucky one? */
2137 if (strstr(*header, crunch_pattern))
2139 log_error(LOG_LEVEL_HEADER, "Crunching server header: %s (contains: %s)", *header, crunch_pattern);
2148 /*********************************************************************
2150 * Function : server_content_type
2152 * Description : Set the content-type for filterable types (text/.*,
2153 * .*xml.*, javascript and image/gif) unless filtering has been
2154 * forbidden (CT_TABOO) while parsing earlier headers.
2155 * NOTE: Since text/plain is commonly used by web servers
2156 * for files whose correct type is unknown, we don't
2157 * set CT_TEXT for it.
2160 * 1 : csp = Current client state (buffers, headers, etc...)
2161 * 2 : header = On input, pointer to header to modify.
2162 * On output, pointer to the modified header, or NULL
2163 * to remove the header. This function frees the
2164 * original string if necessary.
2166 * Returns : JB_ERR_OK on success, or
2167 * JB_ERR_MEMORY on out-of-memory error.
2169 *********************************************************************/
2170 static jb_err server_content_type(struct client_state *csp, char **header)
2172 /* Remove header if it isn't the first Content-Type header */
2173 if ((csp->content_type & CT_DECLARED))
2176 * Another, slightly slower, way to see if
2177 * we already parsed another Content-Type header.
2179 assert(NULL != get_header_value(csp->headers, "Content-Type:"));
2181 log_error(LOG_LEVEL_ERROR,
2182 "Multiple Content-Type headers. Removing and ignoring: \'%s\'",
2190 * Signal that the Content-Type has been set.
2192 csp->content_type |= CT_DECLARED;
2194 if (!(csp->content_type & CT_TABOO))
2197 * XXX: The assumption that text/plain is a sign of
2198 * binary data seems to be somewhat unreasonable nowadays
2199 * and should be dropped after 3.0.8 is out.
2201 if ((strstr(*header, "text/") && !strstr(*header, "plain"))
2202 || strstr(*header, "xml")
2203 || strstr(*header, "application/x-javascript"))
2205 csp->content_type |= CT_TEXT;
2207 else if (strstr(*header, "image/gif"))
2209 csp->content_type |= CT_GIF;
2211 else if (strstr(*header, "image/jpeg"))
2213 csp->content_type |= CT_JPEG;
2218 * Are we messing with the content type?
2220 if (csp->action->flags & ACTION_CONTENT_TYPE_OVERWRITE)
2223 * Make sure the user doesn't accidently
2224 * change the content type of binary documents.
2226 if ((csp->content_type & CT_TEXT) || (csp->action->flags & ACTION_FORCE_TEXT_MODE))
2229 *header = strdup("Content-Type: ");
2230 string_append(header, csp->action->string[ACTION_STRING_CONTENT_TYPE]);
2234 log_error(LOG_LEVEL_HEADER, "Insufficient memory to replace Content-Type!");
2235 return JB_ERR_MEMORY;
2237 log_error(LOG_LEVEL_HEADER, "Modified: %s!", *header);
2241 log_error(LOG_LEVEL_HEADER, "%s not replaced. "
2242 "It doesn't look like a content type that should be filtered. "
2243 "Enable force-text-mode if you know what you're doing.", *header);
2251 /*********************************************************************
2253 * Function : server_transfer_coding
2255 * Description : - Prohibit filtering (CT_TABOO) if transfer coding compresses
2256 * - Raise the CSP_FLAG_CHUNKED flag if coding is "chunked"
2257 * - Remove header if body was chunked but has been
2258 * de-chunked for filtering.
2261 * 1 : csp = Current client state (buffers, headers, etc...)
2262 * 2 : header = On input, pointer to header to modify.
2263 * On output, pointer to the modified header, or NULL
2264 * to remove the header. This function frees the
2265 * original string if necessary.
2267 * Returns : JB_ERR_OK on success, or
2268 * JB_ERR_MEMORY on out-of-memory error.
2270 *********************************************************************/
2271 static jb_err server_transfer_coding(struct client_state *csp, char **header)
2274 * Turn off pcrs and gif filtering if body compressed
2276 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2280 * XXX: Added to test if we could use CT_GZIP and CT_DEFLATE here.
2282 log_error(LOG_LEVEL_INFO, "Marking content type for %s as CT_TABOO because of %s.",
2283 csp->http->cmd, *header);
2284 #endif /* def FEATURE_ZLIB */
2285 csp->content_type = CT_TABOO;
2289 * Raise flag if body chunked
2291 if (strstr(*header, "chunked"))
2293 csp->flags |= CSP_FLAG_CHUNKED;
2296 * If the body was modified, it has been de-chunked first
2297 * and the header must be removed.
2299 * FIXME: If there is more than one transfer encoding,
2300 * only the "chunked" part should be removed here.
2302 if (csp->flags & CSP_FLAG_MODIFIED)
2304 log_error(LOG_LEVEL_HEADER, "Removing: %s", *header);
2313 /*********************************************************************
2315 * Function : server_content_encoding
2317 * Description : This function is run twice for each request,
2318 * unless FEATURE_ZLIB and filtering are disabled.
2320 * The first run is used to check if the content
2321 * is compressed, if FEATURE_ZLIB is disabled
2322 * filtering is then disabled as well, if FEATURE_ZLIB
2323 * is enabled the content is marked for decompression.
2325 * The second run is used to remove the Content-Encoding
2326 * header if the decompression was successful.
2329 * 1 : csp = Current client state (buffers, headers, etc...)
2330 * 2 : header = On input, pointer to header to modify.
2331 * On output, pointer to the modified header, or NULL
2332 * to remove the header. This function frees the
2333 * original string if necessary.
2335 * Returns : JB_ERR_OK on success, or
2336 * JB_ERR_MEMORY on out-of-memory error.
2338 *********************************************************************/
2339 static jb_err server_content_encoding(struct client_state *csp, char **header)
2342 if ((csp->flags & CSP_FLAG_MODIFIED)
2343 && (csp->content_type & (CT_GZIP | CT_DEFLATE)))
2346 * We successfully decompressed the content,
2347 * and have to clean the header now, so the
2348 * client no longer expects compressed data..
2350 * XXX: There is a difference between cleaning
2351 * and removing it completely.
2353 log_error(LOG_LEVEL_HEADER, "Crunching: %s", *header);
2356 else if (strstr(*header, "gzip"))
2358 /* Mark for gzip decompression */
2359 csp->content_type |= CT_GZIP;
2361 else if (strstr(*header, "deflate"))
2363 /* Mark for zlib decompression */
2364 csp->content_type |= CT_DEFLATE;
2366 else if (strstr(*header, "compress"))
2369 * We can't decompress this; therefore we can't filter
2372 csp->content_type |= CT_TABOO;
2374 #else /* !defined(FEATURE_ZLIB) */
2375 if (strstr(*header, "gzip") || strstr(*header, "compress") || strstr(*header, "deflate"))
2378 * Body is compressed, turn off pcrs and gif filtering.
2380 csp->content_type |= CT_TABOO;
2383 * Log a warning if the user expects the content to be filtered.
2385 if ((csp->rlist != NULL) &&
2386 (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
2388 log_error(LOG_LEVEL_INFO,
2389 "Compressed content detected, content filtering disabled. "
2390 "Consider recompiling Privoxy with zlib support or "
2391 "enable the prevent-compression action.");
2394 #endif /* defined(FEATURE_ZLIB) */
2401 /*********************************************************************
2403 * Function : server_content_length
2405 * Description : Adjust Content-Length header if we modified
2409 * 1 : csp = Current client state (buffers, headers, etc...)
2410 * 2 : header = On input, pointer to header to modify.
2411 * On output, pointer to the modified header, or NULL
2412 * to remove the header. This function frees the
2413 * original string if necessary.
2415 * Returns : JB_ERR_OK on success, or
2416 * JB_ERR_MEMORY on out-of-memory error.
2418 *********************************************************************/
2419 static jb_err server_content_length(struct client_state *csp, char **header)
2421 const size_t max_header_length = 80;
2423 /* Regenerate header if the content was modified. */
2424 if (csp->flags & CSP_FLAG_MODIFIED)
2427 *header = (char *) zalloc(max_header_length);
2428 if (*header == NULL)
2430 return JB_ERR_MEMORY;
2433 snprintf(*header, max_header_length, "Content-Length: %d",
2434 (int)csp->content_length);
2435 log_error(LOG_LEVEL_HEADER, "Adjusted Content-Length to %d",
2436 (int)csp->content_length);
2443 /*********************************************************************
2445 * Function : server_content_md5
2447 * Description : Crumble any Content-MD5 headers if the document was
2448 * modified. FIXME: Should we re-compute instead?
2451 * 1 : csp = Current client state (buffers, headers, etc...)
2452 * 2 : header = On input, pointer to header to modify.
2453 * On output, pointer to the modified header, or NULL
2454 * to remove the header. This function frees the
2455 * original string if necessary.
2457 * Returns : JB_ERR_OK on success, or
2458 * JB_ERR_MEMORY on out-of-memory error.
2460 *********************************************************************/
2461 static jb_err server_content_md5(struct client_state *csp, char **header)
2463 if (csp->flags & CSP_FLAG_MODIFIED)
2465 log_error(LOG_LEVEL_HEADER, "Crunching Content-MD5");
2473 /*********************************************************************
2475 * Function : server_content_disposition
2477 * Description : If enabled, blocks or modifies the "Content-Disposition" header.
2478 * Called from `sed'.
2481 * 1 : csp = Current client state (buffers, headers, etc...)
2482 * 2 : header = On input, pointer to header to modify.
2483 * On output, pointer to the modified header, or NULL
2484 * to remove the header. This function frees the
2485 * original string if necessary.
2487 * Returns : JB_ERR_OK on success, or
2488 * JB_ERR_MEMORY on out-of-memory error.
2490 *********************************************************************/
2491 static jb_err server_content_disposition(struct client_state *csp, char **header)
2496 * Are we messing with the Content-Disposition header?
2498 if ((csp->action->flags & ACTION_HIDE_CONTENT_DISPOSITION) == 0)
2504 newval = csp->action->string[ACTION_STRING_CONTENT_DISPOSITION];
2506 if ((newval == NULL) || (0 == strcmpic(newval, "block")))
2509 * Blocking content-disposition header
2511 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2518 * Replacing Content-Disposition header
2521 *header = strdup("Content-Disposition: ");
2522 string_append(header, newval);
2524 if (*header != NULL)
2526 log_error(LOG_LEVEL_HEADER,
2527 "Content-Disposition header crunched and replaced with: %s", *header);
2530 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2534 /*********************************************************************
2536 * Function : server_last_modified
2538 * Description : Changes Last-Modified header to the actual date
2539 * to help hide-if-modified-since.
2540 * Called from `sed'.
2543 * 1 : csp = Current client state (buffers, headers, etc...)
2544 * 2 : header = On input, pointer to header to modify.
2545 * On output, pointer to the modified header, or NULL
2546 * to remove the header. This function frees the
2547 * original string if necessary.
2549 * Returns : JB_ERR_OK on success, or
2550 * JB_ERR_MEMORY on out-of-memory error.
2552 *********************************************************************/
2553 static jb_err server_last_modified(struct client_state *csp, char **header)
2556 char buf[BUFFER_SIZE];
2559 #ifdef HAVE_GMTIME_R
2562 struct tm *timeptr = NULL;
2563 time_t now, last_modified;
2565 long int days, hours, minutes, seconds;
2568 * Are we messing with the Last-Modified header?
2570 if ((csp->action->flags & ACTION_OVERWRITE_LAST_MODIFIED) == 0)
2576 newval = csp->action->string[ACTION_STRING_LAST_MODIFIED];
2578 if (0 == strcmpic(newval, "block") )
2581 * Blocking Last-Modified header. Useless but why not.
2583 log_error(LOG_LEVEL_HEADER, "Crunching %s!", *header);
2587 else if (0 == strcmpic(newval, "reset-to-request-time"))
2590 * Setting Last-Modified Header to now.
2592 get_http_time(0, buf);
2594 *header = strdup("Last-Modified: ");
2595 string_append(header, buf);
2597 if (*header == NULL)
2599 log_error(LOG_LEVEL_HEADER, "Insufficent memory. Last-Modified header got lost, boohoo.");
2603 log_error(LOG_LEVEL_HEADER, "Reset to present time: %s", *header);
2606 else if (0 == strcmpic(newval, "randomize"))
2608 const char *header_time = *header + sizeof("Last-Modified:");
2610 log_error(LOG_LEVEL_HEADER, "Randomizing: %s", *header);
2612 #ifdef HAVE_GMTIME_R
2613 timeptr = gmtime_r(&now, &gmt);
2614 #elif FEATURE_PTHREAD
2615 pthread_mutex_lock(&gmtime_mutex);
2616 timeptr = gmtime(&now);
2617 pthread_mutex_unlock(&gmtime_mutex);
2619 timeptr = gmtime(&now);
2621 if (JB_ERR_OK != parse_header_time(header_time, &last_modified))
2623 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
2628 rtime = (long int)difftime(now, last_modified);
2637 log_error(LOG_LEVEL_HEADER, "Server time in the future.");
2639 rtime = pick_from_range(rtime);
2640 if (negative) rtime *= -1;
2641 last_modified += rtime;
2642 #ifdef HAVE_GMTIME_R
2643 timeptr = gmtime_r(&last_modified, &gmt);
2644 #elif FEATURE_PTHREAD
2645 pthread_mutex_lock(&gmtime_mutex);
2646 timeptr = gmtime(&last_modified);
2647 pthread_mutex_unlock(&gmtime_mutex);
2649 timeptr = gmtime(&last_modified);
2651 strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr);
2653 *header = strdup("Last-Modified: ");
2654 string_append(header, newheader);
2656 if (*header == NULL)
2658 log_error(LOG_LEVEL_ERROR, "Insufficent memory, header crunched without replacement.");
2659 return JB_ERR_MEMORY;
2662 if (LOG_LEVEL_HEADER & debug) /* Save cycles if the user isn't interested. */
2664 days = rtime / (3600 * 24);
2665 hours = rtime / 3600 % 24;
2666 minutes = rtime / 60 % 60;
2667 seconds = rtime % 60;
2669 log_error(LOG_LEVEL_HEADER, "Randomized: %s (added %d da%s %d hou%s %d minut%s %d second%s",
2670 *header, days, (days == 1) ? "y" : "ys", hours, (hours == 1) ? "r" : "rs",
2671 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
2676 log_error(LOG_LEVEL_HEADER, "Randomized ... or not. No time difference to work with.");
2685 /*********************************************************************
2687 * Function : client_accept_encoding
2689 * Description : Rewrite the client's Accept-Encoding header so that
2690 * if doesn't allow compression, if the action applies.
2691 * Note: For HTTP/1.0 the absence of the header is enough.
2694 * 1 : csp = Current client state (buffers, headers, etc...)
2695 * 2 : header = On input, pointer to header to modify.
2696 * On output, pointer to the modified header, or NULL
2697 * to remove the header. This function frees the
2698 * original string if necessary.
2700 * Returns : JB_ERR_OK on success, or
2701 * JB_ERR_MEMORY on out-of-memory error.
2703 *********************************************************************/
2704 static jb_err client_accept_encoding(struct client_state *csp, char **header)
2706 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2708 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress content");
2712 /* Temporarily disable the correct behaviour to
2713 * work around a PHP bug.
2715 * if (!strcmpic(csp->http->ver, "HTTP/1.1"))
2717 * *header = strdup("Accept-Encoding: identity;q=1.0, *;q=0");
2718 * if (*header == NULL)
2720 * return JB_ERR_MEMORY;
2731 /*********************************************************************
2733 * Function : client_te
2735 * Description : Rewrite the client's TE header so that
2736 * if doesn't allow compression, if the action applies.
2739 * 1 : csp = Current client state (buffers, headers, etc...)
2740 * 2 : header = On input, pointer to header to modify.
2741 * On output, pointer to the modified header, or NULL
2742 * to remove the header. This function frees the
2743 * original string if necessary.
2745 * Returns : JB_ERR_OK on success, or
2746 * JB_ERR_MEMORY on out-of-memory error.
2748 *********************************************************************/
2749 static jb_err client_te(struct client_state *csp, char **header)
2751 if ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
2754 log_error(LOG_LEVEL_HEADER, "Suppressed offer to compress transfer");
2761 /*********************************************************************
2763 * Function : client_referrer
2765 * Description : Handle the "referer" config setting properly.
2766 * Called from `sed'.
2769 * 1 : csp = Current client state (buffers, headers, etc...)
2770 * 2 : header = On input, pointer to header to modify.
2771 * On output, pointer to the modified header, or NULL
2772 * to remove the header. This function frees the
2773 * original string if necessary.
2775 * Returns : JB_ERR_OK on success, or
2776 * JB_ERR_MEMORY on out-of-memory error.
2778 *********************************************************************/
2779 static jb_err client_referrer(struct client_state *csp, char **header)
2781 const char *parameter;
2782 /* booleans for parameters we have to check multiple times */
2783 int parameter_conditional_block;
2784 int parameter_conditional_forge;
2786 #ifdef FEATURE_FORCE_LOAD
2788 * Since the referrer can include the prefix even
2789 * if the request itself is non-forced, we must
2790 * clean it unconditionally.
2792 * XXX: strclean is too broad
2794 strclean(*header, FORCE_PREFIX);
2795 #endif /* def FEATURE_FORCE_LOAD */
2797 if ((csp->action->flags & ACTION_HIDE_REFERER) == 0)
2799 /* Nothing left to do */
2803 parameter = csp->action->string[ACTION_STRING_REFERER];
2804 assert(parameter != NULL);
2805 parameter_conditional_block = (0 == strcmpic(parameter, "conditional-block"));
2806 parameter_conditional_forge = (0 == strcmpic(parameter, "conditional-forge"));
2808 if (!parameter_conditional_block && !parameter_conditional_forge)
2811 * As conditional-block and conditional-forge are the only
2812 * parameters that rely on the original referrer, we can
2813 * remove it now for all the others.
2818 if (0 == strcmpic(parameter, "block"))
2820 log_error(LOG_LEVEL_HEADER, "Referer crunched!");
2823 else if (parameter_conditional_block || parameter_conditional_forge)
2825 return handle_conditional_hide_referrer_parameter(header,
2826 csp->http->hostport, parameter_conditional_block);
2828 else if (0 == strcmpic(parameter, "forge"))
2830 return create_forged_referrer(header, csp->http->hostport);
2834 /* interpret parameter as user-supplied referer to fake */
2835 return create_fake_referrer(header, parameter);
2840 /*********************************************************************
2842 * Function : client_accept_language
2844 * Description : Handle the "Accept-Language" config setting properly.
2845 * Called from `sed'.
2848 * 1 : csp = Current client state (buffers, headers, etc...)
2849 * 2 : header = On input, pointer to header to modify.
2850 * On output, pointer to the modified header, or NULL
2851 * to remove the header. This function frees the
2852 * original string if necessary.
2854 * Returns : JB_ERR_OK on success, or
2855 * JB_ERR_MEMORY on out-of-memory error.
2857 *********************************************************************/
2858 static jb_err client_accept_language(struct client_state *csp, char **header)
2863 * Are we messing with the Accept-Language?
2865 if ((csp->action->flags & ACTION_HIDE_ACCEPT_LANGUAGE) == 0)
2867 /*I don't think so*/
2871 newval = csp->action->string[ACTION_STRING_LANGUAGE];
2873 if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
2876 * Blocking Accept-Language header
2878 log_error(LOG_LEVEL_HEADER, "Crunching Accept-Language!");
2885 * Replacing Accept-Language header
2888 *header = strdup("Accept-Language: ");
2889 string_append(header, newval);
2891 if (*header == NULL)
2893 log_error(LOG_LEVEL_ERROR,
2894 "Insufficent memory. Accept-Language header crunched without replacement.");
2898 log_error(LOG_LEVEL_HEADER,
2899 "Accept-Language header crunched and replaced with: %s", *header);
2902 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2906 /*********************************************************************
2908 * Function : crunch_client_header
2910 * Description : Crunch client header if it matches a string supplied by the
2911 * user. Called from `sed'.
2914 * 1 : csp = Current client state (buffers, headers, etc...)
2915 * 2 : header = On input, pointer to header to modify.
2916 * On output, pointer to the modified header, or NULL
2917 * to remove the header. This function frees the
2918 * original string if necessary.
2920 * Returns : JB_ERR_OK on success and always succeeds
2922 *********************************************************************/
2923 static jb_err crunch_client_header(struct client_state *csp, char **header)
2925 const char *crunch_pattern;
2927 /* Do we feel like crunching? */
2928 if ((csp->action->flags & ACTION_CRUNCH_CLIENT_HEADER))
2930 crunch_pattern = csp->action->string[ACTION_STRING_CLIENT_HEADER];
2932 /* Is the current header the lucky one? */
2933 if (strstr(*header, crunch_pattern))
2935 log_error(LOG_LEVEL_HEADER, "Crunching client header: %s (contains: %s)", *header, crunch_pattern);
2943 /*********************************************************************
2945 * Function : client_uagent
2947 * Description : Handle the "user-agent" config setting properly
2948 * and remember its original value to enable browser
2949 * bug workarounds. Called from `sed'.
2952 * 1 : csp = Current client state (buffers, headers, etc...)
2953 * 2 : header = On input, pointer to header to modify.
2954 * On output, pointer to the modified header, or NULL
2955 * to remove the header. This function frees the
2956 * original string if necessary.
2958 * Returns : JB_ERR_OK on success, or
2959 * JB_ERR_MEMORY on out-of-memory error.
2961 *********************************************************************/
2962 static jb_err client_uagent(struct client_state *csp, char **header)
2966 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) == 0)
2971 newval = csp->action->string[ACTION_STRING_USER_AGENT];
2978 *header = strdup("User-Agent: ");
2979 string_append(header, newval);
2981 log_error(LOG_LEVEL_HEADER, "Modified: %s", *header);
2983 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
2987 /*********************************************************************
2989 * Function : client_ua
2991 * Description : Handle "ua-" headers properly. Called from `sed'.
2994 * 1 : csp = Current client state (buffers, headers, etc...)
2995 * 2 : header = On input, pointer to header to modify.
2996 * On output, pointer to the modified header, or NULL
2997 * to remove the header. This function frees the
2998 * original string if necessary.
3000 * Returns : JB_ERR_OK on success, or
3001 * JB_ERR_MEMORY on out-of-memory error.
3003 *********************************************************************/
3004 static jb_err client_ua(struct client_state *csp, char **header)
3006 if ((csp->action->flags & ACTION_HIDE_USER_AGENT) != 0)
3008 log_error(LOG_LEVEL_HEADER, "crunched User-Agent!");
3016 /*********************************************************************
3018 * Function : client_from
3020 * Description : Handle the "from" config setting properly.
3021 * Called from `sed'.
3024 * 1 : csp = Current client state (buffers, headers, etc...)
3025 * 2 : header = On input, pointer to header to modify.
3026 * On output, pointer to the modified header, or NULL
3027 * to remove the header. This function frees the
3028 * original string if necessary.
3030 * Returns : JB_ERR_OK on success, or
3031 * JB_ERR_MEMORY on out-of-memory error.
3033 *********************************************************************/
3034 static jb_err client_from(struct client_state *csp, char **header)
3038 if ((csp->action->flags & ACTION_HIDE_FROM) == 0)
3045 newval = csp->action->string[ACTION_STRING_FROM];
3048 * Are we blocking the e-mail address?
3050 if ((newval == NULL) || (0 == strcmpic(newval, "block")) )
3052 log_error(LOG_LEVEL_HEADER, "crunched From!");
3056 log_error(LOG_LEVEL_HEADER, " modified");
3058 *header = strdup("From: ");
3059 string_append(header, newval);
3061 return (*header == NULL) ? JB_ERR_MEMORY : JB_ERR_OK;
3065 /*********************************************************************
3067 * Function : client_send_cookie
3069 * Description : Crunches the "cookie" header if necessary.
3070 * Called from `sed'.
3072 * XXX: Stupid name, doesn't send squat.
3075 * 1 : csp = Current client state (buffers, headers, etc...)
3076 * 2 : header = On input, pointer to header to modify.
3077 * On output, pointer to the modified header, or NULL
3078 * to remove the header. This function frees the
3079 * original string if necessary.
3081 * Returns : JB_ERR_OK on success, or
3082 * JB_ERR_MEMORY on out-of-memory error.
3084 *********************************************************************/
3085 static jb_err client_send_cookie(struct client_state *csp, char **header)
3087 if (csp->action->flags & ACTION_NO_COOKIE_READ)
3089 log_error(LOG_LEVEL_HEADER, "Crunched outgoing cookie: %s", *header);
3097 /*********************************************************************
3099 * Function : client_x_forwarded
3101 * Description : Handle the "x-forwarded-for" config setting properly,
3102 * also used in the add_client_headers list. Called from `sed'.
3105 * 1 : csp = Current client state (buffers, headers, etc...)
3106 * 2 : header = On input, pointer to header to modify.
3107 * On output, pointer to the modified header, or NULL
3108 * to remove the header. This function frees the
3109 * original string if necessary.
3111 * Returns : JB_ERR_OK on success, or
3112 * JB_ERR_MEMORY on out-of-memory error.
3114 *********************************************************************/
3115 jb_err client_x_forwarded(struct client_state *csp, char **header)
3117 if ((csp->action->flags & ACTION_HIDE_FORWARDED) != 0)
3120 log_error(LOG_LEVEL_HEADER, "crunched x-forwarded-for!");
3127 /*********************************************************************
3129 * Function : client_max_forwards
3131 * Description : If the HTTP method is OPTIONS or TRACE, subtract one
3132 * from the value of the Max-Forwards header field.
3135 * 1 : csp = Current client state (buffers, headers, etc...)
3136 * 2 : header = On input, pointer to header to modify.
3137 * On output, pointer to the modified header, or NULL
3138 * to remove the header. This function frees the
3139 * original string if necessary.
3141 * Returns : JB_ERR_OK on success, or
3142 * JB_ERR_MEMORY on out-of-memory error.
3144 *********************************************************************/
3145 static jb_err client_max_forwards(struct client_state *csp, char **header)
3149 if ((0 == strcmpic(csp->http->gpc, "trace")) ||
3150 (0 == strcmpic(csp->http->gpc, "options")))
3152 assert(*(*header+12) == ':');
3153 if (1 == sscanf(*header+12, ": %u", &max_forwards))
3155 if (max_forwards > 0)
3157 snprintf(*header, strlen(*header)+1, "Max-Forwards: %u", --max_forwards);
3158 log_error(LOG_LEVEL_HEADER, "Max-Forwards value for %s request reduced to %u.",
3159 csp->http->gpc, max_forwards);
3161 else if (max_forwards < 0)
3163 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3169 log_error(LOG_LEVEL_ERROR, "Crunching invalid header: %s", *header);
3178 /*********************************************************************
3180 * Function : client_host
3182 * Description : If the request URI did not contain host and
3183 * port information, parse and evaluate the Host
3186 * Also, kill ill-formed HOST: headers as sent by
3187 * Apple's iTunes software when used with a proxy.
3190 * 1 : csp = Current client state (buffers, headers, etc...)
3191 * 2 : header = On input, pointer to header to modify.
3192 * On output, pointer to the modified header, or NULL
3193 * to remove the header. This function frees the
3194 * original string if necessary.
3196 * Returns : JB_ERR_OK on success, or
3197 * JB_ERR_MEMORY on out-of-memory error.
3199 *********************************************************************/
3200 static jb_err client_host(struct client_state *csp, char **header)
3205 * If the header field name is all upper-case, chances are that it's
3206 * an ill-formed one from iTunes. BTW, killing innocent headers here is
3207 * not a problem -- they are regenerated later.
3209 if ((*header)[1] == 'O')
3211 log_error(LOG_LEVEL_HEADER, "Killed all-caps Host header line: %s", *header);
3216 if (!csp->http->hostport || (*csp->http->hostport == '*') ||
3217 *csp->http->hostport == ' ' || *csp->http->hostport == '\0')
3220 if (NULL == (p = strdup((*header)+6)))
3222 return JB_ERR_MEMORY;
3225 if (NULL == (q = strdup(p)))
3228 return JB_ERR_MEMORY;
3231 freez(csp->http->hostport);
3232 csp->http->hostport = p;
3233 freez(csp->http->host);
3234 csp->http->host = q;
3235 q = strchr(csp->http->host, ':');
3238 /* Terminate hostname and evaluate port string */
3240 csp->http->port = atoi(q);
3244 csp->http->port = csp->http->ssl ? 443 : 80;
3247 log_error(LOG_LEVEL_HEADER, "New host and port from Host field: %s = %s:%d",
3248 csp->http->hostport, csp->http->host, csp->http->port);
3251 /* Signal client_host_adder() to return right away */
3252 csp->flags |= CSP_FLAG_HOST_HEADER_IS_SET;
3258 /*********************************************************************
3260 * Function : client_if_modified_since
3262 * Description : Remove or modify the If-Modified-Since header.
3265 * 1 : csp = Current client state (buffers, headers, etc...)
3266 * 2 : header = On input, pointer to header to modify.
3267 * On output, pointer to the modified header, or NULL
3268 * to remove the header. This function frees the
3269 * original string if necessary.
3271 * Returns : JB_ERR_OK on success, or
3272 * JB_ERR_MEMORY on out-of-memory error.
3274 *********************************************************************/
3275 static jb_err client_if_modified_since(struct client_state *csp, char **header)
3278 #ifdef HAVE_GMTIME_R
3281 struct tm *timeptr = NULL;
3285 long int hours, minutes, seconds;
3289 if ( 0 == strcmpic(*header, "If-Modified-Since: Wed, 08 Jun 1955 12:00:00 GMT"))
3292 * The client got an error message because of a temporary problem,
3293 * the problem is gone and the client now tries to revalidate our
3294 * error message on the real server. The revalidation would always
3295 * end with the transmission of the whole document and there is
3296 * no need to expose the bogus If-Modified-Since header.
3298 log_error(LOG_LEVEL_HEADER, "Crunching useless If-Modified-Since header.");
3301 else if (csp->action->flags & ACTION_HIDE_IF_MODIFIED_SINCE)
3303 newval = csp->action->string[ACTION_STRING_IF_MODIFIED_SINCE];
3305 if ((0 == strcmpic(newval, "block")))
3307 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3310 else /* add random value */
3312 const char *header_time = *header + sizeof("If-Modified-Since:");
3314 if (JB_ERR_OK != parse_header_time(header_time, &tm))
3316 log_error(LOG_LEVEL_HEADER, "Couldn't parse: %s in %s (crunching!)", header_time, *header);
3321 rtime = strtol(newval, &endptr, 0);
3324 log_error(LOG_LEVEL_HEADER, "Randomizing: %s (random range: %d minut%s)",
3325 *header, rtime, (rtime == 1 || rtime == -1) ? "e": "es");
3332 rtime = pick_from_range(rtime);
3336 log_error(LOG_LEVEL_ERROR, "Random range is 0. Assuming time transformation test.",
3339 tm += rtime * (negative ? -1 : 1);
3340 #ifdef HAVE_GMTIME_R
3341 timeptr = gmtime_r(&tm, &gmt);
3342 #elif FEATURE_PTHREAD
3343 pthread_mutex_lock(&gmtime_mutex);
3344 timeptr = gmtime(&tm);
3345 pthread_mutex_unlock(&gmtime_mutex);
3347 timeptr = gmtime(&tm);
3349 strftime(newheader, sizeof(newheader), "%a, %d %b %Y %H:%M:%S GMT", timeptr);
3352 *header = strdup("If-Modified-Since: ");
3353 string_append(header, newheader);
3355 if (*header == NULL)
3357 log_error(LOG_LEVEL_HEADER, "Insufficent memory, header crunched without replacement.");
3358 return JB_ERR_MEMORY;
3361 if (LOG_LEVEL_HEADER & debug) /* Save cycles if the user isn't interested. */
3363 hours = rtime / 3600;
3364 minutes = rtime / 60 % 60;
3365 seconds = rtime % 60;
3367 log_error(LOG_LEVEL_HEADER, "Randomized: %s (%s %d hou%s %d minut%s %d second%s",
3368 *header, (negative) ? "subtracted" : "added", hours, (hours == 1) ? "r" : "rs",
3369 minutes, (minutes == 1) ? "e" : "es", seconds, (seconds == 1) ? ")" : "s)");
3379 /*********************************************************************
3381 * Function : client_if_none_match
3383 * Description : Remove the If-None-Match header.
3386 * 1 : csp = Current client state (buffers, headers, etc...)
3387 * 2 : header = On input, pointer to header to modify.
3388 * On output, pointer to the modified header, or NULL
3389 * to remove the header. This function frees the
3390 * original string if necessary.
3392 * Returns : JB_ERR_OK on success, or
3393 * JB_ERR_MEMORY on out-of-memory error.
3395 *********************************************************************/
3396 static jb_err client_if_none_match(struct client_state *csp, char **header)
3398 if (csp->action->flags & ACTION_CRUNCH_IF_NONE_MATCH)
3400 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3408 /*********************************************************************
3410 * Function : client_x_filter
3412 * Description : Disables filtering if the client set "X-Filter: No".
3413 * Called from `sed'.
3416 * 1 : csp = Current client state (buffers, headers, etc...)
3417 * 2 : header = On input, pointer to header to modify.
3418 * On output, pointer to the modified header, or NULL
3419 * to remove the header. This function frees the
3420 * original string if necessary.
3422 * Returns : JB_ERR_OK on success
3424 *********************************************************************/
3425 jb_err client_x_filter(struct client_state *csp, char **header)
3427 if ( 0 == strcmpic(*header, "X-Filter: No"))
3429 if (!(csp->config->feature_flags & RUNTIME_FEATURE_HTTP_TOGGLE))
3431 log_error(LOG_LEVEL_INFO, "Ignored the client's request to fetch without filtering.");
3435 if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
3437 log_error(LOG_LEVEL_HEADER,
3438 "force-text-mode overruled the client's request to fetch without filtering!");
3442 csp->content_type = CT_TABOO; /* XXX: This hack shouldn't be necessary */
3443 csp->flags |= CSP_FLAG_NO_FILTERING;
3444 log_error(LOG_LEVEL_HEADER, "Accepted the client's request to fetch without filtering.");
3446 log_error(LOG_LEVEL_HEADER, "Crunching %s", *header);
3454 /*********************************************************************
3456 * Function : client_range
3458 * Description : Removes Range, Request-Range and If-Range headers if
3459 * content filtering is enabled. If the client's version
3460 * of the document has been altered by Privoxy, the server
3461 * could interpret the range differently than the client
3462 * intended in which case the user could end up with
3463 * corrupted content.
3466 * 1 : csp = Current client state (buffers, headers, etc...)
3467 * 2 : header = On input, pointer to header to modify.
3468 * On output, pointer to the modified header, or NULL
3469 * to remove the header. This function frees the
3470 * original string if necessary.
3472 * Returns : JB_ERR_OK
3474 *********************************************************************/
3475 static jb_err client_range(struct client_state *csp, char **header)
3477 if (content_filters_enabled(csp))
3479 log_error(LOG_LEVEL_HEADER, "Content filtering is enabled."
3480 " Crunching: \'%s\' to prevent range-mismatch problems.", *header);
3487 /* the following functions add headers directly to the header list */
3489 /*********************************************************************
3491 * Function : client_host_adder
3493 * Description : Adds the Host: header field if it is missing.
3494 * Called from `sed'.
3497 * 1 : csp = Current client state (buffers, headers, etc...)
3499 * Returns : JB_ERR_OK on success, or
3500 * JB_ERR_MEMORY on out-of-memory error.
3502 *********************************************************************/
3503 static jb_err client_host_adder(struct client_state *csp)
3508 if (csp->flags & CSP_FLAG_HOST_HEADER_IS_SET)
3510 /* Header already set by the client, nothing to do. */
3514 if ( !csp->http->hostport || !*(csp->http->hostport))
3516 /* XXX: When does this happen and why is it OK? */
3517 log_error(LOG_LEVEL_INFO, "Weirdness in client_host_adder detected and ignored.");
3522 * remove 'user:pass@' from 'proto://user:pass@host'
3524 if ( (p = strchr( csp->http->hostport, '@')) != NULL )
3530 p = csp->http->hostport;
3533 /* XXX: Just add it, we already made sure that it will be unique */
3534 log_error(LOG_LEVEL_HEADER, "addh-unique: Host: %s", p);
3535 err = enlist_unique_header(csp->headers, "Host", p);
3541 /*********************************************************************
3543 * Function : client_cookie_adder
3545 * Description : Used in the add_client_headers list to add "wafers".
3546 * Called from `sed'.
3549 * 1 : csp = Current client state (buffers, headers, etc...)
3551 * Returns : JB_ERR_OK on success, or
3552 * JB_ERR_MEMORY on out-of-memory error.
3554 *********************************************************************/
3555 jb_err client_cookie_adder(struct client_state *csp)
3558 struct list_entry *wafer;
3559 struct list_entry *wafer_list;
3563 * If the user has not supplied any wafers, and the user has not
3564 * told us to suppress the vanilla wafer, then send the vanilla wafer.
3566 if ((0 != (csp->action->flags & ACTION_VANILLA_WAFER))
3567 && list_is_empty(csp->action->multi[ACTION_MULTI_WAFER]))
3569 enlist(csp->action->multi[ACTION_MULTI_WAFER], VANILLA_WAFER);
3572 wafer_list = csp->action->multi[ACTION_MULTI_WAFER]->first;
3574 if (NULL == wafer_list)
3580 tmp = strdup("Cookie: ");
3582 for (wafer = wafer_list; (NULL != tmp) && (NULL != wafer); wafer = wafer->next)
3584 if (wafer != wafer_list)
3586 /* As this isn't the first wafer, we need a delimiter. */
3587 string_append(&tmp, "; ");
3589 string_join(&tmp, cookie_encode(wafer->str));
3594 return JB_ERR_MEMORY;
3597 log_error(LOG_LEVEL_HEADER, "addh: %s", tmp);
3598 err = enlist(csp->headers, tmp);
3605 /*********************************************************************
3607 * Function : client_accept_encoding_adder
3609 * Description : Add an Accept-Encoding header to the client's request
3610 * that disables compression if the action applies, and
3611 * the header is not already there. Called from `sed'.
3612 * Note: For HTTP/1.0, the absence of the header is enough.
3615 * 1 : csp = Current client state (buffers, headers, etc...)
3617 * Returns : JB_ERR_OK on success, or
3618 * JB_ERR_MEMORY on out-of-memory error.
3620 *********************************************************************/
3621 static jb_err client_accept_encoding_adder(struct client_state *csp)
3623 if ( ((csp->action->flags & ACTION_NO_COMPRESSION) != 0)
3624 && (!strcmpic(csp->http->ver, "HTTP/1.1")) )
3626 return enlist_unique(csp->headers, "Accept-Encoding: identity;q=1.0, *;q=0", 16);
3634 /*********************************************************************
3636 * Function : client_xtra_adder
3638 * Description : Used in the add_client_headers list. Called from `sed'.
3641 * 1 : csp = Current client state (buffers, headers, etc...)
3643 * Returns : JB_ERR_OK on success, or
3644 * JB_ERR_MEMORY on out-of-memory error.
3646 *********************************************************************/
3647 static jb_err client_xtra_adder(struct client_state *csp)
3649 struct list_entry *lst;
3652 for (lst = csp->action->multi[ACTION_MULTI_ADD_HEADER]->first;
3653 lst ; lst = lst->next)
3655 log_error(LOG_LEVEL_HEADER, "addh: %s", lst->str);
3656 err = enlist(csp->headers, lst->str);
3668 /*********************************************************************
3670 * Function : connection_close_adder
3672 * Description : "Temporary" fix for the needed but missing HTTP/1.1
3673 * support. Adds a "Connection: close" header to csp->headers
3674 * unless the header was already present. Called from `sed'.
3676 * FIXME: This whole function shouldn't be neccessary!
3679 * 1 : csp = Current client state (buffers, headers, etc...)
3681 * Returns : JB_ERR_OK on success, or
3682 * JB_ERR_MEMORY on out-of-memory error.
3684 *********************************************************************/
3685 static jb_err connection_close_adder(struct client_state *csp)
3687 const unsigned int flags = csp->flags;
3690 * Return right away if
3692 * - we're parsing server headers and the server header
3693 * "Connection: close" is already set, or if
3695 * - we're parsing client headers and the client header
3696 * "Connection: close" is already set.
3698 if ((flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE
3699 && flags & CSP_FLAG_SERVER_CONNECTION_CLOSE_SET)
3700 ||(!(flags & CSP_FLAG_CLIENT_HEADER_PARSING_DONE)
3701 && flags & CSP_FLAG_CLIENT_CONNECTION_CLOSE_SET))
3706 log_error(LOG_LEVEL_HEADER, "Adding: Connection: close");
3708 return enlist(csp->headers, "Connection: close");
3712 /*********************************************************************
3714 * Function : server_http
3716 * Description : - Save the HTTP Status into csp->http->status
3717 * - Set CT_TABOO to prevent filtering if the answer
3718 * is a partial range (HTTP status 206)
3719 * - Rewrite HTTP/1.1 answers to HTTP/1.0 if +downgrade
3723 * 1 : csp = Current client state (buffers, headers, etc...)
3724 * 2 : header = On input, pointer to header to modify.
3725 * On output, pointer to the modified header, or NULL
3726 * to remove the header. This function frees the
3727 * original string if necessary.
3729 * Returns : JB_ERR_OK on success, or
3730 * JB_ERR_MEMORY on out-of-memory error.
3732 *********************************************************************/
3733 static jb_err server_http(struct client_state *csp, char **header)
3735 sscanf(*header, "HTTP/%*d.%*d %d", &(csp->http->status));
3736 if (csp->http->status == 206)
3738 csp->content_type = CT_TABOO;
3741 if ((csp->action->flags & ACTION_DOWNGRADE) != 0)
3743 /* XXX: Should we do a real validity check here? */
3744 if (strlen(*header) > 8)
3747 log_error(LOG_LEVEL_HEADER, "Downgraded answer to HTTP/1.0");
3752 * XXX: Should we block the request or
3753 * enlist a valid status code line here?
3755 log_error(LOG_LEVEL_INFO, "Malformed server response detected. "
3756 "Downgrading to HTTP/1.0 impossible.");
3764 /*********************************************************************
3766 * Function : server_set_cookie
3768 * Description : Handle the server "cookie" header properly.
3769 * Log cookie to the jar file. Then "crunch",
3770 * accept or rewrite it to a session cookie.
3771 * Called from `sed'.
3773 * TODO: Allow the user to specify a new expiration
3774 * time to cause the cookie to expire even before the
3775 * browser is closed.
3778 * 1 : csp = Current client state (buffers, headers, etc...)
3779 * 2 : header = On input, pointer to header to modify.
3780 * On output, pointer to the modified header, or NULL
3781 * to remove the header. This function frees the
3782 * original string if necessary.
3784 * Returns : JB_ERR_OK on success, or
3785 * JB_ERR_MEMORY on out-of-memory error.
3787 *********************************************************************/
3788 static jb_err server_set_cookie(struct client_state *csp, char **header)
3795 #ifdef FEATURE_COOKIE_JAR
3796 if (csp->config->jar)
3799 * Write timestamp into outbuf.
3801 * Complex because not all OSs have tm_gmtoff or
3802 * the %z field in strftime()
3804 char tempbuf[ BUFFER_SIZE ];
3806 #ifdef HAVE_LOCALTIME_R
3807 tm_now = *localtime_r(&now, &tm_now);
3808 #elif FEATURE_PTHREAD
3809 pthread_mutex_lock(&localtime_mutex);
3810 tm_now = *localtime (&now);
3811 pthread_mutex_unlock(&localtime_mutex);
3813 tm_now = *localtime (&now);
3815 strftime(tempbuf, BUFFER_SIZE-6, "%b %d %H:%M:%S ", &tm_now);
3817 /* strlen("set-cookie: ") = 12 */
3818 fprintf(csp->config->jar, "%s %s\t%s\n", tempbuf, csp->http->host, *header + 12);
3820 #endif /* def FEATURE_COOKIE_JAR */
3822 if ((csp->action->flags & ACTION_NO_COOKIE_SET) != 0)
3824 log_error(LOG_LEVEL_HEADER, "Crunching incoming cookie: %s", *header);
3827 else if ((csp->action->flags & ACTION_NO_COOKIE_KEEP) != 0)
3829 /* Flag whether or not to log a message */
3832 /* A variable to store the tag we're working on */
3835 /* Skip "Set-Cookie:" (11 characters) in header */
3836 cur_tag = *header + 11;
3838 /* skip whitespace between "Set-Cookie:" and value */
3839 while (*cur_tag && ijb_isspace(*cur_tag))
3844 /* Loop through each tag in the cookie */
3848 char *next_tag = strchr(cur_tag, ';');
3849 if (next_tag != NULL)
3851 /* Skip the ';' character itself */
3854 /* skip whitespace ";" and start of tag */
3855 while (*next_tag && ijb_isspace(*next_tag))
3862 /* "Next tag" is the end of the string */
3863 next_tag = cur_tag + strlen(cur_tag);
3867 * Check the expiration date to see
3868 * if the cookie is still valid, if yes,
3869 * rewrite it to a session cookie.
3871 if ((strncmpic(cur_tag, "expires=", 8) == 0) && *(cur_tag + 8))
3873 char *expiration_date = cur_tag + 8; /* Skip "[Ee]xpires=" */
3875 /* Did we detect the date properly? */
3876 if (JB_ERR_OK != parse_header_time(expiration_date, &cookie_time))
3879 * Nope, treat it as if it was still valid.
3881 * XXX: Should we remove the whole cookie instead?
3883 log_error(LOG_LEVEL_ERROR,
3884 "Can't parse \'%s\', send by %s. Unsupported time format?", cur_tag, csp->http->url);
3885 memmove(cur_tag, next_tag, strlen(next_tag) + 1);
3891 * Yes. Check if the cookie is still valid.
3893 * If the cookie is already expired it's probably
3894 * a delete cookie and even if it isn't, the browser
3895 * will discard it anyway.
3899 * XXX: timegm() isn't available on some AmigaOS
3900 * versions and our replacement doesn't work.
3902 * Our options are to either:
3904 * - disable session-cookies-only completely if timegm
3907 * - to simply remove all expired tags, like it has
3908 * been done until Privoxy 3.0.6 and to live with
3909 * the consequence that it can cause login/logout
3910 * problems on servers that don't validate their
3911 * input properly, or
3913 * - to replace it with mktime in which
3914 * case there is a slight chance of valid cookies
3915 * passing as already expired.
3917 * This is the way it's currently done and it's not
3918 * as bad as it sounds. If the missing GMT offset is
3919 * enough to change the result of the expiration check
3920 * the cookie will be only valid for a few hours
3921 * anyway, which in many cases will be shorter
3922 * than a browser session.
3924 if (cookie_time - now < 0)
3926 log_error(LOG_LEVEL_HEADER,
3927 "Cookie \'%s\' is already expired and can pass unmodified.", *header);
3928 /* Just in case some clown sets more then one expiration date */
3934 * Still valid, delete expiration date by copying
3935 * the rest of the string over it.
3937 * (Note that we cannot just use "strcpy(cur_tag, next_tag)",
3938 * since the behaviour of strcpy is undefined for overlapping
3941 memmove(cur_tag, next_tag, strlen(next_tag) + 1);
3943 /* That changed the header, need to issue a log message */
3947 * Note that the next tag has now been moved to *cur_tag,
3948 * so we do not need to update the cur_tag pointer.
3956 /* Move on to next cookie tag */
3963 assert(NULL != *header);
3964 log_error(LOG_LEVEL_HEADER, "Cookie rewritten to a temporary one: %s",
3973 #ifdef FEATURE_FORCE_LOAD
3974 /*********************************************************************
3976 * Function : strclean
3978 * Description : In-Situ-Eliminate all occurances of substring in
3982 * 1 : string = string to clean
3983 * 2 : substring = substring to eliminate
3985 * Returns : Number of eliminations
3987 *********************************************************************/
3988 int strclean(const char *string, const char *substring)
3994 len = strlen(substring);
3996 while((pos = strstr(string, substring)) != NULL)
4003 while (*p++ != '\0');
4010 #endif /* def FEATURE_FORCE_LOAD */
4013 /*********************************************************************
4015 * Function : parse_header_time
4017 * Description : Parses time formats used in HTTP header strings
4018 * to get the numerical respresentation.
4021 * 1 : header_time = HTTP header time as string.
4022 * 2 : result = storage for header_time in seconds
4024 * Returns : JB_ERR_OK if the time format was recognized, or
4025 * JB_ERR_PARSE otherwise.
4027 *********************************************************************/
4028 static jb_err parse_header_time(const char *header_time, time_t *result)
4033 * Zero out gmt to prevent time zone offsets.
4035 * While this is only necessary on some platforms
4036 * (mingw32 for example), I don't know how to
4037 * detect these automatically and doing it everywhere
4040 memset(&gmt, 0, sizeof(gmt));
4042 /* Tue, 02 Jun 2037 20:00:00 */
4043 if ((NULL == strptime(header_time, "%a, %d %b %Y %H:%M:%S", &gmt))
4044 /* Tue, 02-Jun-2037 20:00:00 */
4045 && (NULL == strptime(header_time, "%a, %d-%b-%Y %H:%M:%S", &gmt))
4046 /* Tue, 02-Jun-37 20:00:00 */
4047 && (NULL == strptime(header_time, "%a, %d-%b-%y %H:%M:%S", &gmt))
4048 /* Tuesday, 02-Jun-2037 20:00:00 */
4049 && (NULL == strptime(header_time, "%A, %d-%b-%Y %H:%M:%S", &gmt))
4050 /* Tuesday Jun 02 20:00:00 2037 */
4051 && (NULL == strptime(header_time, "%A %b %d %H:%M:%S %Y", &gmt)))
4053 return JB_ERR_PARSE;
4056 *result = timegm(&gmt);
4063 /*********************************************************************
4065 * Function : get_destination_from_headers
4067 * Description : Parse the "Host:" header to get the request's destination.
4068 * Only needed if the client's request was forcefully
4069 * redirected into Privoxy.
4071 * Code mainly copied from client_host() which is currently
4072 * run too late for this purpose.
4075 * 1 : headers = List of headers (one of them hopefully being
4076 * the "Host:" header)
4077 * 2 : http = storage for the result (host, port and hostport).
4079 * Returns : JB_ERR_MEMORY in case of memory problems,
4080 * JB_ERR_PARSE if the host header couldn't be found,
4081 * JB_ERR_OK otherwise.
4083 *********************************************************************/
4084 jb_err get_destination_from_headers(const struct list *headers, struct http_request *http)
4090 host = get_header_value(headers, "Host:");
4094 log_error(LOG_LEVEL_ERROR, "No \"Host:\" header found.");
4095 return JB_ERR_PARSE;
4098 if (NULL == (p = strdup((host))))
4100 log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
4101 return JB_ERR_MEMORY;
4104 if (NULL == (q = strdup(p)))
4107 log_error(LOG_LEVEL_ERROR, "Out of memory while parsing \"Host:\" header");
4108 return JB_ERR_MEMORY;
4111 freez(http->hostport);
4115 q = strchr(http->host, ':');
4118 /* Terminate hostname and evaluate port string */
4120 http->port = atoi(q);
4124 http->port = http->ssl ? 443 : 80;
4127 /* Rebuild request URL */
4129 http->url = strdup(http->ssl ? "https://" : "http://");
4130 string_append(&http->url, http->hostport);
4131 string_append(&http->url, http->path);
4132 if (http->url == NULL)
4134 return JB_ERR_MEMORY;
4137 log_error(LOG_LEVEL_HEADER, "Destination extracted from \"Host:\" header. New request URL: %s",
4145 /*********************************************************************
4147 * Function : create_forged_referrer
4149 * Description : Helper for client_referrer to forge a referer as
4150 * 'http://[hostname:port/' to fool stupid
4151 * checks for in-site links
4154 * 1 : header = Pointer to header pointer
4155 * 2 : hostport = Host and optionally port as string
4157 * Returns : JB_ERR_OK in case of success, or
4158 * JB_ERR_MEMORY in case of memory problems.
4160 *********************************************************************/
4161 static jb_err create_forged_referrer(char **header, const char *hostport)
4163 assert(NULL == *header);
4165 *header = strdup("Referer: http://");
4166 string_append(header, hostport);
4167 string_append(header, "/");
4169 if (NULL == *header)
4171 return JB_ERR_MEMORY;
4174 log_error(LOG_LEVEL_HEADER, "Referer forged to: %s", *header);
4181 /*********************************************************************
4183 * Function : create_fake_referrer
4185 * Description : Helper for client_referrer to create a fake referrer
4186 * based on a string supplied by the user.
4189 * 1 : header = Pointer to header pointer
4190 * 2 : hosthost = Referrer to fake
4192 * Returns : JB_ERR_OK in case of success, or
4193 * JB_ERR_MEMORY in case of memory problems.
4195 *********************************************************************/
4196 static jb_err create_fake_referrer(char **header, const char *fake_referrer)
4198 assert(NULL == *header);
4200 if ((0 != strncmpic(fake_referrer, "http://", 7)) && (0 != strncmpic(fake_referrer, "https://", 8)))
4202 log_error(LOG_LEVEL_HEADER,
4203 "Parameter: +hide-referrer{%s} is a bad idea, but I don't care.", fake_referrer);
4205 *header = strdup("Referer: ");
4206 string_append(header, fake_referrer);
4208 if (NULL == *header)
4210 return JB_ERR_MEMORY;
4213 log_error(LOG_LEVEL_HEADER, "Referer replaced with: %s", *header);
4220 /*********************************************************************
4222 * Function : handle_conditional_hide_referrer_parameter
4224 * Description : Helper for client_referrer to crunch or forge
4225 * the referrer header if the host has changed.
4228 * 1 : header = Pointer to header pointer
4229 * 2 : host = The target host (may include the port)
4230 * 3 : parameter_conditional_block = Boolean to signal
4231 * if we're in conditional-block mode. If not set,
4232 * we're in conditional-forge mode.
4234 * Returns : JB_ERR_OK in case of success, or
4235 * JB_ERR_MEMORY in case of memory problems.
4237 *********************************************************************/
4238 static jb_err handle_conditional_hide_referrer_parameter(char **header,
4239 const char *host, const int parameter_conditional_block)
4241 char *referer = strdup(*header);
4242 const size_t hostlenght = strlen(host);
4244 if (NULL == referer)
4247 return JB_ERR_MEMORY;
4250 /* referer begins with 'Referer: http[s]://' */
4251 if (hostlenght < (strlen(referer)-17))
4254 * Shorten referer to make sure the referer is blocked
4255 * if www.example.org/www.example.com-shall-see-the-referer/
4256 * links to www.example.com/
4258 referer[hostlenght+17] = '\0';
4260 if (NULL == strstr(referer, host))
4262 /* Host has changed */
4263 if (parameter_conditional_block)
4265 log_error(LOG_LEVEL_HEADER, "New host is: %s. Crunching %s!", host, *header);
4272 return create_forged_referrer(header, host);