-const char filters_rcs[] = "$Id: filters.c,v 1.96 2007/10/19 16:53:28 fabiankeil Exp $";
/*********************************************************************
*
* File : $Source: /cvsroot/ijbswa/current/filters.c,v $
*
* Purpose : Declares functions to parse/crunch headers and pages.
- * Functions declared include:
- * `acl_addr', `add_stats', `block_acl', `block_imageurl',
- * `block_url', `url_actions', `domain_split',
- * `filter_popups', `forward_url', 'redirect_url',
- * `ij_untrusted_url', `intercept_url', `pcrs_filter_respose',
- * `ijb_send_banner', `trust_url', `gif_deanimate_response',
- * `jpeg_inspect_response', `execute_single_pcrs_command',
- * `rewrite_url', `get_last_url'
- *
- * Copyright : Written by and Copyright (C) 2001, 2004-2007 the SourceForge
- * Privoxy team. http://www.privoxy.org/
+ *
+ * Copyright : Written by and Copyright (C) 2001-2020 the
+ * Privoxy team. https://www.privoxy.org/
*
* Based on the Internet Junkbuster originally written
* by and Copyright (C) 1997 Anonymous Coders and
* or write to the Free Software Foundation, Inc., 59
* Temple Place - Suite 330, Boston, MA 02111-1307, USA.
*
- * Revisions :
- * $Log: filters.c,v $
- * Revision 1.96 2007/10/19 16:53:28 fabiankeil
- * Add helper function to check if any content filters are enabled.
- *
- * Revision 1.95 2007/10/17 19:31:20 fabiankeil
- * Omitting the zero chunk that ends the chunk transfer encoding seems
- * to be the new black. Log the problem and continue filtering anyway.
- *
- * Revision 1.94 2007/09/29 13:20:20 fabiankeil
- * Remove two redundant and one useless log messages.
- *
- * Revision 1.93 2007/09/29 10:21:16 fabiankeil
- * - Move get_filter_function() from jcc.c to filters.c
- * so the filter functions can be static.
- * - Don't bother filtering body-less responses.
- *
- * Revision 1.92 2007/09/28 16:38:55 fabiankeil
- * - Execute content filters through execute_content_filter().
- * - Add prepare_for_filtering() so filter functions don't have to
- * care about de-chunking and decompression. As a side effect this enables
- * decompression for gif_deanimate_response() and jpeg_inspect_response().
- * - Change remove_chunked_transfer_coding()'s return type to jb_err.
- * Some clowns feel like chunking empty responses in which case
- * (size == 0) is valid but previously would be interpreted as error.
- *
- * Revision 1.91 2007/09/02 15:31:20 fabiankeil
- * Move match_portlist() from filter.c to urlmatch.c.
- * It's used for url matching, not for filtering.
- *
- * Revision 1.90 2007/09/02 12:44:17 fabiankeil
- * Remove newline at the end of a log_error() message.
- *
- * Revision 1.89 2007/08/05 13:42:23 fabiankeil
- * #1763173 from Stefan Huehner: declare some more functions static.
- *
- * Revision 1.88 2007/06/01 16:41:11 fabiankeil
- * Add forward-override{} to change the forwarding settings through
- * action sections. This is mainly interesting to forward different
- * clients differently (for example based on User-Agent or request
- * origin).
- *
- * Revision 1.87 2007/04/30 15:53:10 fabiankeil
- * Make sure filters with dynamic jobs actually use them.
- *
- * Revision 1.86 2007/04/30 15:03:28 fabiankeil
- * - Introduce dynamic pcrs jobs that can resolve variables.
- * - Don't run redirect functions more than once,
- * unless they are activated more than once.
- *
- * Revision 1.85 2007/03/21 12:24:47 fabiankeil
- * - Log the content size after decompression in decompress_iob()
- * instead of pcrs_filter_response().
- *
- * Revision 1.84 2007/03/20 15:16:34 fabiankeil
- * Use dedicated header filter actions instead of abusing "filter".
- * Replace "filter-client-headers" and "filter-client-headers"
- * with "server-header-filter" and "client-header-filter".
- *
- * Revision 1.83 2007/03/17 15:20:05 fabiankeil
- * New config option: enforce-blocks.
- *
- * Revision 1.82 2007/03/13 11:28:43 fabiankeil
- * - Fix port handling in acl_addr() and use a temporary acl spec
- * copy so error messages don't contain a truncated version.
- * - Log size of iob before and after decompression.
- *
- * Revision 1.81 2007/03/05 14:40:53 fabiankeil
- * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
- * - Hide the "Go there anyway" link for blocked CONNECT
- * requests where going there anyway doesn't work anyway.
- *
- * Revision 1.80 2007/02/07 10:55:20 fabiankeil
- * - Save the reason for generating http_responses.
- * - Block (+block) with status code 403 instead of 404.
- * - Use a different kludge to remember a failed decompression.
- *
- * Revision 1.79 2007/01/31 16:21:38 fabiankeil
- * Search for Max-Forwards headers case-insensitive,
- * don't generate the "501 unsupported" message for invalid
- * Max-Forwards values and don't increase negative ones.
- *
- * Revision 1.78 2007/01/28 13:41:18 fabiankeil
- * - Add HEAD support to finish_http_response.
- * - Add error favicon to internal HTML error messages.
- *
- * Revision 1.77 2007/01/12 15:36:44 fabiankeil
- * Mark *csp as immutable for is_untrusted_url()
- * and is_imageurl(). Closes FR 1237736.
- *
- * Revision 1.76 2007/01/01 19:36:37 fabiankeil
- * Integrate a modified version of Wil Mahan's
- * zlib patch (PR #895531).
- *
- * Revision 1.75 2006/12/29 18:30:46 fabiankeil
- * Fixed gcc43 conversion warnings,
- * changed sprintf calls to snprintf.
- *
- * Revision 1.74 2006/12/24 17:37:38 fabiankeil
- * Adjust comment in pcrs_filter_response()
- * to recent pcrs changes. Hohoho.
- *
- * Revision 1.73 2006/12/23 16:01:02 fabiankeil
- * Don't crash if pcre returns an error code
- * that pcrs didn't expect. Fixes BR 1621173.
- *
- * Revision 1.72 2006/12/22 18:52:53 fabiankeil
- * Modified is_untrusted_url to complain in case of
- * write errors and to give a reason when adding new
- * entries to the trustfile. Closes FR 1097611.
- *
- * Revision 1.71 2006/12/22 14:24:52 fabiankeil
- * Skip empty filter files in pcrs_filter_response,
- * but don't ignore the ones that come afterwards.
- * Fixes parts of BR 1619208.
- *
- * Revision 1.70 2006/12/09 13:33:15 fabiankeil
- * Added some sanity checks for get_last_url().
- * Fixed possible segfault caused by my last commit.
- *
- * Revision 1.69 2006/12/08 12:39:13 fabiankeil
- * Let get_last_url() catch https URLs as well.
- *
- * Revision 1.68 2006/12/05 14:45:48 fabiankeil
- * Make sure get_last_url() behaves like advertised
- * and fast-redirects{} can be combined with redirect{}.
- *
- * Revision 1.67 2006/11/28 15:19:43 fabiankeil
- * Implemented +redirect{s@foo@bar@} to generate
- * a redirect based on a rewritten version of the
- * original URL.
- *
- * Revision 1.66 2006/09/23 13:26:38 roro
- * Replace TABs by spaces in source code.
- *
- * Revision 1.65 2006/09/21 12:54:43 fabiankeil
- * Fix +redirect{}. Didn't work with -fast-redirects.
- *
- * Revision 1.64 2006/08/31 10:55:49 fabiankeil
- * Block requests for untrusted URLs with status
- * code 403 instead of 200.
- *
- * Revision 1.63 2006/08/31 10:11:28 fabiankeil
- * Don't free p which is still in use and will be later
- * freed by free_map(). Don't claim the referrer is unknown
- * when the client didn't set one.
- *
- * Revision 1.62 2006/08/14 00:27:47 david__schmidt
- * Feature request 595948: Re-Filter logging in single line
- *
- * Revision 1.61 2006/08/03 02:46:41 david__schmidt
- * Incorporate Fabian Keil's patch work:\rhttp://www.fabiankeil.de/sourcecode/privoxy/
- *
- * Revision 1.60 2006/07/18 14:48:46 david__schmidt
- * Reorganizing the repository: swapping out what was HEAD (the old 3.1 branch)
- * with what was really the latest development (the v_3_0_branch branch)
- *
- * Revision 1.58.2.9 2006/01/29 23:10:56 david__schmidt
- * Multiple filter file support
- *
- * Revision 1.58.2.8 2005/05/07 21:50:55 david__schmidt
- * A few memory leaks plugged (mostly on error paths)
- *
- * Revision 1.58.2.7 2004/10/03 12:53:32 david__schmidt
- * Add the ability to check jpeg images for invalid
- * lengths of comment blocks. Defensive strategy
- * against the exploit:
- * Microsoft Security Bulletin MS04-028
- * Buffer Overrun in JPEG Processing (GDI+) Could
- * Allow Code Execution (833987)
- * Enabled with +inspect-jpegs in actions files.
- *
- * Revision 1.58.2.6 2003/12/06 22:18:27 gliptak
- * Correcting compile problem with FEATURE_IMAGE_BLOCKING
- *
- * Revision 1.58.2.5 2003/11/11 13:10:31 oes
- * Fixed bug #839859: "See why" link URL now gets url-encoded.
- *
- * Revision 1.58.2.4 2003/02/28 12:52:45 oes
- * Fixed a typo
- *
- * Revision 1.58.2.3 2002/09/25 14:51:51 oes
- * Added basic support for OPTIONS and TRACE HTTP methods:
- * New function direct_response which handles OPTIONS and
- * TRACE requests whose Max-Forwards header field is zero.
- *
- * Revision 1.58.2.2 2002/08/01 17:18:28 oes
- * Fixed BR 537651 / SR 579724 (MSIE image detect improper for IE/Mac)
- *
- * Revision 1.58.2.1 2002/07/26 15:18:53 oes
- * - Bugfix: Executing a filters without jobs no longer results in
- * turing off *all* filters.
- * - Security fix: Malicious web servers can't cause a seg fault
- * through bogus chunk sizes anymore
- *
- * Revision 1.58 2002/04/24 02:11:17 oes
- * Jon's multiple AF patch: url_actions now evaluates rules
- * from all AFs.
- *
- * Revision 1.57 2002/04/08 20:38:34 swa
- * fixed JB spelling
- *
- * Revision 1.56 2002/04/05 15:51:24 oes
- * - bugfix: error-pages now get correct request protocol
- * - fix for invalid HTML in trust info
- *
- * Revision 1.55 2002/04/02 16:13:51 oes
- * Fix: No "Go there anyway" for SSL
- *
- * Revision 1.54 2002/04/02 14:55:56 oes
- * Bugfix: is_untrusted_url() now depends on FEATURE_TRUST, not FEATURE_COOKIE_JAR
- *
- * Revision 1.53 2002/03/26 22:29:54 swa
- * we have a new homepage!
- *
- * Revision 1.52 2002/03/24 16:35:57 jongfoster
- * Removing logo
- *
- * Revision 1.51 2002/03/24 15:23:33 jongfoster
- * Name changes
- *
- * Revision 1.50 2002/03/24 13:25:43 swa
- * name change related issues
- *
- * Revision 1.49 2002/03/16 20:29:14 oes
- * Cosmetics
- *
- * Revision 1.48 2002/03/13 20:25:34 oes
- * Better logging for content filters
- *
- * Revision 1.47 2002/03/13 00:30:52 jongfoster
- * Killing warnings
- * Added option of always sending redirect for imageblock,
- * currently disabled with #if 0.
- *
- * Revision 1.46 2002/03/12 01:42:49 oes
- * Introduced modular filters
- *
- * Revision 1.45 2002/03/08 16:47:50 oes
- * Added choice beween GIF and PNG built-in images
- *
- * Revision 1.44 2002/03/07 03:49:31 oes
- * - Fixed compiler warnings etc
- * - Changed built-in images from GIF to PNG
- * (with regard to Unisys patent issue)
- * - Added a 4x4 pattern PNG which is less intrusive
- * than the logo but also clearly marks the deleted banners
- *
- * Revision 1.43 2002/01/22 23:51:59 jongfoster
- * Replacing strsav() with the safer string_append().
- *
- * Adding missing html_encode() to error message generators. Where encoded
- * and unencoded versions of a string were provided, removing the unencoded
- * one.
- *
- * Revision 1.42 2002/01/17 21:00:32 jongfoster
- * Moving all our URL and URL pattern parsing code to urlmatch.c.
- *
- * Using a single, simple url_match(pattern,url) function - rather than
- * the 3-line match routine which was repeated all over the place.
- *
- * Renaming free_url to free_url_spec, since it frees a struct url_spec.
- *
- * Using parse_http_url() to parse URLs without faking a HTTP
- * request line for parse_http_request().
- *
- * Revision 1.41 2001/11/13 00:14:07 jongfoster
- * Fixing stupid bug now I've figured out what || means.
- * (It always returns 0 or 1, not one of it's paramaters.)
- *
- * Revision 1.40 2001/10/26 17:37:55 oes
- * - Re-enabled Netscape 200/404 bug workaround in block_url():
- * - Removed OS/2 special case
- * - Made block_url() independant from sed() having been run
- * - Made trust_url independant from sed() having been run
- * - Made is_imageurl independant from sed() having been run.
- * It now checks User-Agent: and Accept: by itself.
- *
- *
- * Revision 1.39 2001/10/25 03:40:48 david__schmidt
- * Change in porting tactics: OS/2's EMX porting layer doesn't allow multiple
- * threads to call select() simultaneously. So, it's time to do a real, live,
- * native OS/2 port. See defines for __EMX__ (the porting layer) vs. __OS2__
- * (native). Both versions will work, but using __OS2__ offers multi-threading.
- *
- * Revision 1.38 2001/10/23 21:32:33 jongfoster
- * Adding error-checking to selected functions
- *
- * Revision 1.37 2001/10/22 15:33:56 david__schmidt
- * Special-cased OS/2 out of the Netscape-abort-on-404-in-js problem in
- * filters.c. Added a FIXME in front of the offending code. I'll gladly
- * put in a better/more robust fix for all parties if one is presented...
- * It seems that just returning 200 instead of 404 would pretty much fix
- * it for everyone, but I don't know all the history of the problem.
- *
- * Revision 1.36 2001/10/10 16:44:16 oes
- * Added match_portlist function
- *
- * Revision 1.35 2001/10/07 15:41:23 oes
- * Replaced 6 boolean members of csp with one bitmap (csp->flags)
- *
- * New function remove_chunked_transfer_coding that strips chunked
- * transfer coding to plain and is called by pcrs_filter_response
- * and gif_deanimate_response if neccessary
- *
- * Improved handling of zero-change re_filter runs
- *
- * pcrs_filter_response and gif_deanimate_response now remove
- * chunked transfer codeing before processing the body.
- *
- * Revision 1.34 2001/09/20 15:49:36 steudten
- *
- * Fix BUG: Change int size to size_t size in pcrs_filter_response().
- * See cgi.c fill_template().
- *
- * Revision 1.33 2001/09/16 17:05:14 jongfoster
- * Removing unused #include showarg.h
- *
- * Revision 1.32 2001/09/16 13:21:27 jongfoster
- * Changes to use new list functions.
- *
- * Revision 1.31 2001/09/16 11:38:02 jongfoster
- * Splitting fill_template() into 2 functions:
- * template_load() loads the file
- * template_fill() performs the PCRS regexps.
- * This is because the CGI edit interface has a "table row"
- * template which is used many times in the page - this
- * change means it's only loaded from disk once.
- *
- * Revision 1.30 2001/09/16 11:00:10 jongfoster
- * New function alloc_http_response, for symmetry with free_http_response
- *
- * Revision 1.29 2001/09/13 23:32:40 jongfoster
- * Moving image data to cgi.c rather than cgi.h
- * Fixing a GPF under Win32 (and any other OS that protects global
- * constants from being written to).
- *
- * Revision 1.28 2001/09/10 10:18:51 oes
- * Silenced compiler warnings
- *
- * Revision 1.27 2001/08/05 16:06:20 jongfoster
- * Modifiying "struct map" so that there are now separate header and
- * "map_entry" structures. This means that functions which modify a
- * map no longer need to return a pointer to the modified map.
- * Also, it no longer reverses the order of the entries (which may be
- * important with some advanced template substitutions).
- *
- * Revision 1.26 2001/07/30 22:08:36 jongfoster
- * Tidying up #defines:
- * - All feature #defines are now of the form FEATURE_xxx
- * - Permanently turned off WIN_GUI_EDIT
- * - Permanently turned on WEBDAV and SPLIT_PROXY_ARGS
- *
- * Revision 1.25 2001/07/26 10:09:46 oes
- * Made browser detection a little less naive
- *
- * Revision 1.24 2001/07/25 17:22:51 oes
- * Added workaround for Netscape bug that prevents display of page when loading a component fails.
- *
- * Revision 1.23 2001/07/23 13:40:12 oes
- * Fixed bug that caused document body to be dropped when pcrs joblist was empty.
- *
- * Revision 1.22 2001/07/18 12:29:34 oes
- * - Made gif_deanimate_response respect
- * csp->action->string[ACTION_STRING_DEANIMATE]
- * - Logging cosmetics
- *
- * Revision 1.21 2001/07/13 13:59:53 oes
- * - Introduced gif_deanimate_response which shares the
- * generic content modification interface of pcrs_filter_response
- * and acts as a wrapper to deanimate.c:gif_deanimate()
- * - Renamed re_process_buffer to pcrs_filter_response
- * - pcrs_filter_response now returns NULL on failiure
- * - Removed all #ifdef PCRS
- *
- * Revision 1.20 2001/07/01 17:01:04 oes
- * Added comments and missing return statement in is_untrusted_url()
- *
- * Revision 1.19 2001/06/29 21:45:41 oes
- * Indentation, CRLF->LF, Tab-> Space
- *
- * Revision 1.18 2001/06/29 13:27:38 oes
- * - Cleaned up, renamed and reorderd functions
- * and improved comments
- *
- * - block_url:
- * - Ported to CGI platform. Now delivers
- * http_response or NULL
- * - Unified HTML and GIF generation (moved image detection
- * and GIF generation here from jcc.c:chat())
- * - Fixed HTTP status to:
- * - 403 (Forbidden) for the "blocked" HTML message
- * - 200 (OK) for GIF answers
- * - 302 (Redirect) for redirect to GIF
- *
- * - trust_url:
- * - Ported to CGI platform. Now delivers
- * http_response or NULL
- * - Separated detection of untrusted URL into
- * (bool)is_untrusted_url
- * - Added enforcement of untrusted requests
- *
- * - Moved redirect_url() from cgi.c to here
- * and ported it to the CGI platform
- *
- * - Removed logentry from cancelled commit
- *
- * Revision 1.17 2001/06/09 10:55:28 jongfoster
- * Changing BUFSIZ ==> BUFFER_SIZE
- *
- * Revision 1.16 2001/06/07 23:10:26 jongfoster
- * Allowing unanchored domain patterns to back off and retry
- * if they partially match. Optimized right-anchored patterns.
- * Moving ACL and forward files into config file.
- * Replacing struct gateway with struct forward_spec
- *
- * Revision 1.15 2001/06/03 19:12:00 oes
- * extracted-CGI relevant stuff
- *
- * Revision 1.14 2001/06/01 10:30:55 oes
- * Added optional left-anchoring to domaincmp
- *
- * Revision 1.13 2001/05/31 21:21:30 jongfoster
- * Permissionsfile / actions file changes:
- * - Changed "permission" to "action" throughout
- * - changes to file format to allow string parameters
- * - Moved helper functions to actions.c
- *
- * Revision 1.12 2001/05/31 17:35:20 oes
- *
- * - Enhanced domain part globbing with infix and prefix asterisk
- * matching and optional unanchored operation
- *
- * Revision 1.11 2001/05/29 11:53:23 oes
- * "See why" link added to "blocked" page
- *
- * Revision 1.10 2001/05/29 09:50:24 jongfoster
- * Unified blocklist/imagelist/permissionslist.
- * File format is still under discussion, but the internal changes
- * are (mostly) done.
- *
- * Also modified interceptor behaviour:
- * - We now intercept all URLs beginning with one of the following
- * prefixes (and *only* these prefixes):
- * * http://i.j.b/
- * * http://ijbswa.sf.net/config/
- * * http://ijbswa.sourceforge.net/config/
- * - New interceptors "home page" - go to http://i.j.b/ to see it.
- * - Internal changes so that intercepted and fast redirect pages
- * are not replaced with an image.
- * - Interceptors now have the option to send a binary page direct
- * to the client. (i.e. ijb-send-banner uses this)
- * - Implemented show-url-info interceptor. (Which is why I needed
- * the above interceptors changes - a typical URL is
- * "http://i.j.b/show-url-info?url=www.somesite.com/banner.gif".
- * The previous mechanism would not have intercepted that, and
- * if it had been intercepted then it then it would have replaced
- * it with an image.)
- *
- * Revision 1.9 2001/05/27 22:17:04 oes
- *
- * - re_process_buffer no longer writes the modified buffer
- * to the client, which was very ugly. It now returns the
- * buffer, which it is then written by chat.
- *
- * - content_length now adjusts the Content-Length: header
- * for modified documents rather than crunch()ing it.
- * (Length info in csp->content_length, which is 0 for
- * unmodified documents)
- *
- * - For this to work, sed() is called twice when filtering.
- *
- * Revision 1.8 2001/05/26 17:13:28 jongfoster
- * Filled in a function comment.
- *
- * Revision 1.7 2001/05/26 15:26:15 jongfoster
- * ACL feature now provides more security by immediately dropping
- * connections from untrusted hosts.
- *
- * Revision 1.6 2001/05/26 00:28:36 jongfoster
- * Automatic reloading of config file.
- * Removed obsolete SIGHUP support (Unix) and Reload menu option (Win32).
- * Most of the global variables have been moved to a new
- * struct configuration_spec, accessed through csp->config->globalname
- * Most of the globals remaining are used by the Win32 GUI.
- *
- * Revision 1.5 2001/05/25 22:34:30 jongfoster
- * Hard tabs->Spaces
- *
- * Revision 1.4 2001/05/22 18:46:04 oes
- *
- * - Enabled filtering banners by size rather than URL
- * by adding patterns that replace all standard banner
- * sizes with the "Junkbuster" gif to the re_filterfile
- *
- * - Enabled filtering WebBugs by providing a pattern
- * which kills all 1x1 images
- *
- * - Added support for PCRE_UNGREEDY behaviour to pcrs,
- * which is selected by the (nonstandard and therefore
- * capital) letter 'U' in the option string.
- * It causes the quantifiers to be ungreedy by default.
- * Appending a ? turns back to greedy (!).
- *
- * - Added a new interceptor ijb-send-banner, which
- * sends back the "Junkbuster" gif. Without imagelist or
- * MSIE detection support, or if tinygif = 1, or the
- * URL isn't recognized as an imageurl, a lame HTML
- * explanation is sent instead.
- *
- * - Added new feature, which permits blocking remote
- * script redirects and firing back a local redirect
- * to the browser.
- * The feature is conditionally compiled, i.e. it
- * can be disabled with --disable-fast-redirects,
- * plus it must be activated by a "fast-redirects"
- * line in the config file, has its own log level
- * and of course wants to be displayed by show-proxy-args
- * Note: Boy, all the #ifdefs in 1001 locations and
- * all the fumbling with configure.in and acconfig.h
- * were *way* more work than the feature itself :-(
- *
- * - Because a generic redirect template was needed for
- * this, tinygif = 3 now uses the same.
- *
- * - Moved GIFs, and other static HTTP response templates
- * to project.h
- *
- * - Some minor fixes
- *
- * - Removed some >400 CRs again (Jon, you really worked
- * a lot! ;-)
- *
- * Revision 1.3 2001/05/20 16:44:47 jongfoster
- * Removing last hardcoded Junkbusters.com URLs.
- *
- * Revision 1.2 2001/05/20 01:21:20 jongfoster
- * Version 2.9.4 checkin.
- * - Merged popupfile and cookiefile, and added control over PCRS
- * filtering, in new "permissionsfile".
- * - Implemented LOG_LEVEL_FATAL, so that if there is a configuration
- * file error you now get a message box (in the Win32 GUI) rather
- * than the program exiting with no explanation.
- * - Made killpopup use the PCRS MIME-type checking and HTTP-header
- * skipping.
- * - Removed tabs from "config"
- * - Moved duplicated url parsing code in "loaders.c" to a new funcition.
- * - Bumped up version number.
- *
- * Revision 1.1.1.1 2001/05/15 13:58:52 oes
- * Initial import of version 2.9.3 source tree
- *
- *
*********************************************************************/
-\f
+
#include "config.h"
#include "miscutil.h"
#include "actions.h"
#include "cgi.h"
+#include "jcc.h"
#include "list.h"
#include "deanimate.h"
#include "urlmatch.h"
#include "loaders.h"
+#ifdef FEATURE_CLIENT_TAGS
+#include "client-tags.h"
+#endif
#ifdef _WIN32
#include "win32.h"
#endif
-const char filters_h_rcs[] = FILTERS_H_VERSION;
-
-/* Fix a problem with Solaris. There should be no effect on other
- * platforms.
- * Solaris's isspace() is a macro which uses it's argument directly
- * as an array index. Therefore we need to make sure that high-bit
- * characters generate +ve values, and ideally we also want to make
- * the argument match the declared parameter type of "int".
- */
-#define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
-
-static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size);
+typedef char *(*filter_function_ptr)();
+static filter_function_ptr get_filter_function(const struct client_state *csp);
static jb_err prepare_for_filtering(struct client_state *csp);
+static void apply_url_actions(struct current_action_spec *action,
+ struct http_request *http,
+#ifdef FEATURE_CLIENT_TAGS
+ const struct list *client_tags,
+#endif
+ struct url_actions *b);
#ifdef FEATURE_ACL
+#ifdef HAVE_RFC2553
+/*********************************************************************
+ *
+ * Function : sockaddr_storage_to_ip
+ *
+ * Description : Access internal structure of sockaddr_storage
+ *
+ * Parameters :
+ * 1 : addr = socket address
+ * 2 : ip = IP address as array of octets in network order
+ * (it points into addr)
+ * 3 : len = length of IP address in octets
+ * 4 : port = port number in network order;
+ *
+ * Returns : void
+ *
+ *********************************************************************/
+static void sockaddr_storage_to_ip(const struct sockaddr_storage *addr,
+ uint8_t **ip, unsigned int *len,
+ in_port_t **port)
+{
+ assert(NULL != addr);
+ assert(addr->ss_family == AF_INET || addr->ss_family == AF_INET6);
+
+ switch (addr->ss_family)
+ {
+ case AF_INET:
+ if (NULL != len)
+ {
+ *len = 4;
+ }
+ if (NULL != ip)
+ {
+ *ip = (uint8_t *)
+ &(((struct sockaddr_in *)addr)->sin_addr.s_addr);
+ }
+ if (NULL != port)
+ {
+ *port = &((struct sockaddr_in *)addr)->sin_port;
+ }
+ break;
+
+ case AF_INET6:
+ if (NULL != len)
+ {
+ *len = 16;
+ }
+ if (NULL != ip)
+ {
+ *ip = ((struct sockaddr_in6 *)addr)->sin6_addr.s6_addr;
+ }
+ if (NULL != port)
+ {
+ *port = &((struct sockaddr_in6 *)addr)->sin6_port;
+ }
+ break;
+
+ }
+}
+
+
+/*********************************************************************
+ *
+ * Function : match_sockaddr
+ *
+ * Description : Check whether address matches network (IP address and port)
+ *
+ * Parameters :
+ * 1 : network = socket address of subnework
+ * 2 : netmask = network mask as socket address
+ * 3 : address = checked socket address against given network
+ *
+ * Returns : 0 = doesn't match; 1 = does match
+ *
+ *********************************************************************/
+static int match_sockaddr(const struct sockaddr_storage *network,
+ const struct sockaddr_storage *netmask,
+ const struct sockaddr_storage *address)
+{
+ uint8_t *network_addr, *netmask_addr, *address_addr;
+ unsigned int addr_len;
+ in_port_t *network_port, *netmask_port, *address_port;
+ int i;
+
+ if (network->ss_family != netmask->ss_family)
+ {
+ /* This should never happen */
+ assert(network->ss_family == netmask->ss_family);
+ log_error(LOG_LEVEL_FATAL, "Network and netmask differ in family.");
+ }
+
+ sockaddr_storage_to_ip(network, &network_addr, &addr_len, &network_port);
+ sockaddr_storage_to_ip(netmask, &netmask_addr, NULL, &netmask_port);
+ sockaddr_storage_to_ip(address, &address_addr, NULL, &address_port);
+
+ /* Check for family */
+ if ((network->ss_family == AF_INET) && (address->ss_family == AF_INET6)
+ && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)address_addr))
+ {
+ /* Map AF_INET6 V4MAPPED address into AF_INET */
+ address_addr += 12;
+ addr_len = 4;
+ }
+ else if ((network->ss_family == AF_INET6) && (address->ss_family == AF_INET)
+ && IN6_IS_ADDR_V4MAPPED((struct in6_addr *)network_addr))
+ {
+ /* Map AF_INET6 V4MAPPED network into AF_INET */
+ network_addr += 12;
+ netmask_addr += 12;
+ addr_len = 4;
+ }
+
+ /* XXX: Port check is signaled in netmask */
+ if (*netmask_port && *network_port != *address_port)
+ {
+ return 0;
+ }
+
+ /* TODO: Optimize by checking by words instead of octets */
+ for (i = 0; (i < addr_len) && netmask_addr[i]; i++)
+ {
+ if ((network_addr[i] & netmask_addr[i]) !=
+ (address_addr[i] & netmask_addr[i]))
+ {
+ return 0;
+ }
+ }
+
+ return 1;
+}
+#endif /* def HAVE_RFC2553 */
+
+
/*********************************************************************
*
* Function : block_acl
* Returns : 0 = FALSE (don't block) and 1 = TRUE (do block)
*
*********************************************************************/
-int block_acl(struct access_control_addr *dst, struct client_state *csp)
+int block_acl(const struct access_control_addr *dst, const struct client_state *csp)
{
struct access_control_list *acl = csp->config->acl;
/* search the list */
while (acl != NULL)
{
- if ((csp->ip_addr_long & acl->src->mask) == acl->src->addr)
+ if (
+#ifdef HAVE_RFC2553
+ match_sockaddr(&acl->src->addr, &acl->src->mask, &csp->tcp_addr)
+#else
+ (csp->ip_addr_long & acl->src->mask) == acl->src->addr
+#endif
+ )
{
if (dst == NULL)
{
{
return(0);
}
+ else
+ {
+ return(1);
+ }
}
- else if ( ((dst->addr & acl->dst->mask) == acl->dst->addr)
- && ((dst->port == acl->dst->port) || (acl->dst->port == 0)))
+ else if (
+#ifdef HAVE_RFC2553
+ /*
+ * XXX: An undefined acl->dst is full of zeros and should be
+ * considered a wildcard address. sockaddr_storage_to_ip()
+ * fails on such destinations because of unknown sa_familly
+ * (glibc only?). However this test is not portable.
+ *
+ * So, we signal the acl->dst is wildcard in wildcard_dst.
+ */
+ acl->wildcard_dst ||
+ match_sockaddr(&acl->dst->addr, &acl->dst->mask, &dst->addr)
+#else
+ ((dst->addr & acl->dst->mask) == acl->dst->addr)
+ && ((dst->port == acl->dst->port) || (acl->dst->port == 0))
+#endif
+ )
{
if (acl->action == ACL_PERMIT)
{
int acl_addr(const char *aspec, struct access_control_addr *aca)
{
int i, masklength;
+#ifdef HAVE_RFC2553
+ struct addrinfo hints, *result;
+ uint8_t *mask_data;
+ in_port_t *mask_port;
+ unsigned int addr_len;
+#else
long port;
+#endif /* def HAVE_RFC2553 */
char *p;
char *acl_spec = NULL;
+#ifdef HAVE_RFC2553
+ /* XXX: Depend on ai_family */
+ masklength = 128;
+#else
masklength = 32;
port = 0;
+#endif
/*
* Use a temporary acl spec copy so we can log
* the unmodified original in case of parse errors.
*/
- acl_spec = strdup(aspec);
- if (acl_spec == NULL)
- {
- /* XXX: This will be logged as parse error. */
- return(-1);
- }
+ acl_spec = strdup_or_die(aspec);
if ((p = strchr(acl_spec, '/')) != NULL)
{
*p++ = '\0';
- if (ijb_isdigit(*p) == 0)
+ if (privoxy_isdigit(*p) == 0)
{
freez(acl_spec);
return(-1);
masklength = atoi(p);
}
- if ((masklength < 0) || (masklength > 32))
+ if ((masklength < 0) ||
+#ifdef HAVE_RFC2553
+ (masklength > 128)
+#else
+ (masklength > 32)
+#endif
+ )
{
freez(acl_spec);
return(-1);
}
- if ((p = strchr(acl_spec, ':')) != NULL)
+ if ((*acl_spec == '[') && (NULL != (p = strchr(acl_spec, ']'))))
+ {
+ *p = '\0';
+ memmove(acl_spec, acl_spec + 1, (size_t)(p - acl_spec));
+
+ if (*++p != ':')
+ {
+ p = NULL;
+ }
+ }
+ else
+ {
+ p = strchr(acl_spec, ':');
+ }
+ if (p != NULL)
+ {
+ assert(*p == ':');
+ *p = '\0';
+ p++;
+ }
+
+#ifdef HAVE_RFC2553
+ memset(&hints, 0, sizeof(struct addrinfo));
+ hints.ai_family = AF_UNSPEC;
+ hints.ai_socktype = SOCK_STREAM;
+
+ i = getaddrinfo(acl_spec, p, &hints, &result);
+
+ if (i != 0)
+ {
+ log_error(LOG_LEVEL_ERROR, "Can not resolve [%s]:%s: %s",
+ acl_spec, p, gai_strerror(i));
+ freez(acl_spec);
+ return(-1);
+ }
+ freez(acl_spec);
+
+ /* TODO: Allow multihomed hostnames */
+ memcpy(&(aca->addr), result->ai_addr, result->ai_addrlen);
+ freeaddrinfo(result);
+#else
+ if (p != NULL)
{
char *endptr;
- *p++ = '\0';
port = strtol(p, &endptr, 10);
if (port <= 0 || port > 65535 || *endptr != '\0')
/* XXX: This will be logged as parse error. */
return(-1);
}
+#endif /* def HAVE_RFC2553 */
/* build the netmask */
+#ifdef HAVE_RFC2553
+ /* Clip masklength according to current family. */
+ if ((aca->addr.ss_family == AF_INET) && (masklength > 32))
+ {
+ masklength = 32;
+ }
+
+ aca->mask.ss_family = aca->addr.ss_family;
+ sockaddr_storage_to_ip(&aca->mask, &mask_data, &addr_len, &mask_port);
+
+ if (p)
+ {
+ /* ACL contains a port number, check ports in the future. */
+ *mask_port = 1;
+ }
+
+ /*
+ * XXX: This could be optimized to operate on whole words instead
+ * of octets (128-bit CPU could do it in one iteration).
+ */
+ /*
+ * Octets after prefix can be omitted because of
+ * previous initialization to zeros.
+ */
+ for (i = 0; (i < addr_len) && masklength; i++)
+ {
+ if (masklength >= 8)
+ {
+ mask_data[i] = 0xFF;
+ masklength -= 8;
+ }
+ else
+ {
+ /*
+ * XXX: This assumes MSB of octet is on the left side.
+ * This should be true for all architectures or solved
+ * by the link layer.
+ */
+ mask_data[i] = (uint8_t)~((1 << (8 - masklength)) - 1);
+ masklength = 0;
+ }
+ }
+
+#else
aca->mask = 0;
for (i=1; i <= masklength ; i++)
{
- aca->mask |= (1 << (32 - i));
+ aca->mask |= (1U << (32 - i));
}
/* now mask off the host portion of the ip address
* (i.e. save on the network portion of the address).
*/
aca->addr = aca->addr & aca->mask;
+#endif /* def HAVE_RFC2553 */
return(0);
#endif /* def FEATURE_ACL */
+/*********************************************************************
+ *
+ * Function : connect_port_is_forbidden
+ *
+ * Description : Check to see if CONNECT requests to the destination
+ * port of this request are forbidden. The check is
+ * independent of the actual request method.
+ *
+ * Parameters :
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ *
+ * Returns : True if yes, false otherwise.
+ *
+ *********************************************************************/
+int connect_port_is_forbidden(const struct client_state *csp)
+{
+ return ((csp->action->flags & ACTION_LIMIT_CONNECT) &&
+ !match_portlist(csp->action->string[ACTION_STRING_LIMIT_CONNECT],
+ csp->http->port));
+}
+
+
/*********************************************************************
*
* Function : block_url
}
if (csp->action->flags & ACTION_REDIRECT)
{
- log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
+ log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
}
/*
* Else, prepare a response
/* determine HOW images should be blocked */
p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
- if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
+ if (csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
{
log_error(LOG_LEVEL_ERROR, "handle-as-empty-document overruled by handle-as-image.");
}
-#if 1 /* Two alternative strategies, use this one for now: */
/* and handle accordingly: */
if ((p == NULL) || (0 == strcmpic(p, "pattern")))
{
- rsp->status = strdup("403 Request blocked by Privoxy");
- if (rsp->status == NULL)
- {
- free_http_response(rsp);
- return cgi_error_memory();
- }
+ rsp->status = strdup_or_die("403 Request blocked by Privoxy");
rsp->body = bindup(image_pattern_data, image_pattern_length);
if (rsp->body == NULL)
{
return cgi_error_memory();
}
}
-
else if (0 == strcmpic(p, "blank"))
{
- rsp->status = strdup("403 Request blocked by Privoxy");
- if (rsp->status == NULL)
- {
- free_http_response(rsp);
- return cgi_error_memory();
- }
+ rsp->status = strdup_or_die("403 Request blocked by Privoxy");
rsp->body = bindup(image_blank_data, image_blank_length);
if (rsp->body == NULL)
{
return cgi_error_memory();
}
}
-
else
{
- rsp->status = strdup("302 Local Redirect from Privoxy");
- if (rsp->status == NULL)
- {
- free_http_response(rsp);
- return cgi_error_memory();
- }
+ rsp->status = strdup_or_die("302 Local Redirect from Privoxy");
if (enlist_unique_header(rsp->headers, "Location", p))
{
}
}
-#else /* Following code is disabled for now */
-
- /* and handle accordingly: */
- if ((p == NULL) || (0 == strcmpic(p, "pattern")))
- {
- p = CGI_PREFIX "send-banner?type=pattern";
- }
- else if (0 == strcmpic(p, "blank"))
- {
- p = CGI_PREFIX "send-banner?type=blank";
- }
- rsp->status = strdup("302 Local Redirect from Privoxy");
- if (rsp->status == NULL)
- {
- free_http_response(rsp);
- return cgi_error_memory();
- }
-
- if (enlist_unique_header(rsp->headers, "Location", p))
- {
- free_http_response(rsp);
- return cgi_error_memory();
- }
-#endif /* Preceeding code is disabled for now */
}
- else if(csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
+ else
+#endif /* def FEATURE_IMAGE_BLOCKING */
+ if (csp->action->flags & ACTION_HANDLE_AS_EMPTY_DOCUMENT)
{
/*
- * Send empty document.
+ * Send empty document.
*/
new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
freez(rsp->body);
- rsp->body = strdup(" ");
+ rsp->body = strdup_or_die(" ");
rsp->content_length = 1;
- rsp->status = strdup("403 Request blocked by Privoxy");
- if (rsp->status == NULL)
+ if (csp->config->feature_flags & RUNTIME_FEATURE_EMPTY_DOC_RETURNS_OK)
{
- free_http_response(rsp);
- return cgi_error_memory();
+ /*
+ * Workaround for firefox bug 492459
+ * https://bugzilla.mozilla.org/show_bug.cgi?id=492459
+ * Return a 200 OK status for pages blocked with +handle-as-empty-document
+ * if the "handle-as-empty-doc-returns-ok" runtime config option is set.
+ */
+ rsp->status = strdup_or_die("200 Request blocked by Privoxy");
+ }
+ else
+ {
+ rsp->status = strdup_or_die("403 Request blocked by Privoxy");
}
+
if (new_content_type != 0)
{
log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
}
}
else
-#endif /* def FEATURE_IMAGE_BLOCKING */
/*
* Else, generate an HTML "blocked" message:
{
jb_err err;
struct map * exports;
- char *p;
-
- /*
- * Workaround for stupid Netscape bug which prevents
- * pages from being displayed if loading a referenced
- * JavaScript or style sheet fails. So make it appear
- * as if it succeeded.
- */
- if ( NULL != (p = get_header_value(csp->headers, "User-Agent:"))
- && !strncmpic(p, "mozilla", 7) /* Catch Netscape but */
- && !strstr(p, "Gecko") /* save Mozilla, */
- && !strstr(p, "compatible") /* MSIE */
- && !strstr(p, "Opera")) /* and Opera. */
- {
- rsp->status = strdup("200 Request for blocked URL");
- }
- else
- {
- rsp->status = strdup("403 Request for blocked URL");
- }
- if (rsp->status == NULL)
- {
- free_http_response(rsp);
- return cgi_error_memory();
- }
+ rsp->status = strdup_or_die("403 Request blocked by Privoxy");
exports = default_exports(csp, NULL);
if (exports == NULL)
if (!err) err = map(exports, "hostport", 1, html_encode(csp->http->hostport), 0);
if (!err) err = map(exports, "path", 1, html_encode(csp->http->path), 0);
if (!err) err = map(exports, "path-ue", 1, url_encode(csp->http->path), 0);
-
+ if (!err)
+ {
+ const char *block_reason;
+ if (csp->action->string[ACTION_STRING_BLOCK] != NULL)
+ {
+ block_reason = csp->action->string[ACTION_STRING_BLOCK];
+ }
+ else
+ {
+ assert(connect_port_is_forbidden(csp));
+ block_reason = "Forbidden CONNECT port.";
+ }
+ err = map(exports, "block-reason", 1, html_encode(block_reason), 0);
+ }
if (err)
{
free_map(exports);
return cgi_error_memory();
}
}
- rsp->reason = RSP_REASON_BLOCKED;
+ rsp->crunch_reason = BLOCKED;
return finish_http_response(csp, rsp);
struct map * exports;
char buf[BUFFER_SIZE];
char *p;
- struct url_spec **tl;
- struct url_spec *t;
+ struct pattern_spec **tl;
+ struct pattern_spec *t;
jb_err err;
/*
return cgi_error_memory();
}
- rsp->status = strdup("403 Request blocked by Privoxy");
+ rsp->status = strdup_or_die("403 Request blocked by Privoxy");
exports = default_exports(csp, NULL);
- if (exports == NULL || rsp->status == NULL)
+ if (exports == NULL)
{
free_http_response(rsp);
return cgi_error_memory();
* Export the protocol, host, port, and referrer information
*/
err = map(exports, "hostport", 1, csp->http->hostport, 1);
- if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
+ if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
if (!err) err = map(exports, "path", 1, csp->http->path, 1);
if (NULL != (p = get_header_value(csp->headers, "Referer:")))
/*
* Export the trust list
*/
- p = strdup("");
+ p = strdup_or_die("");
for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
{
snprintf(buf, sizeof(buf), "<li>%s</li>\n", t->spec);
{
struct list_entry *l;
- p = strdup("");
+ p = strdup_or_die("");
for (l = csp->config->trust_info->first; l ; l = l->next)
{
snprintf(buf, sizeof(buf), "<li> <a href=\"%s\">%s</a><br>\n", l->str, l->str);
free_http_response(rsp);
return cgi_error_memory();
}
- rsp->reason = RSP_REASON_UNTRUSTED;
+ rsp->crunch_reason = UNTRUSTED;
return finish_http_response(csp, rsp);
}
* 2 : b = The filter list to compile
*
* Returns : NULL in case of errors, otherwise the
- * pcrs job list.
+ * pcrs job list.
*
*********************************************************************/
pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
{"path", csp->http->path, 1},
{"host", csp->http->host, 1},
{"origin", csp->ip_addr_str, 1},
+ {"listen-address", csp->listen_addr_str, 1},
{NULL, NULL, 1}
};
dummy = pcrs_compile_dynamic_command(pattern->str, variables, &error);
if (NULL == dummy)
{
- assert(error < 0);
log_error(LOG_LEVEL_ERROR,
- "Adding filter job \'%s\' to dynamic filter %s failed: %s",
- pattern->str, b->name, pcrs_strerror(error));
+ "Compiling dynamic pcrs job '%s' for '%s' failed with error code %d: %s",
+ pattern->str, b->name, error, pcrs_strerror(error));
continue;
}
else
* 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
*
*
- * Returns : NULL if the pcrs_command didn't change the url, or
+ * Returns : NULL if the pcrs_command didn't change the url, or
* the result of the modification.
*
*********************************************************************/
*
* Parameters :
* 1 : subject = the string to check
- * 2 : redirect_mode = +fast-redirect{} mode
+ * 2 : redirect_mode = +fast-redirect{} mode
*
* Returns : NULL if no URL was found, or
* the last URL found.
return NULL;
}
- if (0 == strcmpic(redirect_mode, "check-decoded-url"))
- {
- log_error(LOG_LEVEL_REDIRECTS, "Decoding \"%s\" if necessary.", subject);
- new_url = url_decode(subject);
- if (new_url != NULL)
+ if (0 == strcmpic(redirect_mode, "check-decoded-url") && strchr(subject, '%'))
+ {
+ char *url_segment = NULL;
+ char **url_segments;
+ size_t max_segments;
+ int segments;
+
+ log_error(LOG_LEVEL_REDIRECTS,
+ "Checking \"%s\" for encoded redirects.", subject);
+
+ /*
+ * Check each parameter in the URL separately.
+ * Sectionize the URL at "?" and "&",
+ * go backwards through the segments, URL-decode them
+ * and look for a URL in the decoded result.
+ * Stop the search after the first match.
+ *
+ * XXX: This estimate is guaranteed to be high enough as we
+ * let ssplit() ignore empty fields, but also a bit wasteful.
+ */
+ max_segments = strlen(subject) / 2;
+ url_segments = malloc(max_segments * sizeof(char *));
+
+ if (NULL == url_segments)
{
+ log_error(LOG_LEVEL_ERROR,
+ "Out of memory while decoding URL: %s", subject);
freez(subject);
- subject = new_url;
+ return NULL;
}
- else
+
+ segments = ssplit(subject, "?&", url_segments, max_segments);
+
+ while (segments-- > 0)
{
- log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
+ char *dtoken = url_decode(url_segments[segments]);
+ if (NULL == dtoken)
+ {
+ log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", url_segments[segments]);
+ continue;
+ }
+ url_segment = strstr(dtoken, "http://");
+ if (NULL == url_segment)
+ {
+ url_segment = strstr(dtoken, "https://");
+ }
+ if (NULL != url_segment)
+ {
+ url_segment = strdup_or_die(url_segment);
+ freez(dtoken);
+ break;
+ }
+ freez(dtoken);
}
- }
+ freez(subject);
+ freez(url_segments);
- log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
+ if (url_segment == NULL)
+ {
+ return NULL;
+ }
+ subject = url_segment;
+ }
+ else
+ {
+ /* Look for a URL inside this one, without decoding anything. */
+ log_error(LOG_LEVEL_REDIRECTS,
+ "Checking \"%s\" for unencoded redirects.", subject);
+ }
/*
* Find the last URL encoded in the request
))
{
/*
- * Return new URL if we found a redirect
+ * Return new URL if we found a redirect
* or if the subject already was a URL.
*
* The second case makes sure that we can
* properly formatted URL and use it for the redirection
* directly.
*
- * According to RFC 2616 section 14.30 the URL
- * has to be absolute and if the user tries:
- * +redirect{shit/this/will/be/parsed/as/pcrs_command.html}
+ * According to (the now obsolete) RFC 2616 section 14.30
+ * the URL has to be absolute and if the user tries:
+ * +redirect{sadly/this/will/be/parsed/as/pcrs_command.html}
* she would get undefined results anyway.
*
+ * RFC 7231 7.1.2 actually allows relative references,
+ * but those start with a leading slash (RFC 3986 4.2) and
+ * thus can't be mistaken for pcrs commands either.
*/
if (*redirection_string == 's')
#endif /* def FEATURE_FAST_REDIRECTS */
csp->action->flags &= ~ACTION_REDIRECT;
- /* Did any redirect action trigger? */
+ /* Did any redirect action trigger? */
if (new_url)
{
+ if (url_requires_percent_encoding(new_url))
+ {
+ char *encoded_url;
+ log_error(LOG_LEVEL_REDIRECTS, "Percent-encoding redirect URL: %N",
+ strlen(new_url), new_url);
+ encoded_url = percent_encode_url(new_url);
+ freez(new_url);
+ if (encoded_url == NULL)
+ {
+ return cgi_error_memory();
+ }
+ new_url = encoded_url;
+ assert(FALSE == url_requires_percent_encoding(new_url));
+ }
+
if (0 == strcmpic(new_url, csp->http->url))
{
log_error(LOG_LEVEL_ERROR,
return cgi_error_memory();
}
- if ( enlist_unique_header(rsp->headers, "Location", new_url)
- || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
+ rsp->status = strdup_or_die("302 Local Redirect from Privoxy");
+ if (enlist_unique_header(rsp->headers, "Location", new_url))
{
freez(new_url);
free_http_response(rsp);
return cgi_error_memory();
}
- rsp->reason = RSP_REASON_REDIRECTED;
+ rsp->crunch_reason = REDIRECTED;
freez(new_url);
return finish_http_response(csp, rsp);
*
* Function : is_imageurl
*
- * Description : Given a URL, decide whether it is an image or not,
- * using either the info from a previous +image action
- * or, #ifdef FEATURE_IMAGE_DETECT_MSIE, and the browser
- * is MSIE and not on a Mac, tell from the browser's accept
- * header.
+ * Description : Given a URL, decide whether it should be treated
+ * as image URL or not.
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
*
- * Returns : True (nonzero) if URL is an image, false (0)
+ * Returns : True (nonzero) if URL is an image URL, false (0)
* otherwise
*
*********************************************************************/
int is_imageurl(const struct client_state *csp)
{
-#ifdef FEATURE_IMAGE_DETECT_MSIE
- char *tmp;
-
- tmp = get_header_value(csp->headers, "User-Agent:");
- if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
- {
- tmp = get_header_value(csp->headers, "Accept:");
- if (tmp && strstr(tmp, "image/gif"))
- {
- /* Client will accept HTML. If this seems counterintuitive,
- * blame Microsoft.
- */
- return(0);
- }
- else
- {
- return(1);
- }
- }
-#endif /* def FEATURE_IMAGE_DETECT_MSIE */
-
return ((csp->action->flags & ACTION_IMAGE) != 0);
}
{
struct file_list *fl;
struct block_spec *b;
- struct url_spec **trusted_url;
+ struct pattern_spec **trusted_url;
struct http_request rhttp[1];
const char * referer;
jb_err err;
/*
* If not, do we maybe trust its referrer?
*/
- err = parse_http_url(referer, rhttp, csp);
+ err = parse_http_url(referer, rhttp, REQUIRE_PROTOCOL);
if (err)
{
return 1;
{
char * path;
char * path_end;
- char * new_entry = strdup("~");
+ char * new_entry = strdup_or_die("~");
string_append(&new_entry, csp->http->hostport);
/* since this path points into a user's home space
* be sure to include this spec in the trustfile.
*/
- int path_len = path_end - path; /* save offset */
+ long path_len = path_end - path; /* save offset */
path = strdup(path); /* Copy string */
if (path != NULL)
{
#endif /* def FEATURE_TRUST */
+/*********************************************************************
+ *
+ * Function : get_filter
+ *
+ * Description : Get a filter with a given name and type.
+ * Note that taggers are filters, too.
+ *
+ * Parameters :
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ * 2 : requested_name = Name of the content filter to get
+ * 3 : requested_type = Type of the filter to tagger to lookup
+ *
+ * Returns : A pointer to the requested filter
+ * or NULL if the filter wasn't found
+ *
+ *********************************************************************/
+struct re_filterfile_spec *get_filter(const struct client_state *csp,
+ const char *requested_name,
+ enum filter_type requested_type)
+{
+ int i;
+ struct re_filterfile_spec *b;
+ struct file_list *fl;
+
+ for (i = 0; i < MAX_AF_FILES; i++)
+ {
+ fl = csp->rlist[i];
+ if ((NULL == fl) || (NULL == fl->f))
+ {
+ /*
+ * Either there are no filter files left or this
+ * filter file just contains no valid filters.
+ *
+ * Continue to be sure we don't miss valid filter
+ * files that are chained after empty or invalid ones.
+ */
+ continue;
+ }
+
+ for (b = fl->f; b != NULL; b = b->next)
+ {
+ if (b->type != requested_type)
+ {
+ /* The callers isn't interested in this filter type. */
+ continue;
+ }
+ if (strcmp(b->name, requested_name) == 0)
+ {
+ /* The requested filter has been found. Abort search. */
+ return b;
+ }
+ }
+ }
+
+ /* No filter with the given name and type exists. */
+ return NULL;
+
+}
+
+
/*********************************************************************
*
* Function : pcrs_filter_response
*********************************************************************/
static char *pcrs_filter_response(struct client_state *csp)
{
- int hits=0;
+ int hits = 0;
size_t size, prev_size;
char *old = NULL;
char *new = NULL;
pcrs_job *job;
- struct file_list *fl;
struct re_filterfile_spec *b;
struct list_entry *filtername;
- int i, found_filters = 0;
-
- /*
+ /*
* Sanity first
*/
if (csp->iob->cur >= csp->iob->eod)
return(NULL);
}
- /*
- * Need to check the set of re_filterfiles...
- */
- for (i = 0; i < MAX_AF_FILES; i++)
- {
- fl = csp->rlist[i];
- if (NULL != fl)
- {
- if (NULL != fl->f)
- {
- found_filters = 1;
- break;
- }
- }
- }
-
- if (0 == found_filters)
+ if (filters_available(csp) == FALSE)
{
- log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
+ log_error(LOG_LEVEL_ERROR, "Inconsistent configuration: "
+ "content filtering enabled, but no content filters available.");
return(NULL);
}
size = (size_t)(csp->iob->eod - csp->iob->cur);
old = csp->iob->cur;
- for (i = 0; i < MAX_AF_FILES; i++)
- {
- fl = csp->rlist[i];
- if ((NULL == fl) || (NULL == fl->f))
- {
- /*
- * Either there are no filter files
- * left, or this filter file just
- * contains no valid filters.
- *
- * Continue to be sure we don't miss
- * valid filter files that are chained
- * after empty or invalid ones.
- */
- continue;
- }
/*
* For all applying +filter actions, look if a filter by that
* name exists and if yes, execute it's pcrs_joblist on the
* buffer.
*/
- for (b = fl->f; b; b = b->next)
+ for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
+ filtername != NULL; filtername = filtername->next)
{
- if (b->type != FT_CONTENT_FILTER)
+ int current_hits = 0; /* Number of hits caused by this filter */
+ int job_number = 0; /* Which job we're currently executing */
+ int job_hits = 0; /* How many hits the current job caused */
+ pcrs_job *joblist;
+
+ b = get_filter(csp, filtername->str, FT_CONTENT_FILTER);
+ if (b == NULL)
{
- /* Skip header filters */
continue;
}
- for (filtername = csp->action->multi[ACTION_MULTI_FILTER]->first;
- filtername ; filtername = filtername->next)
- {
- if (strcmp(b->name, filtername->str) == 0)
- {
- int current_hits = 0; /* Number of hits caused by this filter */
- int job_number = 0; /* Which job we're currently executing */
- int job_hits = 0; /* How many hits the current job caused */
- pcrs_job *joblist = b->joblist;
+ joblist = b->joblist;
- if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
+ if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
- if (NULL == joblist)
- {
- log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
- continue;
- }
+ if (NULL == joblist)
+ {
+ log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
+ continue;
+ }
- prev_size = size;
- /* Apply all jobs from the joblist */
- for (job = joblist; NULL != job; job = job->next)
- {
- job_number++;
- job_hits = pcrs_execute(job, old, size, &new, &size);
+ prev_size = size;
+ /* Apply all jobs from the joblist */
+ for (job = joblist; NULL != job; job = job->next)
+ {
+ job_number++;
+ job_hits = pcrs_execute(job, old, size, &new, &size);
- if (job_hits >= 0)
- {
- /*
- * That went well. Continue filtering
- * and use the result of this job as
- * input for the next one.
- */
- current_hits += job_hits;
- if (old != csp->iob->cur)
- {
- freez(old);
- }
- old = new;
- }
- else
- {
- /*
- * This job caused an unexpected error. Inform the user
- * and skip the rest of the jobs in this filter. We could
- * continue with the next job, but usually the jobs
- * depend on each other or are similar enough to
- * fail for the same reason.
- *
- * At the moment our pcrs expects the error codes of pcre 3.4,
- * but newer pcre versions can return additional error codes.
- * As a result pcrs_strerror()'s error message might be
- * "Unknown error ...", therefore we print the numerical value
- * as well.
- *
- * XXX: Is this important enough for LOG_LEVEL_ERROR or
- * should we use LOG_LEVEL_RE_FILTER instead?
- */
- log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
- b->name, job_number, pcrs_strerror(job_hits), job_hits);
- break;
- }
+ if (job_hits >= 0)
+ {
+ /*
+ * That went well. Continue filtering
+ * and use the result of this job as
+ * input for the next one.
+ */
+ current_hits += job_hits;
+ if (old != csp->iob->cur)
+ {
+ freez(old);
}
+ old = new;
+ }
+ else
+ {
+ /*
+ * This job caused an unexpected error. Inform the user
+ * and skip the rest of the jobs in this filter. We could
+ * continue with the next job, but usually the jobs
+ * depend on each other or are similar enough to
+ * fail for the same reason.
+ *
+ * At the moment our pcrs expects the error codes of pcre 3.4,
+ * but newer pcre versions can return additional error codes.
+ * As a result pcrs_strerror()'s error message might be
+ * "Unknown error ...", therefore we print the numerical value
+ * as well.
+ *
+ * XXX: Is this important enough for LOG_LEVEL_ERROR or
+ * should we use LOG_LEVEL_RE_FILTER instead?
+ */
+ log_error(LOG_LEVEL_ERROR, "Skipped filter \'%s\' after job number %u: %s (%d)",
+ b->name, job_number, pcrs_strerror(job_hits), job_hits);
+ break;
+ }
+ }
- if (b->dynamic) pcrs_free_joblist(joblist);
+ if (b->dynamic) pcrs_free_joblist(joblist);
- log_error(LOG_LEVEL_RE_FILTER,
- "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
- csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
+ log_error(LOG_LEVEL_RE_FILTER,
+ "filtering %s%s (size %d) with \'%s\' produced %d hits (new size %d).",
+ csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
- hits += current_hits;
- }
- }
- }
+ hits += current_hits;
}
/*
csp->flags |= CSP_FLAG_MODIFIED;
csp->content_length = size;
- IOB_RESET(csp);
+ clear_iob(csp->iob);
return(new);
}
+#ifdef FEATURE_EXTERNAL_FILTERS
/*********************************************************************
*
- * Function : gif_deanimate_response
+ * Function : get_external_filter
*
- * Description : Deanimate the GIF image that has been accumulated in
- * csp->iob->buf, set csp->content_length to the modified
- * size and raise the CSP_FLAG_MODIFIED flag.
+ * Description : Lookup the code to execute for an external filter.
+ * Masks the misuse of the re_filterfile_spec.
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
+ * 2 : name = Name of the content filter to get
+ *
+ * Returns : A pointer to the requested code
+ * or NULL if the filter wasn't found
+ *
+ *********************************************************************/
+static const char *get_external_filter(const struct client_state *csp,
+ const char *name)
+{
+ struct re_filterfile_spec *external_filter;
+
+ external_filter = get_filter(csp, name, FT_EXTERNAL_CONTENT_FILTER);
+ if (external_filter == NULL)
+ {
+ log_error(LOG_LEVEL_FATAL,
+ "Didn't find stuff to execute for external filter: %s",
+ name);
+ }
+
+ return external_filter->patterns->first->str;
+
+}
+
+
+/*********************************************************************
+ *
+ * Function : set_privoxy_variables
+ *
+ * Description : Sets a couple of privoxy-specific environment variables
+ *
+ * Parameters :
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ *
+ * Returns : N/A
+ *
+ *********************************************************************/
+static void set_privoxy_variables(const struct client_state *csp)
+{
+ int i;
+ struct {
+ const char *name;
+ const char *value;
+ } env[] = {
+ { "PRIVOXY_URL", csp->http->url },
+ { "PRIVOXY_PATH", csp->http->path },
+ { "PRIVOXY_HOST", csp->http->host },
+ { "PRIVOXY_ORIGIN", csp->ip_addr_str },
+ { "PRIVOXY_LISTEN_ADDRESS", csp->listen_addr_str },
+ };
+
+ for (i = 0; i < SZ(env); i++)
+ {
+ if (setenv(env[i].name, env[i].value, 1))
+ {
+ log_error(LOG_LEVEL_ERROR, "Failed to set %s=%s: %E",
+ env[i].name, env[i].value);
+ }
+ }
+}
+
+
+/*********************************************************************
+ *
+ * Function : execute_external_filter
+ *
+ * Description : Pipe content into external filter and return the output
+ *
+ * Parameters :
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ * 2 : name = Name of the external filter to execute
+ * 3 : content = The original content to filter
+ * 4 : size = The size of the content buffer
*
* Returns : a pointer to the (newly allocated) modified buffer.
- * or NULL in case something went wrong.
+ * or NULL if there were no hits or something went wrong
*
*********************************************************************/
-static char *gif_deanimate_response(struct client_state *csp)
+static char *execute_external_filter(const struct client_state *csp,
+ const char *name, char *content, size_t *size)
{
- struct binbuffer *in, *out;
- char *p;
- size_t size;
+ char cmd[200];
+ char file_name[FILENAME_MAX];
+ FILE *fp;
+ char *filter_output;
+ int fd;
+ int ret;
+ size_t new_size;
+ const char *external_filter;
- size = (size_t)(csp->iob->eod - csp->iob->cur);
+ if (csp->config->temporary_directory == NULL)
+ {
+ log_error(LOG_LEVEL_ERROR,
+ "No temporary-directory configured. Can't execute filter: %s",
+ name);
+ return NULL;
+ }
+
+ external_filter = get_external_filter(csp, name);
- if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
- || (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
+ if (sizeof(file_name) < snprintf(file_name, sizeof(file_name),
+ "%s/privoxy-XXXXXXXX", csp->config->temporary_directory))
{
- log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
+ log_error(LOG_LEVEL_ERROR, "temporary-directory path too long");
return NULL;
}
- in->buffer = csp->iob->cur;
- in->size = size;
+ fd = mkstemp(file_name);
+ if (fd == -1)
+ {
+ log_error(LOG_LEVEL_ERROR, "mkstemp() failed to create %s: %E", file_name);
+ return NULL;
+ }
- if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
+ fp = fdopen(fd, "w");
+ if (fp == NULL)
{
- log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
- freez(in);
- buf_free(out);
- return(NULL);
+ log_error(LOG_LEVEL_ERROR, "fdopen() failed: %E");
+ unlink(file_name);
+ return NULL;
}
- else
+
+ /*
+ * The size may be zero if a previous filter discarded everything.
+ *
+ * This isn't necessary unintentional, so we just don't try
+ * to fwrite() nothing and let the user deal with the rest.
+ */
+ if ((*size != 0) && fwrite(content, *size, 1, fp) != 1)
{
- if ((int)size == out->offset)
+ log_error(LOG_LEVEL_ERROR, "fwrite(..., %d, 1, ..) failed: %E", *size);
+ unlink(file_name);
+ fclose(fp);
+ return NULL;
+ }
+ fclose(fp);
+
+ if (sizeof(cmd) < snprintf(cmd, sizeof(cmd), "%s < %s", external_filter, file_name))
+ {
+ log_error(LOG_LEVEL_ERROR,
+ "temporary-directory or external filter path too long");
+ unlink(file_name);
+ return NULL;
+ }
+
+ log_error(LOG_LEVEL_RE_FILTER, "Executing '%s': %s", name, cmd);
+
+ /*
+ * The locking is necessary to prevent other threads
+ * from overwriting the environment variables before
+ * the popen fork. Afterwards this no longer matters.
+ */
+ privoxy_mutex_lock(&external_filter_mutex);
+ set_privoxy_variables(csp);
+ fp = popen(cmd, "r");
+ privoxy_mutex_unlock(&external_filter_mutex);
+ if (fp == NULL)
+ {
+ log_error(LOG_LEVEL_ERROR, "popen(\"%s\", \"r\") failed: %E", cmd);
+ unlink(file_name);
+ return NULL;
+ }
+
+ /* Allocate at least one byte */
+ filter_output = malloc_or_die(*size + 1);
+
+ new_size = 0;
+ while (!feof(fp) && !ferror(fp))
+ {
+ size_t len;
+ /* Could be bigger ... */
+ enum { READ_LENGTH = 2048 };
+
+ if (new_size + READ_LENGTH >= *size)
{
- log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
+ char *p;
+
+ /* Could be considered wasteful if the content is 'large'. */
+ *size += (*size >= READ_LENGTH) ? *size : READ_LENGTH;
+
+ p = realloc(filter_output, *size);
+ if (p == NULL)
+ {
+ log_error(LOG_LEVEL_ERROR, "Out of memory while reading "
+ "external filter output. Using what we got so far.");
+ break;
+ }
+ filter_output = p;
}
- else
+ assert(new_size + READ_LENGTH < *size);
+ len = fread(&filter_output[new_size], 1, READ_LENGTH, fp);
+ if (len > 0)
{
- log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
- }
- csp->content_length = out->offset;
- csp->flags |= CSP_FLAG_MODIFIED;
- p = out->buffer;
- freez(in);
- freez(out);
- return(p);
+ new_size += len;
+ }
+ }
+
+ ret = pclose(fp);
+ if (ret == -1)
+ {
+ log_error(LOG_LEVEL_ERROR, "Executing %s failed: %E", cmd);
}
+ else
+ {
+ log_error(LOG_LEVEL_RE_FILTER,
+ "Executing '%s' resulted in return value %d. "
+ "Read %d of up to %d bytes.", name, (ret >> 8), new_size, *size);
+ }
+
+ unlink(file_name);
+ *size = new_size;
+
+ return filter_output;
}
+#endif /* def FEATURE_EXTERNAL_FILTERS */
/*********************************************************************
*
- * Function : jpeg_inspect_response
+ * Function : gif_deanimate_response
*
- * Description :
+ * Description : Deanimate the GIF image that has been accumulated in
+ * csp->iob->buf, set csp->content_length to the modified
+ * size and raise the CSP_FLAG_MODIFIED flag.
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
*
- * Returns : a pointer to the (newly allocated) modified buffer
+ * Returns : a pointer to the (newly allocated) modified buffer.
* or NULL in case something went wrong.
*
*********************************************************************/
-static char *jpeg_inspect_response(struct client_state *csp)
+#ifdef FUZZ
+char *gif_deanimate_response(struct client_state *csp)
+#else
+static char *gif_deanimate_response(struct client_state *csp)
+#endif
{
- struct binbuffer *in = NULL;
- struct binbuffer *out = NULL;
- char *p = NULL;
+ struct binbuffer *in, *out;
+ char *p;
size_t size;
size = (size_t)(csp->iob->eod - csp->iob->cur);
- if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
- {
- log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
- return NULL;
- }
-
- if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
- {
- log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
- return NULL;
- }
+ in = zalloc_or_die(sizeof(*in));
+ out = zalloc_or_die(sizeof(*out));
in->buffer = csp->iob->cur;
in->size = size;
- /*
- * Calling jpeg_inspect has the side-effect of creating and
- * modifying the image buffer of "out" directly.
- */
- if (jpeg_inspect(in, out))
+ if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
{
- log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
+ log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
freez(in);
buf_free(out);
return(NULL);
-
}
else
{
+ if ((int)size == out->offset)
+ {
+ log_error(LOG_LEVEL_DEANIMATE, "GIF not changed.");
+ }
+ else
+ {
+ log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
+ }
csp->content_length = out->offset;
csp->flags |= CSP_FLAG_MODIFIED;
p = out->buffer;
* Function : get_filter_function
*
* Description : Decides which content filter function has
- * to be applied (if any).
- *
- * XXX: Doesn't handle filter_popups()
- * because of the different prototype. Probably
- * we should ditch filter_popups() anyway, it's
- * even less reliable than popup blocking based
- * on pcrs filters.
+ * to be applied (if any). Only considers functions
+ * for internal filters which are mutually-exclusive.
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
* NULL if no content filter is active
*
*********************************************************************/
-filter_function_ptr get_filter_function(struct client_state *csp)
+static filter_function_ptr get_filter_function(const struct client_state *csp)
{
filter_function_ptr filter_function = NULL;
- /*
- * Are we enabling text mode by force?
- */
- if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
- {
- /*
- * Do we really have to?
- */
- if (csp->content_type & CT_TEXT)
- {
- log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
- }
- else
- {
- csp->content_type |= CT_TEXT;
- log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
- }
- }
-
- if (!(csp->content_type & CT_DECLARED))
- {
- /*
- * The server didn't bother to declare a MIME-Type.
- * Assume it's text that can be filtered.
- *
- * This also regulary happens with 304 responses,
- * therefore logging anything here would cause
- * too much noise.
- */
- csp->content_type |= CT_TEXT;
- }
-
/*
* Choose the applying filter function based on
* the content type and action settings.
*/
if ((csp->content_type & CT_TEXT) &&
- (csp->rlist != NULL) &&
(!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER])))
{
filter_function = pcrs_filter_response;
}
- else if ((csp->content_type & CT_GIF) &&
+ else if ((csp->content_type & CT_GIF) &&
(csp->action->flags & ACTION_DEANIMATE))
{
filter_function = gif_deanimate_response;
}
- else if ((csp->content_type & CT_JPEG) &&
- (csp->action->flags & ACTION_JPEG_INSPECT))
- {
- filter_function = jpeg_inspect_response;
- }
return filter_function;
}
* Function : remove_chunked_transfer_coding
*
* Description : In-situ remove the "chunked" transfer coding as defined
- * in rfc2616 from a buffer.
+ * in RFC 7230 4.1 from a buffer. XXX: The implementation
+ * is neither complete nor compliant (TODO #129).
*
* Parameters :
* 1 : buffer = Pointer to the text buffer
* JB_ERR_PARSE otherwise
*
*********************************************************************/
+#ifdef FUZZ
+extern jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
+#else
static jb_err remove_chunked_transfer_coding(char *buffer, size_t *size)
+#endif
{
size_t newsize = 0;
unsigned int chunksize = 0;
char *from_p, *to_p;
+ const char *end_of_buffer = buffer + *size;
+
+ if (*size == 0)
+ {
+ log_error(LOG_LEVEL_FATAL, "Invalid chunked input. Buffer is empty.");
+ return JB_ERR_PARSE;
+ }
assert(buffer);
from_p = to_p = buffer;
return JB_ERR_PARSE;
}
- while (chunksize > 0)
+ while (chunksize > 0U)
{
+ /*
+ * If the chunk-size is valid, we should have at least
+ * chunk-size bytes of chunk-data and five bytes of
+ * meta data (chunk-size, CRLF, CRLF) left in the buffer.
+ */
+ if (chunksize + 5 >= *size - newsize)
+ {
+ log_error(LOG_LEVEL_ERROR,
+ "Chunk size %u exceeds buffered data left. "
+ "Already digested %u of %u buffered bytes.",
+ chunksize, (unsigned int)newsize, (unsigned int)*size);
+ return JB_ERR_PARSE;
+ }
+
+ /*
+ * Skip the chunk-size, the optional chunk-ext and the CRLF
+ * that is supposed to be located directly before the start
+ * of chunk-data.
+ */
if (NULL == (from_p = strstr(from_p, "\r\n")))
{
log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
return JB_ERR_PARSE;
}
+ from_p += 2;
- if ((newsize += chunksize) >= *size)
+ /*
+ * The previous strstr() does not enforce chunk-validity
+ * and is sattisfied as long a CRLF is left in the buffer.
+ *
+ * Make sure the bytes we consider chunk-data are within
+ * the valid range.
+ */
+ if (from_p + chunksize >= end_of_buffer)
{
- log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
+ log_error(LOG_LEVEL_ERROR,
+ "End of chunk is beyond the end of the buffer.");
return JB_ERR_PARSE;
}
- from_p += 2;
memmove(to_p, from_p, (size_t) chunksize);
+ newsize += chunksize;
to_p = buffer + newsize;
- from_p += chunksize + 2;
+ from_p += chunksize;
+ /*
+ * Not merging this check with the previous one allows us
+ * to keep chunks without trailing CRLF. It's not clear
+ * if we actually have to care about those, though.
+ */
+ if (from_p + 2 >= end_of_buffer)
+ {
+ log_error(LOG_LEVEL_ERROR, "Not enough room for trailing CRLF.");
+ return JB_ERR_PARSE;
+ }
+ from_p += 2;
if (sscanf(from_p, "%x", &chunksize) != 1)
{
log_error(LOG_LEVEL_INFO, "Invalid \"chunked\" transfer encoding detected and ignored.");
break;
}
}
-
+
/* XXX: Should get its own loglevel. */
log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", *size, newsize);
* If the body has a supported transfer-encoding,
* decompress it, adjusting size and iob->eod.
*/
- if (csp->content_type & (CT_GZIP|CT_DEFLATE))
+ if ((csp->content_type & (CT_GZIP|CT_DEFLATE))
+#ifdef FEATURE_BROTLI
+ || (csp->content_type & CT_BROTLI)
+#endif
+ )
{
if (0 == csp->iob->eod - csp->iob->cur)
{
else
{
/*
- * Unset CT_GZIP and CT_DEFLATE to remember not
- * to modify the Content-Encoding header later.
+ * Unset content types to remember not to
+ * modify the Content-Encoding header later.
*/
csp->content_type &= ~CT_GZIP;
csp->content_type &= ~CT_DEFLATE;
+#ifdef FEATURE_BROTLI
+ csp->content_type &= ~CT_BROTLI;
+#endif
}
}
#endif
/*********************************************************************
*
- * Function : execute_content_filter
+ * Function : execute_content_filters
*
* Description : Executes a given content filter.
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
- * 2 : content_filter = The filter function to execute
*
* Returns : Pointer to the modified buffer, or
* NULL if filtering failed or wasn't necessary.
*
*********************************************************************/
-char *execute_content_filter(struct client_state *csp, filter_function_ptr content_filter)
+char *execute_content_filters(struct client_state *csp)
{
+ char *content;
+ filter_function_ptr content_filter;
+
+ assert(content_filters_enabled(csp->action));
+
if (0 == csp->iob->eod - csp->iob->cur)
{
/*
return NULL;
}
- return ((*content_filter)(csp));
+ content_filter = get_filter_function(csp);
+ content = (content_filter != NULL) ? (*content_filter)(csp) : NULL;
+
+#ifdef FEATURE_EXTERNAL_FILTERS
+ if ((csp->content_type & CT_TEXT) &&
+ !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]))
+ {
+ struct list_entry *filtername;
+ size_t size = (size_t)csp->content_length;
+
+ if (content == NULL)
+ {
+ content = csp->iob->cur;
+ size = (size_t)(csp->iob->eod - csp->iob->cur);
+ }
+
+ for (filtername = csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER]->first;
+ filtername ; filtername = filtername->next)
+ {
+ char *result = execute_external_filter(csp, filtername->str, content, &size);
+ if (result != NULL)
+ {
+ if (content != csp->iob->cur)
+ {
+ free(content);
+ }
+ content = result;
+ }
+ }
+ csp->flags |= CSP_FLAG_MODIFIED;
+ csp->content_length = size;
+ }
+#endif /* def FEATURE_EXTERNAL_FILTERS */
+
+ return content;
+
}
/*********************************************************************
*
- * Function : url_actions
+ * Function : get_url_actions
*
* Description : Gets the actions for this URL.
*
* Parameters :
- * 1 : http = http_request request for blocked URLs
- * 2 : csp = Current client state (buffers, headers, etc...)
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ * 2 : http = http_request request for blocked URLs
*
* Returns : N/A
*
*********************************************************************/
-void url_actions(struct http_request *http,
- struct client_state *csp)
+void get_url_actions(struct client_state *csp, struct http_request *http)
{
struct file_list *fl;
struct url_actions *b;
int i;
- init_current_action(csp->action);
+#ifdef FEATURE_HTTPS_INSPECTION
+ if (!csp->http->client_ssl)
+#endif
+ {
+ /*
+ * When filtering TLS traffic this function gets called a
+ * second time after the encrypted headers have been received.
+ *
+ * Only initialize the first time. The second time we apply
+ * the newly set actions on top of the ones that were set
+ * the first time.
+ */
+ init_current_action(csp->action);
+ }
for (i = 0; i < MAX_AF_FILES; i++)
{
return;
}
+#ifdef FEATURE_CLIENT_TAGS
+ apply_url_actions(csp->action, http, csp->client_tags, b);
+#else
apply_url_actions(csp->action, http, b);
+#endif
}
return;
}
-
/*********************************************************************
*
* Function : apply_url_actions
* Parameters :
* 1 : action = Destination.
* 2 : http = Current URL
- * 3 : b = list of URL actions to apply
+ * 3 : client_tags = list of client tags
+ * 4 : b = list of URL actions to apply
*
* Returns : N/A
*
*********************************************************************/
-void apply_url_actions(struct current_action_spec *action,
- struct http_request *http,
- struct url_actions *b)
+static void apply_url_actions(struct current_action_spec *action,
+ struct http_request *http,
+#ifdef FEATURE_CLIENT_TAGS
+ const struct list *client_tags,
+#endif
+ struct url_actions *b)
{
if (b == NULL)
{
{
merge_current_action(action, b->action);
}
+#ifdef FEATURE_CLIENT_TAGS
+ if (client_tag_match(b->url, client_tags))
+ {
+ merge_current_action(action, b->action);
+ }
+#endif
}
}
* the lifetime of this request. Save its location
* in csp as well, so sweep() can free it later on.
*/
- fwd = csp->fwd = zalloc(sizeof(*fwd));
- if (NULL == fwd)
- {
- log_error(LOG_LEVEL_FATAL,
- "can't allocate memory for forward-override{%s}", forward_override_line);
- /* Never get here - LOG_LEVEL_FATAL causes program exit */
- }
+ fwd = csp->fwd = zalloc_or_die(sizeof(*fwd));
- vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
+ vec_count = ssplit(forward_settings, " \t", vec, SZ(vec));
if ((vec_count == 2) && !strcasecmp(vec[0], "forward"))
{
fwd->type = SOCKS_NONE;
/* Parse the parent HTTP proxy host:port */
http_parent = vec[1];
+ }
+ else if ((vec_count == 2) && !strcasecmp(vec[0], "forward-webserver"))
+ {
+ fwd->type = FORWARD_WEBSERVER;
+
+ /* Parse the parent HTTP server host:port */
+ http_parent = vec[1];
+
}
else if (vec_count == 3)
{
fwd->type = SOCKS_4A;
socks_proxy = vec[1];
}
+ else if (!strcasecmp(vec[0], "forward-socks5"))
+ {
+ fwd->type = SOCKS_5;
+ socks_proxy = vec[1];
+ }
+ else if (!strcasecmp(vec[0], "forward-socks5t"))
+ {
+ fwd->type = SOCKS_5T;
+ socks_proxy = vec[1];
+ }
if (NULL != socks_proxy)
{
- /* Parse the SOCKS proxy host[:port] */
- fwd->gateway_host = strdup(socks_proxy);
-
- if (NULL != (socks_proxy = strchr(fwd->gateway_host, ':')))
- {
- *socks_proxy++ = '\0';
- fwd->gateway_port = strtol(socks_proxy, NULL, 0);
- }
-
- if (fwd->gateway_port <= 0)
- {
- fwd->gateway_port = 1080;
- }
+ /* Parse the SOCKS proxy [user:pass@]host[:port] */
+ fwd->gateway_port = 1080;
+ parse_forwarder_address(socks_proxy,
+ &fwd->gateway_host, &fwd->gateway_port,
+ &fwd->auth_username, &fwd->auth_password);
http_parent = vec[2];
}
/* Parse http forwarding settings */
if (strcmp(http_parent, ".") != 0)
{
- fwd->forward_host = strdup(http_parent);
-
- if (NULL != (http_parent = strchr(fwd->forward_host, ':')))
- {
- *http_parent++ = '\0';
- fwd->forward_port = strtol(http_parent, NULL, 0);
- }
-
- if (fwd->forward_port <= 0)
- {
- fwd->forward_port = 8000;
- }
+ fwd->forward_port = 8000;
+ parse_forwarder_address(http_parent,
+ &fwd->forward_host, &fwd->forward_port,
+ NULL, NULL);
}
assert (NULL != fwd);
*
* Description : Should we forward this to another proxy?
*
- * XXX: Should be changed to make use of csp->fwd.
- *
* Parameters :
- * 1 : http = http_request request for current URL
- * 2 : csp = Current client state (buffers, headers, etc...)
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ * 2 : http = http_request request for current URL
*
* Returns : Pointer to forwarding information.
*
*********************************************************************/
-const struct forward_spec * forward_url(struct http_request *http,
- struct client_state *csp)
+const struct forward_spec *forward_url(struct client_state *csp,
+ const struct http_request *http)
{
- static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
+ static const struct forward_spec fwd_default[1]; /* Zero'ed due to being static. */
struct forward_spec *fwd = csp->config->forward;
if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
/*********************************************************************
*
- * Function : direct_response
+ * Function : direct_response
*
* Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
* request and if so, return a HTTP 501 to the client.
* requests properly. Still, what we do here is rfc-
* compliant, whereas ignoring or forwarding are not.
*
- * Parameters :
+ * Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
*
* Returns : http_response if , NULL if nonmatch or handler fail
{
for (p = csp->headers->first; (p != NULL) ; p = p->next)
{
- if (!strncmpic("Max-Forwards:", p->str, 13))
+ if (!strncmpic(p->str, "Max-Forwards:", 13))
{
unsigned int max_forwards;
{
return cgi_error_memory();
}
-
- if (NULL == (rsp->status = strdup("501 Not Implemented")))
- {
- free_http_response(rsp);
- return cgi_error_memory();
- }
+ rsp->status = strdup_or_die("501 Not Implemented");
rsp->is_static = 1;
- rsp->reason = RSP_REASON_UNSUPPORTED;
+ rsp->crunch_reason = UNSUPPORTED;
return(finish_http_response(csp, rsp));
}
}
+/*********************************************************************
+ *
+ * Function : content_requires_filtering
+ *
+ * Description : Checks whether there are any content filters
+ * enabled for the current request and if they
+ * can actually be applied..
+ *
+ * Parameters :
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ *
+ * Returns : TRUE for yes, FALSE otherwise
+ *
+ *********************************************************************/
+int content_requires_filtering(struct client_state *csp)
+{
+ if ((csp->content_type & CT_TABOO)
+ && !(csp->action->flags & ACTION_FORCE_TEXT_MODE))
+ {
+ return FALSE;
+ }
+
+ /*
+ * Are we enabling text mode by force?
+ */
+ if (csp->action->flags & ACTION_FORCE_TEXT_MODE)
+ {
+ /*
+ * Do we really have to?
+ */
+ if (csp->content_type & CT_TEXT)
+ {
+ log_error(LOG_LEVEL_HEADER, "Text mode is already enabled.");
+ }
+ else
+ {
+ csp->content_type |= CT_TEXT;
+ log_error(LOG_LEVEL_HEADER, "Text mode enabled by force. Take cover!");
+ }
+ }
+
+ if (!(csp->content_type & CT_DECLARED))
+ {
+ /*
+ * The server didn't bother to declare a MIME-Type.
+ * Assume it's text that can be filtered.
+ *
+ * This also regulary happens with 304 responses,
+ * therefore logging anything here would cause
+ * too much noise.
+ */
+ csp->content_type |= CT_TEXT;
+ }
+
+ /*
+ * Choose the applying filter function based on
+ * the content type and action settings.
+ */
+ if ((csp->content_type & CT_TEXT) &&
+ (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER]) ||
+ !list_is_empty(csp->action->multi[ACTION_MULTI_EXTERNAL_FILTER])))
+ {
+ return TRUE;
+ }
+ else if ((csp->content_type & CT_GIF) &&
+ (csp->action->flags & ACTION_DEANIMATE))
+ {
+ return TRUE;
+ }
+
+ return FALSE;
+
+}
+
+
/*********************************************************************
*
* Function : content_filters_enabled
* Description : Checks whether there are any content filters
* enabled for the current request.
*
- * Parameters :
- * 1 : csp = Current client state (buffers, headers, etc...)
+ * Parameters :
+ * 1 : action = Action spec to check.
*
* Returns : TRUE for yes, FALSE otherwise
*
*********************************************************************/
-inline int content_filters_enabled(const struct client_state *csp)
+int content_filters_enabled(const struct current_action_spec *action)
+{
+ return ((action->flags & ACTION_DEANIMATE) ||
+ !list_is_empty(action->multi[ACTION_MULTI_FILTER]) ||
+ !list_is_empty(action->multi[ACTION_MULTI_EXTERNAL_FILTER]));
+}
+
+
+/*********************************************************************
+ *
+ * Function : filters_available
+ *
+ * Description : Checks whether there are any filters available.
+ *
+ * Parameters :
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ *
+ * Returns : TRUE for yes, FALSE otherwise.
+ *
+ *********************************************************************/
+int filters_available(const struct client_state *csp)
{
- return (((csp->rlist != NULL) &&
- (!list_is_empty(csp->action->multi[ACTION_MULTI_FILTER]))) ||
- (csp->action->flags & (ACTION_DEANIMATE|ACTION_JPEG_INSPECT|ACTION_NO_POPUPS)));
+ int i;
+ for (i = 0; i < MAX_AF_FILES; i++)
+ {
+ const struct file_list *fl = csp->rlist[i];
+ if ((NULL != fl) && (NULL != fl->f))
+ {
+ return TRUE;
+ }
+ }
+ return FALSE;
}
+
/*
Local Variables:
tab-width: 3