-const char filters_rcs[] = "$Id: filters.c,v 1.20 2001/07/01 17:01:04 oes Exp $";
+const char filters_rcs[] = "$Id: filters.c,v 1.69 2006/12/08 12:39:13 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', `domaincmp', `dsplit',
+ * `block_url', `url_actions', `domain_split',
* `filter_popups', `forward_url', 'redirect_url',
* `ij_untrusted_url', `intercept_url', `pcrs_filter_respose',
- * `show_proxy_args', 'ijb_send_banner', and `trust_url'
+ * `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 the SourceForge
- * IJBSWA team. http://ijbswa.sourceforge.net
+ * Copyright : Written by and Copyright (C) 2001, 2004-2006 the SourceForge
+ * Privoxy team. http://www.privoxy.org/
*
* Based on the Internet Junkbuster originally written
- * by and Copyright (C) 1997 Anonymous Coders and
+ * by and Copyright (C) 1997 Anonymous Coders and
* Junkbusters Corporation. http://www.junkbusters.com
*
- * This program is free software; you can redistribute it
+ * This program is free software; you can redistribute it
* and/or modify it under the terms of the GNU General
* Public License as published by the Free Software
* Foundation; either version 2 of the License, or (at
*
* Revisions :
* $Log: filters.c,v $
+ * 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()
*
* a lot! ;-)
*
* Revision 1.3 2001/05/20 16:44:47 jongfoster
- * Removing last hardcoded JunkBusters.com URLs.
+ * Removing last hardcoded Junkbusters.com URLs.
*
* Revision 1.2 2001/05/20 01:21:20 jongfoster
* Version 2.9.4 checkin.
#include <stdlib.h>
#include <ctype.h>
#include <string.h>
+#include <assert.h>
#ifndef _WIN32
+#ifndef __OS2__
#include <unistd.h>
+#endif /* ndef __OS2__ */
#include <netinet/in.h>
#else
#include <winsock2.h>
-#endif
+#endif /* ndef _WIN32 */
+
+#ifdef __OS2__
+#include <utils.h>
+#endif /* def __OS2__ */
#include "project.h"
#include "filters.h"
#include "encode.h"
-#include "jcc.h"
-#include "showargs.h"
#include "parsers.h"
#include "ssplit.h"
-#include "gateway.h"
-#include "jbsockets.h"
#include "errlog.h"
#include "jbsockets.h"
#include "miscutil.h"
#include "cgi.h"
#include "list.h"
#include "deanimate.h"
+#include "urlmatch.h"
#ifdef _WIN32
#include "win32.h"
#define ijb_isdigit(__X) isdigit((int)(unsigned char)(__X))
-#ifdef ACL_FILES
+#ifdef FEATURE_ACL
/*********************************************************************
*
* Function : block_acl
*
* Function : acl_addr
*
- * Description : Called from `load_aclfile' to parse an ACL address.
+ * Description : Called from `load_config' to parse an ACL address.
*
* Parameters :
* 1 : aspec = String specifying ACL address.
masklength = 32;
port = 0;
- if ((p = strchr(aspec, '/')))
+ if ((p = strchr(aspec, '/')) != NULL)
{
*p++ = '\0';
return(-1);
}
- if ((p = strchr(aspec, ':')))
+ if ((p = strchr(aspec, ':')) != NULL)
{
*p++ = '\0';
aca->addr = ntohl(resolve_hostname_to_ip(aspec));
- if (aca->addr == -1)
+ if (aca->addr == INADDR_NONE)
{
- log_error(LOG_LEVEL_ERROR, "can't resolve address for %s", aspec);
return(-1);
}
return(0);
}
-#endif /* def ACL_FILES */
+#endif /* def FEATURE_ACL */
+
+
+/*********************************************************************
+ *
+ * Function : match_portlist
+ *
+ * Description : Check if a given number is covered by a comma
+ * separated list of numbers and ranges (a,b-c,d,..)
+ *
+ * Parameters :
+ * 1 : portlist = String with list
+ * 2 : port = port to check
+ *
+ * Returns : 0 => no match
+ * 1 => match
+ *
+ *********************************************************************/
+int match_portlist(const char *portlist, int port)
+{
+ char *min, *max, *next, *portlist_copy;
+
+ min = next = portlist_copy = strdup(portlist);
+
+ /*
+ * Zero-terminate first item and remember offset for next
+ */
+ if (NULL != (next = strchr(portlist_copy, (int) ',')))
+ {
+ *next++ = '\0';
+ }
+
+ /*
+ * Loop through all items, checking for match
+ */
+ while(min)
+ {
+ if (NULL == (max = strchr(min, (int) '-')))
+ {
+ /*
+ * No dash, check for equality
+ */
+ if (port == atoi(min))
+ {
+ free(portlist_copy);
+ return(1);
+ }
+ }
+ else
+ {
+ /*
+ * This is a range, so check if between min and max,
+ * or, if max was omitted, between min and 65K
+ */
+ *max++ = '\0';
+ if(port >= atoi(min) && port <= (atoi(max) ? atoi(max) : 65535))
+ {
+ free(portlist_copy);
+ return(1);
+ }
+
+ }
+
+ /*
+ * Jump to next item
+ */
+ min = next;
+
+ /*
+ * Zero-terminate next item and remember offset for n+1
+ */
+ if ((NULL != next) && (NULL != (next = strchr(next, (int) ','))))
+ {
+ *next++ = '\0';
+ }
+ }
+
+ free(portlist_copy);
+ return 0;
+
+}
/*********************************************************************
*********************************************************************/
struct http_response *block_url(struct client_state *csp)
{
- char *p;
struct http_response *rsp;
- struct map *exports = NULL;
+ const char *new_content_type = NULL;
- /*
+ /*
* If it's not blocked, don't block it ;-)
*/
if ((csp->action->flags & ACTION_BLOCK) == 0)
{
- return(NULL);
+ return NULL;
}
-
- /*
+ if (csp->action->flags & ACTION_REDIRECT)
+ {
+ log_error(LOG_LEVEL_ERROR, "redirect{} overruled by block.");
+ }
+ /*
* Else, prepare a response
*/
- if (NULL == ( rsp = (struct http_response *)zalloc(sizeof(*rsp))))
+ if (NULL == (rsp = alloc_http_response()))
{
- return NULL;
+ return cgi_error_memory();
}
/*
* If it's an image-url, send back an image or redirect
* as specified by the relevant +image action
*/
-#ifdef IMAGE_BLOCKING
+#ifdef FEATURE_IMAGE_BLOCKING
if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
&& is_imageurl(csp))
{
+ char *p;
/* determine HOW images should be blocked */
p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
+ 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, "logo")))
+ if ((p == NULL) || (0 == strcmpic(p, "pattern")))
{
- rsp->body = bindup(JBGIF, sizeof(JBGIF));
- rsp->content_length = sizeof(JBGIF);
- enlist_unique_header(rsp->headers, "Content-Type", "image/gif");
+ rsp->status = strdup("403 Request blocked by Privoxy");
+ if (rsp->status == NULL)
+ {
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
+ rsp->body = bindup(image_pattern_data, image_pattern_length);
+ if (rsp->body == NULL)
+ {
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
+ rsp->content_length = image_pattern_length;
+
+ if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
+ {
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
}
else if (0 == strcmpic(p, "blank"))
{
- rsp->body = bindup(BLANKGIF, sizeof(BLANKGIF));
- rsp->content_length = sizeof(BLANKGIF);
- enlist_unique_header(rsp->headers, "Content-Type", "image/gif");
+ rsp->status = strdup("403 Request blocked by Privoxy");
+ if (rsp->status == NULL)
+ {
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
+ rsp->body = bindup(image_blank_data, image_blank_length);
+ if (rsp->body == NULL)
+ {
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
+ rsp->content_length = image_blank_length;
+
+ if (enlist_unique_header(rsp->headers, "Content-Type", BUILTIN_IMAGE_MIMETYPE))
+ {
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
}
else
{
- rsp->status = strdup("302 Local Redirect from Junkbuster");
- enlist_unique_header(rsp->headers, "Location", p);
+ 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();
+ }
+ }
+
+#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)
+ {
+ /*
+ * Send empty document.
+ */
+ new_content_type = csp->action->string[ACTION_STRING_CONTENT_TYPE];
+
+ freez(rsp->body);
+ rsp->body = strdup(" ");
+ rsp->content_length = 1;
+
+ rsp->status = strdup("403 Request blocked by Privoxy");
+ if (rsp->status == NULL)
+ {
+ free_http_response(rsp);
+ return cgi_error_memory();
}
- }
+ if (new_content_type != 0)
+ {
+ log_error(LOG_LEVEL_HEADER, "Overwriting Content-Type with %s", new_content_type);
+ if (enlist_unique_header(rsp->headers, "Content-Type", new_content_type))
+ {
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
+ }
+
+ }
else
-#endif /* def IMAGE_BLOCKING */
+#endif /* def FEATURE_IMAGE_BLOCKING */
- /*
+ /*
* Else, generate an HTML "blocked" message:
*/
{
+ jb_err err;
+ struct map * exports;
+ char *p;
- exports = default_exports(csp, NULL);
-#ifdef FORCE_LOAD
- exports = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
-#else
- exports = map_block_killer(exports, "force-support");
-#endif /* ndef FORCE_LOAD */
+ /*
+ * 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("404 Request for blocked URL");
+ }
- exports = map(exports, "hostport", 1, csp->http->hostport, 1);
- exports = map(exports, "hostport-html", 1, html_encode(csp->http->hostport), 0);
- exports = map(exports, "path", 1, csp->http->path, 1);
- exports = map(exports, "path-html", 1, html_encode(csp->http->path), 0);
+ if (rsp->status == NULL)
+ {
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
- rsp->body = fill_template(csp, "blocked", exports);
- free_map(exports);
-
- rsp->status = strdup("403 Request for blocked URL");
+ exports = default_exports(csp, NULL);
+ if (exports == NULL)
+ {
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
+
+#ifdef FEATURE_FORCE_LOAD
+ err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
+ if (csp->http->ssl != 0)
+#endif /* ndef FEATURE_FORCE_LOAD */
+ {
+ err = map_block_killer(exports, "force-support");
+ }
+
+ if (!err) err = map(exports, "protocol", 1, csp->http->ssl ? "https://" : "http://", 1);
+ 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)
+ {
+ free_map(exports);
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
+
+ err = template_fill_for_cgi(csp, "blocked", exports, rsp);
+ if (err)
+ {
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
}
- return(finish_http_response(rsp));
+ return finish_http_response(rsp);
}
-#ifdef TRUST_FILES
+#ifdef FEATURE_TRUST
/*********************************************************************
*
* Function : trust_url FIXME: I should be called distrust_url
*
* Description : Calls is_untrusted_url to determine if the URL is trusted
- * and if not, returns a HTTP 304 response with a reject message.
+ * and if not, returns a HTTP 403 response with a reject message.
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
struct http_response *trust_url(struct client_state *csp)
{
struct http_response *rsp;
- struct map *exports = NULL;
- char buf[BUFFER_SIZE], *p = NULL;
- struct url_spec **tl, *t;
+ struct map * exports;
+ char buf[BUFFER_SIZE];
+ char *p;
+ struct url_spec **tl;
+ struct url_spec *t;
+ jb_err err;
/*
* Don't bother to work on trusted URLs
return NULL;
}
- /*
+ /*
* Else, prepare a response:
*/
- if (NULL == ( rsp = (struct http_response *)zalloc(sizeof(*rsp))))
+ if (NULL == (rsp = alloc_http_response()))
{
- return NULL;
+ return cgi_error_memory();
}
+
+ rsp->status = strdup("403 Request blocked by Privoxy");
exports = default_exports(csp, NULL);
+ if (exports == NULL || rsp->status == NULL)
+ {
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
- /*
- * Export the host, port, and referrer information
+ /*
+ * Export the protocol, host, port, and referrer information
*/
- exports = map(exports, "hostport", 1, csp->http->hostport, 1);
- exports = map(exports, "path", 1, csp->http->path, 1);
- exports = map(exports, "hostport-html", 1, html_encode(csp->http->hostport), 0);
- exports = map(exports, "path-html", 1, html_encode(csp->http->path), 0);
+ 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, "path", 1, csp->http->path, 1);
- if (csp->referrer && strlen(csp->referrer) > 9)
+ if (NULL != (p = get_header_value(csp->headers, "Referer:")))
{
- exports = map(exports, "referrer", 1, csp->referrer + 9, 1);
- exports = map(exports, "referrer-html", 1, html_encode(csp->referrer + 9), 0);
+ if (!err) err = map(exports, "referrer", 1, html_encode(p), 0);
}
else
{
- exports = map(exports, "referrer", 1, "unknown", 1);
- exports = map(exports, "referrer-html", 1, "unknown", 1);
+ if (!err) err = map(exports, "referrer", 1, "none set", 1);
+ }
+
+ if (err)
+ {
+ free_map(exports);
+ free_http_response(rsp);
+ return cgi_error_memory();
}
/*
* Export the trust list
*/
- for (tl = csp->config->trust_list; (t = *tl) ; tl++)
+ p = strdup("");
+ for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
{
sprintf(buf, "<li>%s</li>\n", t->spec);
- p = strsav(p, buf);
+ string_append(&p, buf);
+ }
+ err = map(exports, "trusted-referrers", 1, p, 0);
+
+ if (err)
+ {
+ free_map(exports);
+ free_http_response(rsp);
+ return cgi_error_memory();
}
- exports = map(exports, "trusted-referrers", 1, p, 0);
- p = NULL;
/*
* Export the trust info, if available
*/
- if (csp->config->trust_info->next)
+ if (csp->config->trust_info->first)
{
- struct list *l;
+ struct list_entry *l;
- for (l = csp->config->trust_info->next; l ; l = l->next)
+ p = strdup("");
+ for (l = csp->config->trust_info->first; l ; l = l->next)
{
- sprintf(buf, "<li> <a href=%s>%s</a><br>\n",l->str, l->str);
- p = strsav(p, buf);
+ sprintf(buf, "<li> <a href=\"%s\">%s</a><br>\n",l->str, l->str);
+ string_append(&p, buf);
}
- exports = map(exports, "trust-info", 1, p, 0);
+ err = map(exports, "trust-info", 1, p, 0);
}
else
{
- exports = map_block_killer(exports, "have-trust-info");
+ err = map_block_killer(exports, "have-trust-info");
}
-
+
+ if (err)
+ {
+ free_map(exports);
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
+
/*
* Export the force prefix or the force conditional block killer
*/
-#ifdef FORCE_LOAD
- exports = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
-#else
- exports = map_block_killer(exports, "force-support");
-#endif /* ndef FORCE_LOAD */
+#ifdef FEATURE_FORCE_LOAD
+ err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
+#else /* ifndef FEATURE_FORCE_LOAD */
+ err = map_block_killer(exports, "force-support");
+#endif /* ndef FEATURE_FORCE_LOAD */
+
+ if (err)
+ {
+ free_map(exports);
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
/*
* Build the response
*/
- rsp->body = fill_template(csp, "untrusted", exports);
- free_map(exports);
-
- return(finish_http_response(rsp));
+ err = template_fill_for_cgi(csp, "untrusted", exports, rsp);
+ if (err)
+ {
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
+ return finish_http_response(rsp);
}
-#endif /* def TRUST_FILES */
-
+#endif /* def FEATURE_TRUST */
-#ifdef FAST_REDIRECTS
/*********************************************************************
*
- * Function : redirect_url
+ * Function : execute_single_pcrs_command
*
- * Description : Checks for redirection URLs and returns a HTTP redirect
- * to the destination URL, if necessary
+ * Description : Apply single pcrs command to the subject.
+ * The subject itself is left untouched, memory for the result
+ * is malloc()ed and it is the caller's responsibility to free
+ * the result when it's no longer needed.
*
* Parameters :
- * 1 : csp = Current client state (buffers, headers, etc...)
+ * 1 : subject = the subject (== original) string
+ * 2 : pcrs_command = the pcrs command as string (s@foo@bar@)
+ * 3 : hits = int* for returning the number of modifications
*
- * Returns : NULL if URL was clean, HTTP redirect otherwise.
+ * Returns : NULL in case of errors, otherwise the
+ * result of the pcrs command.
*
*********************************************************************/
-struct http_response *redirect_url(struct client_state *csp)
+char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
{
- char *p, *q;
- struct http_response *rsp;
+ int error;
+ size_t size;
+ char *result = NULL;
+ pcrs_job *job;
- p = q = csp->http->path;
- log_error(LOG_LEVEL_REDIRECTS, "checking path for redirects: %s", p);
+ assert(subject);
+ assert(pcrs_command);
- /*
- * find the last URL encoded in the request
- */
- while (p = strstr(p, "http://"))
+ *hits = 0;
+ size = strlen(subject);
+
+ if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
{
- q = p++;
+ log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
+ pcrs_command, error);
}
-
- /*
- * if there was any, generate and return a HTTP redirect
- */
- if (q != csp->http->path)
+ else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
{
- log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
-
- if (NULL == ( rsp = zalloc(sizeof(*rsp))))
- {
- return NULL;
- }
-
- rsp->status = strdup("302 Local Redirect from Junkbuster");
- enlist_unique_header(rsp->headers, "Location", q);
-
- return(finish_http_response(rsp));
+ log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
+ *hits = 0;
+ freez(result);
}
- else
+
+ if (job)
{
- return(NULL);
+ job = pcrs_free_job(job);
}
-}
-#endif /* def FAST_REDIRECTS */
+ return result;
+}
-#ifdef IMAGE_BLOCKING
/*********************************************************************
*
- * Function : is_imageurl
+ * Function : rewrite_url
+ *
+ * Description : Rewrites a URL with a single pcrs command
+ * and returns the result if it differs from the
+ * original and isn't obviously invalid.
*
- * Description : Given a URL, decide whether it is an image or not,
- * using either the info from a previous +image action
- * or, #ifdef DETECT_MSIE_IMAGES, the info from the
- * browser's accept header.
- *
* Parameters :
- * 1 : csp = Current client state (buffers, headers, etc...)
+ * 1 : old_url = URL to rewrite.
+ * 2 : pcrs_command = pcrs command formatted as string (s@foo@bar@)
+ *
+ *
+ * Returns : NULL if the pcrs_command didn't change the url, or
+ * the result of the modification.
+ *
+ *********************************************************************/
+char *rewrite_url(char *old_url, const char *pcrs_command)
+{
+ char *new_url = NULL;
+ int hits;
+
+ assert(old_url);
+ assert(pcrs_command);
+
+ new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
+
+ if (hits == 0)
+ {
+ log_error(LOG_LEVEL_REDIRECTS,
+ "pcrs command \"%s\" didn't change \"%s\".",
+ pcrs_command, old_url, new_url);
+ freez(new_url);
+ }
+ else if (strncmpic(new_url, "http://", 7) && strncmpic(new_url, "https://", 8))
+ {
+ log_error(LOG_LEVEL_ERROR,
+ "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s), "
+ "but the result doesn't look like a valid URL and will be ignored.",
+ pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
+ freez(new_url);
+ }
+ else
+ {
+ log_error(LOG_LEVEL_REDIRECTS,
+ "pcrs command \"%s\" changed \"%s\" to \"%s\" (%u hi%s).",
+ pcrs_command, old_url, new_url, hits, (hits == 1) ? "t" : "ts");
+ }
+
+ return new_url;
+
+}
+
+
+#ifdef FEATURE_FAST_REDIRECTS
+/*********************************************************************
+ *
+ * Function : get_last_url
+ *
+ * Description : Search for the last URL inside a string.
+ * If the string already is a URL, it will
+ * be the first URL found.
+ *
+ * Parameters :
+ * 1 : subject = the string to check
+ * 2 : redirect_mode = +fast-redirect{} mode
+ *
+ * Returns : NULL if no URL was found, or
+ * the last URL found.
+ *
+ *********************************************************************/
+char *get_last_url(char *subject, const char *redirect_mode)
+{
+ char *new_url = NULL;
+ char *tmp;
+
+ assert(subject);
+ assert(redirect_mode);
+
+ subject = strdup(subject);
+ if (subject == NULL)
+ {
+ log_error(LOG_LEVEL_ERROR, "Out of memory while searching for redirects.");
+ 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)
+ {
+ freez(subject);
+ subject = new_url;
+ }
+ else
+ {
+ log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
+ }
+ }
+
+ log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
+
+ /*
+ * Find the last URL encoded in the request
+ */
+ tmp = subject;
+ while ((tmp = strstr(tmp, "http://")) != NULL)
+ {
+ new_url = tmp++;
+ }
+ tmp = (new_url != NULL) ? new_url : subject;
+ while ((tmp = strstr(tmp, "https://")) != NULL)
+ {
+ new_url = tmp++;
+ }
+
+ if ((new_url != NULL)
+ && ( (new_url != subject)
+ || (0 == strncmpic(subject, "http://", 7))
+ || (0 == strncmpic(subject, "https://", 8))
+ ))
+ {
+ /*
+ * Return new URL if we found a redirect
+ * or if the subject already was a URL.
+ *
+ * The second case makes sure that we can
+ * chain get_last_url after another redirection check
+ * (like rewrite_url) without losing earlier redirects.
+ */
+ new_url = strdup(new_url);
+ freez(subject);
+ return new_url;
+ }
+
+ freez(subject);
+ return NULL;
+
+}
+#endif /* def FEATURE_FAST_REDIRECTS */
+
+
+/*********************************************************************
+ *
+ * Function : redirect_url
+ *
+ * Description : Checks if Privoxy should answer the request with
+ * a HTTP redirect and generates the redirect if
+ * necessary.
+ *
+ * Parameters :
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ *
+ * Returns : NULL if the request can pass, HTTP redirect otherwise.
+ *
+ *********************************************************************/
+struct http_response *redirect_url(struct client_state *csp)
+{
+ struct http_response *rsp;
+#ifdef FEATURE_FAST_REDIRECTS
+ /*
+ * XXX: Do we still need FEATURE_FAST_REDIRECTS
+ * as compile-time option? The user can easily disable
+ * it in his action file.
+ */
+ char * redirect_mode;
+#endif /* def FEATURE_FAST_REDIRECTS */
+ char *old_url = NULL;
+ char *new_url = NULL;
+ char *redirection_string;
+
+ if ((csp->action->flags & ACTION_REDIRECT))
+ {
+ redirection_string = csp->action->string[ACTION_STRING_REDIRECT];
+
+ /*
+ * If the redirection string begins with 's',
+ * assume it's a pcrs command, otherwise treat it as
+ * 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}
+ * she would get undefined results anyway.
+ *
+ */
+
+ if (*redirection_string == 's')
+ {
+ old_url = csp->http->url;
+ new_url = rewrite_url(old_url, redirection_string);
+ }
+ else
+ {
+ log_error(LOG_LEVEL_REDIRECTS,
+ "No pcrs command recognized, assuming that \"%s\" is already properly formatted.",
+ redirection_string);
+ new_url = strdup(redirection_string);
+ }
+ }
+
+#ifdef FEATURE_FAST_REDIRECTS
+ if ((csp->action->flags & ACTION_FAST_REDIRECTS))
+ {
+ redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
+
+ /*
+ * If it exists, use the previously rewritten URL as input
+ * otherwise just use the old path.
+ */
+ old_url = new_url ? new_url : strdup(csp->http->path);
+ new_url = get_last_url(old_url, redirect_mode);
+ freez(old_url);
+ }
+#endif /* def FEATURE_FAST_REDIRECTS */
+
+ /* Did any redirect action trigger? */
+ if (new_url)
+ {
+ if (0 == strcmpic(new_url, csp->http->url))
+ {
+ log_error(LOG_LEVEL_ERROR,
+ "New URL \"%s\" and old URL \"%s\" are the same. Redirection loop prevented.",
+ csp->http->url, new_url);
+ freez(new_url);
+ }
+ else
+ {
+ log_error(LOG_LEVEL_REDIRECTS, "New URL is: %s", new_url);
+
+ if (NULL == (rsp = alloc_http_response()))
+ {
+ freez(new_url);
+ return cgi_error_memory();
+ }
+
+ if ( enlist_unique_header(rsp->headers, "Location", new_url)
+ || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
+ {
+ freez(new_url);
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
+ freez(new_url);
+ return finish_http_response(rsp);
+ }
+ }
+
+ /* Only reached if no redirect is required */
+ return NULL;
+
+}
+
+
+#ifdef FEATURE_IMAGE_BLOCKING
+/*********************************************************************
+ *
+ * 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.
+ *
+ * Parameters :
+ * 1 : csp = Current client state (buffers, headers, etc...)
*
* Returns : True (nonzero) if URL is an image, false (0)
* otherwise
*********************************************************************/
int is_imageurl(struct client_state *csp)
{
-#ifdef DETECT_MSIE_IMAGES
- if ((csp->accept_types
- & (ACCEPT_TYPE_IS_MSIE|ACCEPT_TYPE_MSIE_IMAGE|ACCEPT_TYPE_MSIE_HTML))
- == (ACCEPT_TYPE_IS_MSIE|ACCEPT_TYPE_MSIE_IMAGE))
- {
- return 1;
- }
- else if ((csp->accept_types
- & (ACCEPT_TYPE_IS_MSIE|ACCEPT_TYPE_MSIE_IMAGE|ACCEPT_TYPE_MSIE_HTML))
- == (ACCEPT_TYPE_IS_MSIE|ACCEPT_TYPE_MSIE_HTML))
+#ifdef FEATURE_IMAGE_DETECT_MSIE
+ char *tmp;
+
+ tmp = get_header_value(csp->headers, "User-Agent:");
+ if (tmp && strstr(tmp, "MSIE") && !strstr(tmp, "Mac_"))
{
- return 0;
+ 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
+#endif /* def FEATURE_IMAGE_DETECT_MSIE */
return ((csp->action->flags & ACTION_IMAGE) != 0);
}
-#endif /* def IMAGE_BLOCKING */
+#endif /* def FEATURE_IMAGE_BLOCKING */
-#ifdef TRUST_FILES
+#ifdef FEATURE_TRUST
/*********************************************************************
*
* Function : is_untrusted_url
{
struct file_list *fl;
struct block_spec *b;
- struct url_spec url[1], **tl, *t;
+ struct url_spec **trusted_url;
struct http_request rhttp[1];
- char *p, *h;
+ const char * referer;
+ jb_err err;
/*
* If we don't have a trustlist, we trust everybody
*/
if (((fl = csp->tlist) == NULL) || ((b = fl->f) == NULL))
{
- return(0);
+ return 0;
}
+ memset(rhttp, '\0', sizeof(*rhttp));
/*
* Do we trust the request URL itself?
*/
- *url = dsplit(csp->http->host);
-
- /* if splitting the domain fails, punt */
- if (url->dbuf == NULL) return(0);
-
- memset(rhttp, '\0', sizeof(*rhttp));
-
for (b = b->next; b ; b = b->next)
{
- if ((b->url->port == 0) || (b->url->port == csp->http->port))
+ if (url_match(b->url, csp->http))
{
- if ((b->url->domain[0] == '\0') || (domaincmp(b->url, url) == 0))
- {
- if ((b->url->path == NULL) ||
-#ifdef REGEX
- (regexec(b->url->preg, csp->http->path, 0, NULL, 0) == 0)
-#else
- (strncmp(b->url->path, csp->http->path, b->url->pathlen) == 0)
-#endif
- )
- {
- freez(url->dbuf);
- freez(url->dvec);
-
- if (b->reject == 0) return(0);
-
- return(1);
- }
- }
+ return b->reject;
}
}
- freez(url->dbuf);
- freez(url->dvec);
-
- if ((csp->referrer == NULL)|| (strlen(csp->referrer) <= 9))
+ if (NULL == (referer = get_header_value(csp->headers, "Referer:")))
{
/* no referrer was supplied */
- return(1);
- }
-
- /* forge a URL from the referrer so we can use
- * convert_url() to parse it into its components.
- */
-
- p = NULL;
- p = strsav(p, "GET ");
- p = strsav(p, csp->referrer + 9); /* skip over "Referer: " */
- p = strsav(p, " HTTP/1.0");
-
- parse_http_request(p, rhttp, csp);
- freez(p);
-
- if (rhttp->cmd == NULL)
- {
- return(1);
+ return 1;
}
/*
* If not, do we maybe trust its referrer?
*/
- *url = dsplit(rhttp->host);
-
- /* if splitting the domain fails, punt */
- if (url->dbuf == NULL) return(1);
+ err = parse_http_url(referer, rhttp, csp);
+ if (err)
+ {
+ return 1;
+ }
- for (tl = csp->config->trust_list; (t = *tl) ; tl++)
+ for (trusted_url = csp->config->trust_list; *trusted_url != NULL; trusted_url++)
{
- if ((t->port == 0) || (t->port == rhttp->port))
+ if (url_match(*trusted_url, rhttp))
{
- if ((t->domain[0] == '\0') || domaincmp(t, url) == 0)
- {
- if ((t->path == NULL) ||
-#ifdef REGEX
- (regexec(t->preg, rhttp->path, 0, NULL, 0) == 0)
-#else
- (strncmp(t->path, rhttp->path, t->pathlen) == 0)
-#endif
- )
- {
- /* if the URL's referrer is from a trusted referrer, then
- * add the target spec to the trustfile as an unblocked
- * domain and return NULL (which means it's OK).
- */
+ /* if the URL's referrer is from a trusted referrer, then
+ * add the target spec to the trustfile as an unblocked
+ * domain and return NULL (which means it's OK).
+ */
+
+ FILE *fp;
- FILE *fp;
+ if (NULL != (fp = fopen(csp->config->trustfile, "a")))
+ {
+ char * path;
+ char * path_end;
+ char * new_entry = strdup("~");
- freez(url->dbuf);
- freez(url->dvec);
+ string_append(&new_entry, csp->http->hostport);
- if ((fp = fopen(csp->config->trustfile, "a")))
+ path = csp->http->path;
+ if ( (path[0] == '/')
+ && (path[1] == '~')
+ && ((path_end = strchr(path + 2, '/')) != NULL))
+ {
+ /* 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 */
+ path = strdup(path); /* Copy string */
+ if (path != NULL)
{
- h = NULL;
-
- h = strsav(h, "~");
- h = strsav(h, csp->http->hostport);
-
- p = csp->http->path;
- if ((*p++ == '/')
- && (*p++ == '~'))
- {
- /* since this path points into a user's home space
- * be sure to include this spec in the trustfile.
- */
- if ((p = strchr(p, '/')))
- {
- *p = '\0';
- h = strsav(h, csp->http->path); /* FIXME: p?! */
- h = strsav(h, "/");
- }
- }
-
- fprintf(fp, "%s\n", h);
- freez(h);
- fclose(fp);
+ path_end = path + path_len; /* regenerate ptr to new buffer */
+ *(path_end + 1) = '\0'; /* Truncate path after '/' */
}
- return(0);
+ string_join(&new_entry, path);
}
+
+ if (new_entry != NULL)
+ {
+ fprintf(fp, "%s\n", new_entry);
+ free(new_entry);
+ }
+ else
+ {
+ /* FIXME: No way to handle out-of memory, so mostly ignoring it */
+ log_error(LOG_LEVEL_ERROR, "Out of memory adding pattern to trust file");
+ }
+
+ fclose(fp);
}
+ return 0;
}
}
- return(1);
+ return 1;
}
-#endif /* def TRUST_FILES */
+#endif /* def FEATURE_TRUST */
/*********************************************************************
*
* Function : pcrs_filter_response
*
- * Description : Apply all the pcrs jobs from the joblist (re_filterfile)
- * to the text buffer that's been accumulated in
- * csp->iob->buf and set csp->content_length to the modified
- * size.
+ * Description : Execute all text substitutions from all applying
+ * +filter actions on the text buffer that's been accumulated
+ * in csp->iob->buf. If this changes the contents, 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.
- * or NULL in case something went wrong
- *
+ * or NULL if there were no hits or something went wrong
+ *
*********************************************************************/
char *pcrs_filter_response(struct client_state *csp)
{
int hits=0;
- int size = csp->iob->eod - csp->iob->cur;
+ size_t size, prev_size;
char *old = csp->iob->cur, *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 (size <= 0)
+ /*
+ * Sanity first
+ */
+ if (csp->iob->cur >= csp->iob->eod)
{
return(NULL);
}
+ size = csp->iob->eod - csp->iob->cur;
- if ( ( NULL == (fl = csp->rlist) ) || ( NULL == (b = fl->f) ) )
+ /*
+ * 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)
{
log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
- return(NULL);
+ return(NULL);
}
- log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) ...",
- csp->http->hostport, csp->http->path, size);
+ /*
+ * If the body has a "chunked" transfer-encoding,
+ * get rid of it first, adjusting size and iob->eod
+ */
+ if (csp->flags & CSP_FLAG_CHUNKED)
+ {
+ log_error(LOG_LEVEL_RE_FILTER, "Need to de-chunk first");
+ if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
+ {
+ return(NULL);
+ }
+ csp->iob->eod = csp->iob->cur + size;
+ csp->flags |= CSP_FLAG_MODIFIED;
+ }
- /* Apply all jobs from the joblist */
- for (job = b->joblist; NULL != job; job = job->next)
+ for (i = 0; i < MAX_AF_FILES; i++)
{
- hits += pcrs_execute(job, old, size, &new, &size);
- if (old != csp->iob->cur) free(old);
- old=new;
+ fl = csp->rlist[i];
+ if ((NULL == fl) || (NULL == fl->f))
+ break;
+ /*
+ * 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 ; filtername = filtername->next)
+ {
+ if (strcmp(b->name, filtername->str) == 0)
+ {
+ int current_hits = 0;
+
+ if ( NULL == b->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 = b->joblist; NULL != job; job = job->next)
+ {
+ current_hits += pcrs_execute(job, old, size, &new, &size);
+ if (old != csp->iob->cur) free(old);
+ old=new;
+ }
+
+ log_error(LOG_LEVEL_RE_FILTER, "re_filtering %s%s (size %d) with filter %s produced %d hits (new size %d).",
+ csp->http->hostport, csp->http->path, prev_size, b->name, current_hits, size);
+
+ hits += current_hits;
+ }
+ }
+ }
}
- log_error(LOG_LEVEL_RE_FILTER, " produced %d hits (new size %d).", hits, size);
+ /*
+ * If there were no hits, destroy our copy and let
+ * chat() use the original in csp->iob
+ */
+ if (!hits)
+ {
+ free(new);
+ return(NULL);
+ }
+ csp->flags |= CSP_FLAG_MODIFIED;
csp->content_length = size;
-
- /* fwiw, reset the iob */
IOB_RESET(csp);
+
return(new);
}
*
* Function : gif_deanimate_response
*
- * Description : Deanimate the GIF image that has been accumulated in
- * csp->iob->buf and set csp->content_length to the modified
- * size.
+ * 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.
* or NULL in case something went wrong.
- *
+ *
*********************************************************************/
char *gif_deanimate_response(struct client_state *csp)
{
struct binbuffer *in, *out;
char *p;
- int size = csp->iob->eod - csp->iob->cur;
+ size_t size = csp->iob->eod - csp->iob->cur;
+
+ /*
+ * If the body has a "chunked" transfer-encoding,
+ * get rid of it first, adjusting size and iob->eod
+ */
+ if (csp->flags & CSP_FLAG_CHUNKED)
+ {
+ log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
+ if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
+ {
+ return(NULL);
+ }
+ csp->iob->eod = csp->iob->cur + size;
+ csp->flags |= CSP_FLAG_MODIFIED;
+ }
if ( (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
|| (NULL == (out = (struct binbuffer *)zalloc(sizeof *out))) )
{
- log_error(LOG_LEVEL_DEANIMATE, "failed! (No Mem!)");
+ log_error(LOG_LEVEL_DEANIMATE, "failed! (no mem)");
return NULL;
}
in->buffer = csp->iob->cur;
in->size = size;
- if (gif_deanimate(in, out))
+ if (gif_deanimate(in, out, strncmp("last", csp->action->string[ACTION_STRING_DEANIMATE], 4)))
{
- log_error(LOG_LEVEL_DEANIMATE, "failed! (size %d)", size);
+ log_error(LOG_LEVEL_DEANIMATE, "failed! (gif parsing)");
free(in);
buf_free(out);
return(NULL);
}
else
{
- log_error(LOG_LEVEL_DEANIMATE, "Success! GIF shrunk from %d bytes to %d.", size, out->offset);
+ 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;
free(in);
free(out);
return(p);
- }
+ }
}
/*********************************************************************
*
- * Function : url_actions
+ * Function : jpeg_inspect_response
*
- * Description : Gets the actions for this URL.
+ * Description :
*
* 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...)
*
- * Returns : N/A
+ * Returns : a pointer to the (newly allocated) modified buffer
+ * or NULL in case something went wrong.
*
*********************************************************************/
-void url_actions(struct http_request *http,
- struct client_state *csp)
+char *jpeg_inspect_response(struct client_state *csp)
{
- struct file_list *fl;
- struct url_actions *b;
+ struct binbuffer *in = NULL, *out = NULL;
+ char *p = NULL;
+ size_t size = csp->iob->eod - csp->iob->cur;
- init_current_action(csp->action);
+ /*
+ * If the body has a "chunked" transfer-encoding,
+ * get rid of it first, adjusting size and iob->eod
+ */
+ if (csp->flags & CSP_FLAG_CHUNKED)
+ {
+ log_error(LOG_LEVEL_DEANIMATE, "Need to de-chunk first");
+ if (0 == (size = remove_chunked_transfer_coding(csp->iob->cur, size)))
+ {
+ return(NULL);
+ }
+ csp->iob->eod = csp->iob->cur + size;
+ csp->flags |= CSP_FLAG_MODIFIED;
+ }
- if (((fl = csp->actions_list) == NULL) || ((b = fl->f) == NULL))
+ if (NULL == (in = (struct binbuffer *)zalloc(sizeof *in )))
{
- return;
+ log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 1)");
+ return NULL;
}
- apply_url_actions(csp->action, http, b);
+ if (NULL == (out = (struct binbuffer *)zalloc(sizeof *out)))
+ {
+ log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg no mem 2)");
+ return NULL;
+ }
+
+ 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))
+ {
+ log_error(LOG_LEVEL_DEANIMATE, "failed! (jpeg parsing)");
+ free(in);
+ buf_free(out);
+ return(NULL);
+
+ }
+ else
+ {
+ csp->content_length = out->offset;
+ csp->flags |= CSP_FLAG_MODIFIED;
+ p = out->buffer;
+ free(in);
+ free(out);
+ return(p);
+ }
}
/*********************************************************************
*
- * Function : apply_url_actions
+ * Function : remove_chunked_transfer_coding
*
- * Description : Applies a list of URL actions.
+ * Description : In-situ remove the "chunked" transfer coding as defined
+ * in rfc2616 from a buffer.
*
* Parameters :
- * 1 : action = Destination.
- * 2 : http = Current URL
- * 3 : b = list of URL actions to apply
+ * 1 : buffer = Pointer to the text buffer
+ * 2 : size = Number of bytes to be processed
*
- * Returns : N/A
+ * Returns : The new size, i.e. the number of bytes from buffer which
+ * are occupied by the stripped body, or 0 in case something
+ * went wrong
*
*********************************************************************/
-void apply_url_actions(struct current_action_spec *action,
- struct http_request *http,
- struct url_actions *b)
+int remove_chunked_transfer_coding(char *buffer, const size_t size)
{
- struct url_spec url[1];
-
- if (b == NULL)
- {
- /* Should never happen */
- return;
- }
+ size_t newsize = 0;
+ unsigned int chunksize = 0;
+ char *from_p, *to_p;
- *url = dsplit(http->host);
+ assert(buffer);
+ from_p = to_p = buffer;
- /* if splitting the domain fails, punt */
- if (url->dbuf == NULL)
+ if (sscanf(buffer, "%x", &chunksize) != 1)
{
- return;
+ log_error(LOG_LEVEL_ERROR, "Invalid first chunksize while stripping \"chunked\" transfer coding");
+ return(0);
}
- for (b = b->next; NULL != b; b = b->next)
+ while (chunksize > 0)
{
- if ((b->url->port == 0) || (b->url->port == http->port))
+ if (NULL == (from_p = strstr(from_p, "\r\n")))
{
- if ((b->url->domain[0] == '\0') || (domaincmp(b->url, url) == 0))
- {
- if ((b->url->path == NULL) ||
-#ifdef REGEX
- (regexec(b->url->preg, http->path, 0, NULL, 0) == 0)
-#else
- (strncmp(b->url->path, http->path, b->url->pathlen) == 0)
-#endif
- )
- {
- merge_current_action(action, b->action);
- }
- }
+ log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
+ return(0);
+ }
+
+ if ((newsize += chunksize) >= size)
+ {
+ log_error(LOG_LEVEL_ERROR, "Chunksize exceeds buffer in \"chunked\" transfer coding");
+ return(0);
+ }
+ from_p += 2;
+
+ memmove(to_p, from_p, (size_t) chunksize);
+ to_p = buffer + newsize;
+ from_p += chunksize + 2;
+
+ if (sscanf(from_p, "%x", &chunksize) != 1)
+ {
+ log_error(LOG_LEVEL_ERROR, "Parse error while stripping \"chunked\" transfer coding");
+ return(0);
}
}
- freez(url->dbuf);
- freez(url->dvec);
+ /* FIXME: Should this get its own loglevel? */
+ log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
+ return(newsize);
+
}
/*********************************************************************
*
- * Function : forward_url
+ * Function : url_actions
*
- * Description : Should we forward this to another proxy?
+ * Description : Gets the actions for this URL.
*
* Parameters :
- * 1 : http = http_request request for current URL
+ * 1 : http = http_request request for blocked URLs
* 2 : csp = Current client state (buffers, headers, etc...)
*
- * Returns : Pointer to forwarding information.
+ * Returns : N/A
*
*********************************************************************/
-const struct forward_spec * forward_url(struct http_request *http,
- struct client_state *csp)
+void url_actions(struct http_request *http,
+ struct client_state *csp)
{
- static const struct forward_spec fwd_default[1] = { 0 }; /* All zeroes */
- struct forward_spec *fwd = csp->config->forward;
- struct url_spec url[1];
-
- if (fwd == NULL)
- {
- return(fwd_default);
- }
-
- *url = dsplit(http->host);
+ struct file_list *fl;
+ struct url_actions *b;
+ int i;
- /* if splitting the domain fails, punt */
- if (url->dbuf == NULL)
- {
- return(fwd_default);
- }
+ init_current_action(csp->action);
- while (fwd != NULL)
+ for (i = 0; i < MAX_AF_FILES; i++)
{
- if ((fwd->url->port == 0) || (fwd->url->port == http->port))
+ if (((fl = csp->actions_list[i]) == NULL) || ((b = fl->f) == NULL))
{
- if ((fwd->url->domain[0] == '\0') || (domaincmp(fwd->url, url) == 0))
- {
- if ((fwd->url->path == NULL) ||
-#ifdef REGEX
- (regexec(fwd->url->preg, http->path, 0, NULL, 0) == 0)
-#else
- (strncmp(fwd->url->path, http->path, fwd->url->pathlen) == 0)
-#endif
- )
- {
- freez(url->dbuf);
- freez(url->dvec);
- return(fwd);
- }
- }
+ return;
}
- fwd = fwd->next;
- }
- freez(url->dbuf);
- freez(url->dvec);
- return(fwd_default);
+ apply_url_actions(csp->action, http, b);
+ }
+ return;
}
/*********************************************************************
*
- * Function : dsplit
+ * Function : apply_url_actions
*
- * Description : Takes a domain and returns a pointer to a url_spec
- * structure populated with dbuf, dcnt and dvec. The
- * other fields in the structure that is returned are zero.
+ * Description : Applies a list of URL actions.
*
* Parameters :
- * 1 : domain = a URL address
+ * 1 : action = Destination.
+ * 2 : http = Current URL
+ * 3 : b = list of URL actions to apply
*
- * Returns : url_spec structure populated with dbuf, dcnt and dvec.
+ * Returns : N/A
*
*********************************************************************/
-struct url_spec dsplit(char *domain)
+void apply_url_actions(struct current_action_spec *action,
+ struct http_request *http,
+ struct url_actions *b)
{
- struct url_spec ret[1];
- char *v[BUFFER_SIZE];
- int size;
- char *p;
-
- memset(ret, '\0', sizeof(*ret));
-
- if (domain[strlen(domain) - 1] == '.')
- {
- ret->unanchored |= ANCHOR_RIGHT;
- }
-
- if (domain[0] == '.')
- {
- ret->unanchored |= ANCHOR_LEFT;
- }
-
- ret->dbuf = strdup(domain);
-
- /* map to lower case */
- for (p = ret->dbuf; *p ; p++) *p = tolower(*p);
-
- /* split the domain name into components */
- ret->dcnt = ssplit(ret->dbuf, ".", v, SZ(v), 1, 1);
-
- if (ret->dcnt <= 0)
+ if (b == NULL)
{
- memset(ret, '\0', sizeof(ret));
- return(*ret);
+ /* Should never happen */
+ return;
}
- /* save a copy of the pointers in dvec */
- size = ret->dcnt * sizeof(*ret->dvec);
-
- if ((ret->dvec = (char **)malloc(size)))
+ for (b = b->next; NULL != b; b = b->next)
{
- memcpy(ret->dvec, v, size);
+ if (url_match(b->url, http))
+ {
+ merge_current_action(action, b->action);
+ }
}
-
- return(*ret);
-
}
/*********************************************************************
*
- * Function : simple_domaincmp
+ * Function : forward_url
*
- * Description : Domain-wise Compare fqdn's. The comparison is
- * both left- and right-anchored. The individual
- * domain names are compared with simplematch().
- * This is only used by domaincmp.
+ * Description : Should we forward this to another proxy?
*
* Parameters :
- * 1 : pv = array of patterns to compare
- * 2 : fv = array of domain components to compare
- * 3 : len = length of the arrays (both arrays are the
- * same length - if they weren't, it couldn't
- * possibly be a match).
+ * 1 : http = http_request request for current URL
+ * 2 : csp = Current client state (buffers, headers, etc...)
*
- * Returns : 0 => domains are equivalent, else no match.
+ * Returns : Pointer to forwarding information.
*
*********************************************************************/
-static int simple_domaincmp(char **pv, char **fv, int len)
+const struct forward_spec * forward_url(struct http_request *http,
+ struct client_state *csp)
{
- int n;
+ static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
+ struct forward_spec *fwd = csp->config->forward;
+
+ if (fwd == NULL)
+ {
+ return fwd_default;
+ }
- for (n = 0; n < len; n++)
+ while (fwd != NULL)
{
- if (simplematch(pv[n], fv[n]))
+ if (url_match(fwd->url, http))
{
- return 1;
+ return fwd;
}
+ fwd = fwd->next;
}
- return 0;
-
+ return fwd_default;
}
/*********************************************************************
*
- * Function : domaincmp
+ * Function : direct_response
*
- * Description : Domain-wise Compare fqdn's. Governed by the bimap in
- * pattern->unachored, the comparison is un-, left-,
- * right-anchored, or both.
- * The individual domain names are compared with
- * simplematch().
+ * Description : Check if Max-Forwards == 0 for an OPTIONS or TRACE
+ * request and if so, return a HTTP 501 to the client.
*
- * Parameters :
- * 1 : pattern = a domain that may contain a '*' as a wildcard.
- * 2 : fqdn = domain name against which the patterns are compared.
+ * FIXME: I have a stupid name and I should handle the
+ * requests properly. Still, what we do here is rfc-
+ * compliant, whereas ignoring or forwarding are not.
*
- * Returns : 0 => domains are equivalent, else no match.
+ * Parameters :
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ *
+ * Returns : http_response if , NULL if nonmatch or handler fail
*
*********************************************************************/
-int domaincmp(struct url_spec *pattern, struct url_spec *fqdn)
+struct http_response *direct_response(struct client_state *csp)
{
- char **pv, **fv; /* vectors */
- int plen, flen;
- int unanchored = pattern->unanchored & (ANCHOR_RIGHT | ANCHOR_LEFT);
-
- plen = pattern->dcnt;
- flen = fqdn->dcnt;
-
- if (flen < plen)
- {
- /* fqdn is too short to match this pattern */
- return 1;
- }
-
- pv = pattern->dvec;
- fv = fqdn->dvec;
+ struct http_response *rsp;
+ struct list_entry *p;
- if (unanchored == ANCHOR_LEFT)
- {
- /*
- * Right anchored.
- *
- * Convert this into a fully anchored pattern with
- * the fqdn and pattern the same length
- */
- fv += (flen - plen); /* flen - plen >= 0 due to check above */
- return simple_domaincmp(pv, fv, plen);
- }
- else if (unanchored == 0)
- {
- /* Fully anchored, check length */
- if (flen != plen)
- {
- return 1;
- }
- return simple_domaincmp(pv, fv, plen);
- }
- else if (unanchored == ANCHOR_RIGHT)
- {
- /* Left anchored, ignore all extra in fqdn */
- return simple_domaincmp(pv, fv, plen);
- }
- else
+ if ((0 == strcmpic(csp->http->gpc, "trace"))
+ || (0 == strcmpic(csp->http->gpc, "options")))
{
- /* Unanchored */
- int n;
- int maxn = flen - plen;
- for (n = 0; n <= maxn; n++)
+ for (p = csp->headers->first; (p != NULL) ; p = p->next)
{
- if (!simple_domaincmp(pv, fv, plen))
+ if (!strncmp("Max-Forwards:", p->str, 13)
+ && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
{
- return 0;
+ /* FIXME: We could handle at least TRACE here,
+ but that would require a verbatim copy of
+ the request which we don't have anymore */
+
+ log_error(LOG_LEVEL_HEADER, "Found Max-Forwards:0 in OPTIONS or TRACE request -- Returning 501");
+
+ /* Get mem for response or fail*/
+ if (NULL == (rsp = alloc_http_response()))
+ {
+ return cgi_error_memory();
+ }
+
+ if (NULL == (rsp->status = strdup("501 Not Implemented")))
+ {
+ free_http_response(rsp);
+ return cgi_error_memory();
+ }
+
+ rsp->is_static = 1;
+ return(finish_http_response(rsp));
}
- /*
- * Doesn't match from start of fqdn
- * Try skipping first part of fqdn
- */
- fv++;
}
- return 1;
}
-
+ return NULL;
}