X-Git-Url: http://www.privoxy.org/gitweb/?a=blobdiff_plain;f=filters.c;h=15a9f31a5ecf4684c2b28153812652fe8b5e78a8;hb=e8c410da3b4ea069c0d3cb3692e049f51d957a3b;hp=1a1da409823f7a6ed0df34740e5f3da4222b71b4;hpb=656cce68348695162120584f5c12ceeed16b2792;p=privoxy.git
diff --git a/filters.c b/filters.c
index 1a1da409..15a9f31a 100644
--- a/filters.c
+++ b/filters.c
@@ -1,4 +1,4 @@
-const char filters_rcs[] = "$Id: filters.c,v 1.63 2006/08/31 10:11:28 fabiankeil Exp $";
+const char filters_rcs[] = "$Id: filters.c,v 1.88 2007/06/01 16:41:11 fabiankeil Exp $";
/*********************************************************************
*
* File : $Source: /cvsroot/ijbswa/current/filters.c,v $
@@ -10,9 +10,10 @@ const char filters_rcs[] = "$Id: filters.c,v 1.63 2006/08/31 10:11:28 fabiankeil
* `filter_popups', `forward_url', 'redirect_url',
* `ij_untrusted_url', `intercept_url', `pcrs_filter_respose',
* `ijb_send_banner', `trust_url', `gif_deanimate_response',
- * `jpeg_inspect_response'
+ * `jpeg_inspect_response', `execute_single_pcrs_command',
+ * `rewrite_url', `get_last_url'
*
- * Copyright : Written by and Copyright (C) 2001, 2004 the SourceForge
+ * Copyright : Written by and Copyright (C) 2001, 2004-2007 the SourceForge
* Privoxy team. http://www.privoxy.org/
*
* Based on the Internet Junkbuster originally written
@@ -39,6 +40,112 @@ const char filters_rcs[] = "$Id: filters.c,v 1.63 2006/08/31 10:11:28 fabiankeil
*
* Revisions :
* $Log: filters.c,v $
+ * 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
@@ -488,6 +595,7 @@ const char filters_rcs[] = "$Id: filters.c,v 1.63 2006/08/31 10:11:28 fabiankeil
#include "list.h"
#include "deanimate.h"
#include "urlmatch.h"
+#include "loaders.h"
#ifdef _WIN32
#include "win32.h"
@@ -579,20 +687,33 @@ int block_acl(struct access_control_addr *dst, struct client_state *csp)
* Returns : 0 => Ok, everything else is an error.
*
*********************************************************************/
-int acl_addr(char *aspec, struct access_control_addr *aca)
+int acl_addr(const char *aspec, struct access_control_addr *aca)
{
- int i, masklength, port;
+ int i, masklength;
+ long port;
char *p;
+ char *acl_spec = NULL;
masklength = 32;
port = 0;
- if ((p = strchr(aspec, '/')) != NULL)
+ /*
+ * 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)
{
- *p++ = '\0';
+ /* XXX: This will be logged as parse error. */
+ return(-1);
+ }
+ if ((p = strchr(acl_spec, '/')) != NULL)
+ {
+ *p++ = '\0';
if (ijb_isdigit(*p) == 0)
{
+ free(acl_spec);
return(-1);
}
masklength = atoi(p);
@@ -600,26 +721,32 @@ int acl_addr(char *aspec, struct access_control_addr *aca)
if ((masklength < 0) || (masklength > 32))
{
+ free(acl_spec);
return(-1);
}
- if ((p = strchr(aspec, ':')) != NULL)
+ if ((p = strchr(acl_spec, ':')) != NULL)
{
+ char *endptr;
+
*p++ = '\0';
+ port = strtol(p, &endptr, 10);
- if (ijb_isdigit(*p) == 0)
+ if (port <= 0 || port > 65535 || *endptr != '\0')
{
+ free(acl_spec);
return(-1);
}
- port = atoi(p);
}
- aca->port = port;
+ aca->port = (unsigned long)port;
- aca->addr = ntohl(resolve_hostname_to_ip(aspec));
+ aca->addr = ntohl(resolve_hostname_to_ip(acl_spec));
+ free(acl_spec);
if (aca->addr == INADDR_NONE)
{
+ /* XXX: This will be logged as parse error. */
return(-1);
}
@@ -765,7 +892,7 @@ struct http_response *block_url(struct client_state *csp)
if (((csp->action->flags & ACTION_IMAGE_BLOCKER) != 0)
&& is_imageurl(csp))
{
- char *p;
+ char *p;
/* determine HOW images should be blocked */
p = csp->action->string[ACTION_STRING_IMAGE_BLOCKER];
@@ -889,7 +1016,6 @@ struct http_response *block_url(struct client_state *csp)
return cgi_error_memory();
}
}
-
}
else
#endif /* def FEATURE_IMAGE_BLOCKING */
@@ -918,7 +1044,7 @@ struct http_response *block_url(struct client_state *csp)
}
else
{
- rsp->status = strdup("404 Request for blocked URL");
+ rsp->status = strdup("403 Request for blocked URL");
}
if (rsp->status == NULL)
@@ -936,7 +1062,15 @@ struct http_response *block_url(struct client_state *csp)
#ifdef FEATURE_FORCE_LOAD
err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
- if (csp->http->ssl != 0)
+ /*
+ * Export the force conditional block killer if
+ *
+ * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
+ * - Privoxy is configured to enforce blocks, or
+ * - it's a CONNECT request and enforcing wouldn't work anyway.
+ */
+ if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
+ || (0 == strcmpic(csp->http->gpc, "connect")))
#endif /* ndef FEATURE_FORCE_LOAD */
{
err = map_block_killer(exports, "force-support");
@@ -961,8 +1095,9 @@ struct http_response *block_url(struct client_state *csp)
return cgi_error_memory();
}
}
+ rsp->reason = RSP_REASON_BLOCKED;
- return finish_http_response(rsp);
+ return finish_http_response(csp, rsp);
}
@@ -1044,7 +1179,7 @@ struct http_response *trust_url(struct client_state *csp)
p = strdup("");
for (tl = csp->config->trust_list; (t = *tl) != NULL ; tl++)
{
- sprintf(buf, "%s\n", t->spec);
+ snprintf(buf, sizeof(buf), "%s\n", t->spec);
string_append(&p, buf);
}
err = map(exports, "trusted-referrers", 1, p, 0);
@@ -1066,7 +1201,7 @@ struct http_response *trust_url(struct client_state *csp)
p = strdup("");
for (l = csp->config->trust_info->first; l ; l = l->next)
{
- sprintf(buf, " %s
\n",l->str, l->str);
+ snprintf(buf, sizeof(buf), " %s
\n", l->str, l->str);
string_append(&p, buf);
}
err = map(exports, "trust-info", 1, p, 0);
@@ -1084,10 +1219,22 @@ struct http_response *trust_url(struct client_state *csp)
}
/*
- * Export the force prefix or the force conditional block killer
+ * Export the force conditional block killer if
+ *
+ * - Privoxy was compiled without FEATURE_FORCE_LOAD, or
+ * - Privoxy is configured to enforce blocks, or
+ * - it's a CONNECT request and enforcing wouldn't work anyway.
*/
#ifdef FEATURE_FORCE_LOAD
- err = map(exports, "force-prefix", 1, FORCE_PREFIX, 1);
+ if ((csp->config->feature_flags & RUNTIME_FEATURE_ENFORCE_BLOCKS)
+ || (0 == strcmpic(csp->http->gpc, "connect")))
+ {
+ err = map_block_killer(exports, "force-support");
+ }
+ else
+ {
+ 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 */
@@ -1108,107 +1255,357 @@ struct http_response *trust_url(struct client_state *csp)
free_http_response(rsp);
return cgi_error_memory();
}
+ rsp->reason = RSP_REASON_UNTRUSTED;
- return finish_http_response(rsp);
+ return finish_http_response(csp, rsp);
}
#endif /* def FEATURE_TRUST */
-#ifdef FEATURE_FAST_REDIRECTS
/*********************************************************************
*
- * Function : redirect_url
+ * Function : compile_dynamic_pcrs_job_list
*
- * Description : Checks for redirection URLs and returns a HTTP redirect
- * to the destination URL, if necessary
+ * Description : Compiles a dynamic pcrs job list (one with variables
+ * resolved at request time)
*
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
+ * 2 : b = The filter list to compile
*
- * Returns : NULL if URL was clean, HTTP redirect otherwise.
+ * Returns : NULL in case of errors, otherwise the
+ * pcrs job list.
*
*********************************************************************/
-struct http_response *redirect_url(struct client_state *csp)
+pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
{
- char *p, *q;
- struct http_response *rsp;
- char *redirect_mode = NULL;
- int x, y;
+ struct list_entry *pattern;
+ pcrs_job *job_list = NULL;
+ pcrs_job *dummy = NULL;
+ pcrs_job *lastjob = NULL;
+ int error = 0;
- if ((csp->action->flags & ACTION_REDIRECT))
+ const struct pcrs_variable variables[] =
{
- q = csp->action->string[ACTION_STRING_REDIRECT];
- }
- else
+ {"url", csp->http->url, 1},
+ {"path", csp->http->path, 1},
+ {"host", csp->http->host, 1},
+ {"origin", csp->ip_addr_str, 1},
+ {NULL, NULL, 1}
+ };
+
+ for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
{
- redirect_mode = csp->action->string[ACTION_STRING_FAST_REDIRECTS];
- if (0 == strcmpic(redirect_mode, "check-decoded-url"))
- {
- p = q = csp->http->path;
- log_error(LOG_LEVEL_REDIRECTS, "Decoding path: %s if necessary.", p);
- while (*p)
+ assert(pattern->str != NULL);
+
+ 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));
+ continue;
+ }
+ else
+ {
+ if (error == PCRS_WARN_TRUNCATION)
{
- if (*p == '%') /* Escape sequence? */
- {
- /* Yes, translate from hexadecimal to decimal */
- p++;
- /* First byte */
- x=((int)*p++)-48;
- if (x>9) x-=7;
- x<<=4;
- /* Second byte */
- y=((int)*p++)-48;
- if (y>9)y-=7;
- /* Merge */
- *q++=(char)(x|y);
- }
- else
- {
- /* No, forward character. */
- *q++=*p++;
- }
+ log_error(LOG_LEVEL_ERROR,
+ "At least one of the variables in \'%s\' had to "
+ "be truncated before compilation", pattern->str);
}
- *q='\0';
+ if (job_list == NULL)
+ {
+ job_list = dummy;
+ }
+ else
+ {
+ lastjob->next = dummy;
+ }
+ lastjob = dummy;
}
- p = q = csp->http->path;
- log_error(LOG_LEVEL_REDIRECTS, "Checking path for redirects: %s", p);
+ }
- /*
- * find the last URL encoded in the request
- */
- while ((p = strstr(p, "http://")) != NULL)
+ return job_list;
+}
+
+
+/*********************************************************************
+ *
+ * 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.
+ *
+ * Parameters :
+ * 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 = pcrs_execute_single_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);
+ freez(new_url);
+ }
+ else if (hits < 0)
+ {
+ log_error(LOG_LEVEL_REDIRECTS,
+ "executing pcrs command \"%s\" to rewrite %s failed: %s",
+ pcrs_command, old_url, pcrs_strerror(hits));
+ 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
{
- q = p++;
+ log_error(LOG_LEVEL_ERROR, "Unable to decode \"%s\".", subject);
}
}
+
+ log_error(LOG_LEVEL_REDIRECTS, "Checking \"%s\" for redirects.", subject);
+
/*
- * if there was any, generate and return a HTTP redirect
+ * Find the last URL encoded in the request
*/
- if (q != csp->http->path)
+ tmp = subject;
+ while ((tmp = strstr(tmp, "http://")) != NULL)
+ {
+ new_url = tmp++;
+ }
+ tmp = (new_url != NULL) ? new_url : subject;
+ while ((tmp = strstr(tmp, "https://")) != NULL)
{
- log_error(LOG_LEVEL_REDIRECTS, "redirecting to: %s", q);
+ new_url = tmp++;
+ }
- if (NULL == (rsp = alloc_http_response()))
+ 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')
{
- return cgi_error_memory();
+ old_url = csp->http->url;
+ new_url = rewrite_url(old_url, redirection_string);
}
-
- if ( enlist_unique_header(rsp->headers, "Location", q)
- || (NULL == (rsp->status = strdup("302 Local Redirect from Privoxy"))) )
+ else
{
- free_http_response(rsp);
- return cgi_error_memory();
+ 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];
- return finish_http_response(rsp);
+ /*
+ * If it exists, use the previously rewritten URL as input
+ * otherwise just use the old path.
+ */
+ old_url = (new_url != NULL) ? new_url : strdup(csp->http->path);
+ new_url = get_last_url(old_url, redirect_mode);
+ freez(old_url);
}
- else
+
+ /*
+ * Disable redirect checkers, so that they
+ * will be only run more than once if the user
+ * also enables them through tags.
+ *
+ * From a performance point of view
+ * it doesn't matter, but the duplicated
+ * log messages are annoying.
+ */
+ csp->action->flags &= ~ACTION_FAST_REDIRECTS;
+#endif /* def FEATURE_FAST_REDIRECTS */
+ csp->action->flags &= ~ACTION_REDIRECT;
+
+ /* Did any redirect action trigger? */
+ if (new_url)
{
- return NULL;
+ 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();
+ }
+ rsp->reason = RSP_REASON_REDIRECTED;
+ freez(new_url);
+
+ return finish_http_response(csp, rsp);
+ }
}
+ /* Only reached if no redirect is required */
+ return NULL;
+
}
-#endif /* def FEATURE_FAST_REDIRECTS */
#ifdef FEATURE_IMAGE_BLOCKING
@@ -1229,7 +1626,7 @@ struct http_response *redirect_url(struct client_state *csp)
* otherwise
*
*********************************************************************/
-int is_imageurl(struct client_state *csp)
+int is_imageurl(const struct client_state *csp)
{
#ifdef FEATURE_IMAGE_DETECT_MSIE
char *tmp;
@@ -1276,7 +1673,7 @@ int is_imageurl(struct client_state *csp)
* Returns : 0 => trusted, 1 => untrusted
*
*********************************************************************/
-int is_untrusted_url(struct client_state *csp)
+int is_untrusted_url(const struct client_state *csp)
{
struct file_list *fl;
struct block_spec *b;
@@ -1328,7 +1725,7 @@ int is_untrusted_url(struct client_state *csp)
{
/* 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).
+ * domain and return 0 (which means it's OK).
*/
FILE *fp;
@@ -1359,9 +1756,19 @@ int is_untrusted_url(struct client_state *csp)
string_join(&new_entry, path);
}
+ /*
+ * Give a reason for generating this entry.
+ */
+ string_append(&new_entry, " # Trusted referrer was: ");
+ string_append(&new_entry, referer);
+
if (new_entry != NULL)
{
- fprintf(fp, "%s\n", new_entry);
+ if (-1 == fprintf(fp, "%s\n", new_entry))
+ {
+ log_error(LOG_LEVEL_ERROR, "Failed to append \'%s\' to trustfile \'%s\': %E",
+ new_entry, csp->config->trustfile);
+ }
free(new_entry);
}
else
@@ -1372,9 +1779,15 @@ int is_untrusted_url(struct client_state *csp)
fclose(fp);
}
+ else
+ {
+ log_error(LOG_LEVEL_ERROR, "Failed to append new entry for \'%s\' to trustfile \'%s\': %E",
+ csp->http->hostport, csp->config->trustfile);
+ }
return 0;
}
}
+
return 1;
}
#endif /* def FEATURE_TRUST */
@@ -1390,6 +1803,11 @@ int is_untrusted_url(struct client_state *csp)
* csp->content_length to the modified size and raise the
* CSP_FLAG_MODIFIED flag.
*
+ * XXX: Currently pcrs_filter_response is also responsible
+ * for dechunking and decompressing. Both should be
+ * done in separate functions so other content modifiers
+ * profit as well, even if pcrs filtering is disabled.
+ *
* Parameters :
* 1 : csp = Current client state (buffers, headers, etc...)
*
@@ -1418,7 +1836,7 @@ char *pcrs_filter_response(struct client_state *csp)
{
return(NULL);
}
- size = csp->iob->eod - csp->iob->cur;
+ size = (size_t)(csp->iob->eod - csp->iob->cur);
/*
* Need to check the set of re_filterfiles...
@@ -1439,7 +1857,7 @@ char *pcrs_filter_response(struct client_state *csp)
if (0 == found_filters)
{
log_error(LOG_LEVEL_ERROR, "Unable to get current state of regexp filtering.");
- return(NULL);
+ return(NULL);
}
/*
@@ -1457,11 +1875,61 @@ char *pcrs_filter_response(struct client_state *csp)
csp->flags |= CSP_FLAG_MODIFIED;
}
+#ifdef FEATURE_ZLIB
+ /*
+ * If the body has a compressed transfer-encoding,
+ * uncompress it first, adjusting size and iob->eod.
+ * Note that decompression occurs after de-chunking.
+ */
+ if (csp->content_type & (CT_GZIP | CT_DEFLATE))
+ {
+ /* Notice that we at least tried to decompress. */
+ if (JB_ERR_OK != decompress_iob(csp))
+ {
+ /*
+ * We failed to decompress the data; there's no point
+ * in continuing since we can't filter.
+ *
+ * XXX: Actually the Accept-Encoding header may
+ * just be incorrect in which case we could continue
+ * with filtering.
+ *
+ * Unset CT_GZIP and CT_DEFLATE to remember not
+ * to modify the Content-Encoding header later.
+ */
+ csp->content_type &= ~CT_GZIP;
+ csp->content_type &= ~CT_DEFLATE;
+ return(NULL);
+ }
+
+ /*
+ * Decompression gives us a completely new iob,
+ * so we need to update.
+ */
+ size = (size_t)(csp->iob->eod - csp->iob->cur);
+ old = csp->iob->cur;
+
+ csp->flags |= CSP_FLAG_MODIFIED;
+ csp->content_type &= ~CT_TABOO;
+ }
+#endif
+
for (i = 0; i < MAX_AF_FILES; i++)
{
fl = csp->rlist[i];
if ((NULL == fl) || (NULL == fl->f))
- break;
+ {
+ /*
+ * 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
@@ -1469,14 +1937,25 @@ char *pcrs_filter_response(struct client_state *csp)
*/
for (b = fl->f; b; b = b->next)
{
+ if (b->type != FT_CONTENT_FILTER)
+ {
+ /* 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;
+ 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;
+
+ if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
- if ( NULL == b->joblist )
+ if (NULL == joblist)
{
log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
continue;
@@ -1484,15 +1963,54 @@ char *pcrs_filter_response(struct client_state *csp)
prev_size = size;
/* Apply all jobs from the joblist */
- for (job = b->joblist; NULL != job; job = job->next)
+ for (job = joblist; NULL != job; job = job->next)
{
- current_hits += pcrs_execute(job, old, size, &new, &size);
- if (old != csp->iob->cur) free(old);
- old=new;
+ 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)
+ {
+ free(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;
+ }
}
- 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);
+ 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);
hits += current_hits;
}
@@ -1538,7 +2056,9 @@ char *gif_deanimate_response(struct client_state *csp)
{
struct binbuffer *in, *out;
char *p;
- size_t size = csp->iob->eod - csp->iob->cur;
+ size_t size;
+
+ size = (size_t)(csp->iob->eod - csp->iob->cur);
/*
* If the body has a "chunked" transfer-encoding,
@@ -1608,9 +2128,12 @@ char *gif_deanimate_response(struct client_state *csp)
*********************************************************************/
char *jpeg_inspect_response(struct client_state *csp)
{
- struct binbuffer *in = NULL, *out = NULL;
+ struct binbuffer *in = NULL;
+ struct binbuffer *out = NULL;
char *p = NULL;
- size_t size = csp->iob->eod - csp->iob->cur;
+ size_t size;
+
+ size = (size_t)(csp->iob->eod - csp->iob->cur);
/*
* If the body has a "chunked" transfer-encoding,
@@ -1683,7 +2206,7 @@ char *jpeg_inspect_response(struct client_state *csp)
* went wrong
*
*********************************************************************/
-int remove_chunked_transfer_coding(char *buffer, const size_t size)
+size_t remove_chunked_transfer_coding(char *buffer, const size_t size)
{
size_t newsize = 0;
unsigned int chunksize = 0;
@@ -1801,12 +2324,155 @@ void apply_url_actions(struct current_action_spec *action,
}
+/*********************************************************************
+ *
+ * Function : get_forward_override_settings
+ *
+ * Description : Returns forward settings as specified with the
+ * forward-override{} action. forward-override accepts
+ * forward lines similar to the one used in the
+ * configuration file, but without the URL pattern.
+ *
+ * For example:
+ *
+ * forward / .
+ *
+ * in the configuration file can be replaced with
+ * the action section:
+ *
+ * {+forward-override{forward .}}
+ * /
+ *
+ * Parameters :
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ *
+ * Returns : Pointer to forwarding structure in case of success.
+ * Invalid syntax is fatal.
+ *
+ *********************************************************************/
+static const struct forward_spec *get_forward_override_settings(struct client_state *csp)
+{
+ const char *forward_override_line = csp->action->string[ACTION_STRING_FORWARD_OVERRIDE];
+ char forward_settings[BUFFER_SIZE];
+ char *http_parent = NULL;
+ /* variable names were chosen for consistency reasons. */
+ struct forward_spec *fwd = NULL;
+ int vec_count;
+ char *vec[3];
+
+ assert(csp->action->flags & ACTION_FORWARD_OVERRIDE);
+ /* Should be enforced by load_one_actions_file() */
+ assert(strlen(forward_override_line) < sizeof(forward_settings) - 1);
+
+ /* Create a copy ssplit can modify */
+ strlcpy(forward_settings, forward_override_line, sizeof(forward_settings));
+
+ if (NULL != csp->fwd)
+ {
+ /*
+ * XXX: Currently necessary to prevent memory
+ * leaks when the show-url-info cgi page is visited.
+ */
+ unload_forward_spec(csp->fwd);
+ }
+
+ /*
+ * allocate a new forward node, valid only for
+ * 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 */
+ }
+
+ vec_count = ssplit(forward_settings, " \t", vec, SZ(vec), 1, 1);
+ 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 == 3)
+ {
+ char *socks_proxy = NULL;
+
+ if (!strcasecmp(vec[0], "forward-socks4"))
+ {
+ fwd->type = SOCKS_4;
+ socks_proxy = vec[1];
+ }
+ else if (!strcasecmp(vec[0], "forward-socks4a"))
+ {
+ fwd->type = SOCKS_4A;
+ 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;
+ }
+
+ http_parent = vec[2];
+ }
+ }
+
+ if (NULL == http_parent)
+ {
+ log_error(LOG_LEVEL_FATAL,
+ "Invalid forward-override syntax in: %s", forward_override_line);
+ /* Never get here - LOG_LEVEL_FATAL causes program exit */
+ }
+
+ /* 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;
+ }
+ }
+
+ assert (NULL != fwd);
+
+ log_error(LOG_LEVEL_CONNECT,
+ "Overriding forwarding settings based on \'%s\'", forward_override_line);
+
+ return fwd;
+}
+
/*********************************************************************
*
* Function : forward_url
*
* 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...)
@@ -1820,6 +2486,11 @@ const struct forward_spec * forward_url(struct http_request *http,
static const struct forward_spec fwd_default[1] = { FORWARD_SPEC_INITIALIZER };
struct forward_spec *fwd = csp->config->forward;
+ if (csp->action->flags & ACTION_FORWARD_OVERRIDE)
+ {
+ return get_forward_override_settings(csp);
+ }
+
if (fwd == NULL)
{
return fwd_default;
@@ -1865,29 +2536,42 @@ struct http_response *direct_response(struct client_state *csp)
{
for (p = csp->headers->first; (p != NULL) ; p = p->next)
{
- if (!strncmp("Max-Forwards:", p->str, 13)
- && (*(p->str+13) != '\0') && (atoi(p->str+13) == 0))
+ if (!strncmpic("Max-Forwards:", p->str, 13))
{
- /* FIXME: We could handle at least TRACE here,
- but that would require a verbatim copy of
- the request which we don't have anymore */
+ unsigned int max_forwards;
- 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()))
+ /*
+ * If it's a Max-Forwards value of zero,
+ * we have to intercept the request.
+ */
+ if (1 == sscanf(p->str+12, ": %u", &max_forwards) && max_forwards == 0)
{
- return cgi_error_memory();
- }
+ /*
+ * 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,
+ "Detected header \'%s\' in OPTIONS or TRACE request. Returning 501.",
+ p->str);
+
+ /* 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();
- }
+ 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));
+ rsp->is_static = 1;
+ rsp->reason = RSP_REASON_UNSUPPORTED;
+
+ return(finish_http_response(csp, rsp));
+ }
}
}
}