-const char filters_rcs[] = "$Id: filters.c,v 1.76 2007/01/01 19:36:37 fabiankeil Exp $";
+const char filters_rcs[] = "$Id: filters.c,v 1.90 2007/09/02 12:44:17 fabiankeil Exp $";
/*********************************************************************
*
* File : $Source: /cvsroot/ijbswa/current/filters.c,v $
* `jpeg_inspect_response', `execute_single_pcrs_command',
* `rewrite_url', `get_last_url'
*
- * Copyright : Written by and Copyright (C) 2001, 2004-2006 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
*
* Revisions :
* $Log: filters.c,v $
+ * Revision 1.90 2007/09/02 12:44:17 fabiankeil
+ * Remove newline at the end of a log_error() message.
+ *
+ * Revision 1.89 2007/08/05 13:42:23 fabiankeil
+ * #1763173 from Stefan Huehner: declare some more functions static.
+ *
+ * Revision 1.88 2007/06/01 16:41:11 fabiankeil
+ * Add forward-override{} to change the forwarding settings through
+ * action sections. This is mainly interesting to forward different
+ * clients differently (for example based on User-Agent or request
+ * origin).
+ *
+ * Revision 1.87 2007/04/30 15:53:10 fabiankeil
+ * Make sure filters with dynamic jobs actually use them.
+ *
+ * Revision 1.86 2007/04/30 15:03:28 fabiankeil
+ * - Introduce dynamic pcrs jobs that can resolve variables.
+ * - Don't run redirect functions more than once,
+ * unless they are activated more than once.
+ *
+ * Revision 1.85 2007/03/21 12:24:47 fabiankeil
+ * - Log the content size after decompression in decompress_iob()
+ * instead of pcrs_filter_response().
+ *
+ * Revision 1.84 2007/03/20 15:16:34 fabiankeil
+ * Use dedicated header filter actions instead of abusing "filter".
+ * Replace "filter-client-headers" and "filter-client-headers"
+ * with "server-header-filter" and "client-header-filter".
+ *
+ * Revision 1.83 2007/03/17 15:20:05 fabiankeil
+ * New config option: enforce-blocks.
+ *
+ * Revision 1.82 2007/03/13 11:28:43 fabiankeil
+ * - Fix port handling in acl_addr() and use a temporary acl spec
+ * copy so error messages don't contain a truncated version.
+ * - Log size of iob before and after decompression.
+ *
+ * Revision 1.81 2007/03/05 14:40:53 fabiankeil
+ * - Cosmetical changes for LOG_LEVEL_RE_FILTER messages.
+ * - Hide the "Go there anyway" link for blocked CONNECT
+ * requests where going there anyway doesn't work anyway.
+ *
+ * Revision 1.80 2007/02/07 10:55:20 fabiankeil
+ * - Save the reason for generating http_responses.
+ * - Block (+block) with status code 403 instead of 404.
+ * - Use a different kludge to remember a failed decompression.
+ *
+ * Revision 1.79 2007/01/31 16:21:38 fabiankeil
+ * Search for Max-Forwards headers case-insensitive,
+ * don't generate the "501 unsupported" message for invalid
+ * Max-Forwards values and don't increase negative ones.
+ *
+ * Revision 1.78 2007/01/28 13:41:18 fabiankeil
+ * - Add HEAD support to finish_http_response.
+ * - Add error favicon to internal HTML error messages.
+ *
+ * Revision 1.77 2007/01/12 15:36:44 fabiankeil
+ * Mark *csp as immutable for is_untrusted_url()
+ * and is_imageurl(). Closes FR 1237736.
+ *
* Revision 1.76 2007/01/01 19:36:37 fabiankeil
* Integrate a modified version of Wil Mahan's
* zlib patch (PR #895531).
#include "list.h"
#include "deanimate.h"
#include "urlmatch.h"
+#include "loaders.h"
#ifdef _WIN32
#include "win32.h"
* 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);
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);
}
#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;
-
-}
-
-
/*********************************************************************
*
* Function : block_url
}
else
{
- rsp->status = strdup("404 Request for blocked URL");
+ rsp->status = strdup("403 Request for blocked URL");
}
if (rsp->status == NULL)
#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");
return cgi_error_memory();
}
}
+ rsp->reason = RSP_REASON_BLOCKED;
- return finish_http_response(rsp);
+ return finish_http_response(csp, rsp);
}
}
/*
- * 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 */
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 */
+
/*********************************************************************
*
- * Function : execute_single_pcrs_command
+ * Function : compile_dynamic_pcrs_job_list
*
- * 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.
+ * Description : Compiles a dynamic pcrs job list (one with variables
+ * resolved at request time)
*
* Parameters :
- * 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
+ * 1 : csp = Current client state (buffers, headers, etc...)
+ * 2 : b = The filter list to compile
*
* Returns : NULL in case of errors, otherwise the
- * result of the pcrs command.
+ * pcrs job list.
*
*********************************************************************/
-char *execute_single_pcrs_command(char *subject, const char *pcrs_command, int *hits)
+pcrs_job *compile_dynamic_pcrs_job_list(const struct client_state *csp, const struct re_filterfile_spec *b)
{
- int error;
- size_t size;
- char *result = NULL;
- pcrs_job *job;
-
- assert(subject);
- assert(pcrs_command);
-
- *hits = 0;
- size = strlen(subject);
+ struct list_entry *pattern;
+ pcrs_job *job_list = NULL;
+ pcrs_job *dummy = NULL;
+ pcrs_job *lastjob = NULL;
+ int error = 0;
- if (NULL == (job = pcrs_compile_command(pcrs_command, &error)))
- {
- log_error(LOG_LEVEL_ERROR, "Failed to compile pcrs command \"%s\". Error: %d.",
- pcrs_command, error);
- }
- else if ((*hits = pcrs_execute(job, subject, size, &result, &size)) < 0)
+ const struct pcrs_variable variables[] =
{
- log_error(LOG_LEVEL_ERROR, "Failed to execute pcrs command: %s", pcrs_strerror(*hits));
- *hits = 0;
- freez(result);
- }
+ {"url", csp->http->url, 1},
+ {"path", csp->http->path, 1},
+ {"host", csp->http->host, 1},
+ {"origin", csp->ip_addr_str, 1},
+ {NULL, NULL, 1}
+ };
- if (job)
+ for (pattern = b->patterns->first; pattern != NULL; pattern = pattern->next)
{
- job = pcrs_free_job(job);
- }
+ assert(pattern->str != NULL);
- return result;
+ 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)
+ {
+ log_error(LOG_LEVEL_ERROR,
+ "At least one of the variables in \'%s\' had to "
+ "be truncated before compilation", pattern->str);
+ }
+ if (job_list == NULL)
+ {
+ job_list = dummy;
+ }
+ else
+ {
+ lastjob->next = dummy;
+ }
+ lastjob = dummy;
+ }
+ }
+ return job_list;
}
+
/*********************************************************************
*
* Function : rewrite_url
assert(old_url);
assert(pcrs_command);
- new_url = execute_single_pcrs_command(old_url, pcrs_command, &hits);
+ 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, new_url);
+ 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))
new_url = get_last_url(old_url, redirect_mode);
freez(old_url);
}
+
+ /*
+ * 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)
free_http_response(rsp);
return cgi_error_memory();
}
+ rsp->reason = RSP_REASON_REDIRECTED;
freez(new_url);
- return finish_http_response(rsp);
+
+ return finish_http_response(csp, rsp);
}
}
return 0;
}
}
+
return 1;
}
#endif /* def FEATURE_TRUST */
* uncompress it first, adjusting size and iob->eod.
* Note that decompression occurs after de-chunking.
*/
- if (csp->content_type & CT_GZIP || csp->content_type & CT_DEFLATE)
+ 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. This is
- * slightly tricky because we need to remember not to
- * modify the Content-Encoding header later; using
- * CT_TABOO flag is a kludge for this purpose.
+ * 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_TABOO;
+ csp->content_type &= ~CT_GZIP;
+ csp->content_type &= ~CT_DEFLATE;
return(NULL);
}
- log_error(LOG_LEVEL_RE_FILTER, "Decompressing successful");
/*
* Decompression gives us a completely new iob,
*/
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)
{
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 ( NULL == b->joblist )
+ if (b->dynamic) joblist = compile_dynamic_pcrs_job_list(csp, b);
+
+ if (NULL == joblist)
{
log_error(LOG_LEVEL_RE_FILTER, "Filter %s has empty joblist. Nothing to do.", b->name);
continue;
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)
{
job_number++;
job_hits = pcrs_execute(job, old, size, &new, &size);
}
}
+ if (b->dynamic) pcrs_free_joblist(joblist);
+
log_error(LOG_LEVEL_RE_FILTER,
- "re_filtering %s%s (size %d) with filter %s produced %d hits (new size %d).",
+ "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;
}
/* FIXME: Should this get its own loglevel? */
- log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d\n", size, newsize);
+ log_error(LOG_LEVEL_RE_FILTER, "De-chunking successful. Shrunk from %d to %d", size, newsize);
return(newsize);
}
}
+/*********************************************************************
+ *
+ * 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...)
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;
{
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));
+ }
}
}
}
projects / privoxy.git / blobdiff