-const char urlmatch_rcs[] = "$Id: urlmatch.c,v 1.78 2013/11/24 14:25:19 fabiankeil Exp $";
/*********************************************************************
*
* File : $Source: /cvsroot/ijbswa/current/urlmatch.c,v $
* Purpose : Declares functions to match URLs against URL
* patterns.
*
- * Copyright : Written by and Copyright (C) 2001-2011
- * the Privoxy team. http://www.privoxy.org/
+ * Copyright : Written by and Copyright (C) 2001-2020
+ * the Privoxy team. https://www.privoxy.org/
*
* Based on the Internet Junkbuster originally written
* by and Copyright (C) 1997 Anonymous Coders and
#include <assert.h>
#include <string.h>
-#if !defined(_WIN32) && !defined(__OS2__)
+#if !defined(_WIN32)
#include <unistd.h>
#endif
#include "miscutil.h"
#include "errlog.h"
-const char urlmatch_h_rcs[] = URLMATCH_H_VERSION;
-
enum regex_anchoring
{
NO_ANCHORING,
RIGHT_ANCHORED,
RIGHT_ANCHORED_HOST
};
-static jb_err compile_host_pattern(struct pattern_spec *url, const char *host_pattern);
+static jb_err compile_vanilla_host_pattern(struct pattern_spec *url, const char *host_pattern);
+#ifdef FEATURE_PCRE_HOST_PATTERNS
+static jb_err compile_pcre_host_pattern(struct pattern_spec *url, const char *host_pattern);
+#endif
/*********************************************************************
*
freez(http->url);
freez(http->hostport);
freez(http->path);
- freez(http->ver);
+ freez(http->version);
freez(http->host_ip_addr_str);
-#ifndef FEATURE_EXTENDED_HOST_PATTERNS
freez(http->dbuffer);
freez(http->dvec);
http->dcount = 0;
-#endif
}
-#ifndef FEATURE_EXTENDED_HOST_PATTERNS
/*********************************************************************
*
* Function : init_domain_components
return JB_ERR_OK;
}
-#endif /* ndef FEATURE_EXTENDED_HOST_PATTERNS */
/*********************************************************************
else if (strncmpic(url_noproto, "https://", 8) == 0)
{
/*
- * Should only happen when called from cgi_show_url_info().
+ * Should only happen when called from cgi_show_url_info()
+ * or when the request was https-inspected and the request
+ * line got rewritten.
*/
url_noproto += 8;
http->ssl = 1;
/*
* Got a path.
*
- * NOTE: The following line ignores the path for HTTPS URLS.
- * This means that you get consistent behaviour if you type a
- * https URL in and it's parsed by the function. (When the
- * URL is actually retrieved, SSL hides the path part).
+ * If FEATURE_HTTPS_INSPECTION isn't available, ignore the
+ * path for https URLs so that we get consistent behaviour
+ * if a https URL is parsed. When the URL is actually
+ * retrieved, https hides the path part.
*/
- http->path = strdup_or_die(http->ssl ? "/" : url_path);
+ http->path = strdup_or_die(
+#ifndef FEATURE_HTTPS_INSPECTION
+ http->ssl ? "/" :
+#endif
+ url_path
+ );
*url_path = '\0';
- http->hostport = strdup_or_die(url_noproto);
+ http->hostport = string_tolower(url_noproto);
}
else
{
* or CONNECT requests
*/
http->path = strdup_or_die("/");
- http->hostport = strdup_or_die(url_noproto);
+ http->hostport = string_tolower(url_noproto);
}
freez(buf);
+
+ if (http->hostport == NULL)
+ {
+ return JB_ERR_PARSE;
+ }
}
if (!host_available)
freez(buf);
}
-#ifdef FEATURE_EXTENDED_HOST_PATTERNS
- return JB_ERR_OK;
-#else
/* Split domain name so we can compare it against wildcards */
return init_domain_components(http);
-#endif /* def FEATURE_EXTENDED_HOST_PATTERNS */
}
*/
"VERSION-CONTROL", "REPORT", "CHECKOUT", "CHECKIN", "UNCHECKOUT",
"MKWORKSPACE", "UPDATE", "LABEL", "MERGE", "BASELINE-CONTROL", "MKACTIVITY",
+ /*
+ * The PATCH method is defined by RFC5789, the format of the
+ * actual patch in the body depends on the application, but from
+ * Privoxy's point of view it doesn't matter.
+ */
+ "PATCH",
};
int i;
}
+/*********************************************************************
+ *
+ * Function : normalize_http_version
+ *
+ * Description : Take a supported HTTP version string and remove
+ * leading zeroes etc., reject unsupported versions.
+ *
+ * This is an explicit RFC 2616 (3.1) MUST and
+ * RFC 7230 mandates that intermediaries send their
+ * own HTTP-version in forwarded messages.
+ *
+ * Parameters :
+ * 1 : http_version = HTTP version string
+ *
+ * Returns : JB_ERR_OK on success
+ * JB_ERR_PARSE if the HTTP version is unsupported
+ *
+ *********************************************************************/
+static jb_err normalize_http_version(char *http_version)
+{
+ unsigned int major_version;
+ unsigned int minor_version;
+
+ if (2 != sscanf(http_version, "HTTP/%u.%u", &major_version, &minor_version))
+ {
+ log_error(LOG_LEVEL_ERROR, "Unsupported HTTP version: %s", http_version);
+ return JB_ERR_PARSE;
+ }
+
+ if (major_version != 1 || (minor_version != 0 && minor_version != 1))
+ {
+ log_error(LOG_LEVEL_ERROR, "The only supported HTTP "
+ "versions are 1.0 and 1.1. This rules out: %s", http_version);
+ return JB_ERR_PARSE;
+ }
+
+ assert(strlen(http_version) >= 8);
+ snprintf(http_version, 9, "HTTP/%u.%u", major_version, minor_version);
+
+ return JB_ERR_OK;
+
+}
+
+
/*********************************************************************
*
* Function : parse_http_request
jb_err parse_http_request(const char *req, struct http_request *http)
{
char *buf;
- char *v[10]; /* XXX: Why 10? We should only need three. */
+ char *v[3];
int n;
jb_err err;
return JB_ERR_PARSE;
}
- if (strcmpic(v[2], "HTTP/1.1") && strcmpic(v[2], "HTTP/1.0"))
+ if (JB_ERR_OK != normalize_http_version(v[2]))
{
- log_error(LOG_LEVEL_ERROR, "The only supported HTTP "
- "versions are 1.0 and 1.1. This rules out: %s", v[2]);
freez(buf);
return JB_ERR_PARSE;
}
*/
http->cmd = strdup_or_die(req);
http->gpc = strdup_or_die(v[0]);
- http->ver = strdup_or_die(v[2]);
+ http->version = strdup_or_die(v[2]);
+ http->ocmd = strdup_or_die(http->cmd);
freez(buf);
* 4 : regex = Where the compiled regex should be stored.
*
* Returns : JB_ERR_OK - Success
- * JB_ERR_MEMORY - Out of memory
* JB_ERR_PARSE - Cannot parse regex
*
*********************************************************************/
struct pattern_spec *url, regex_t **regex)
{
int errcode;
- char rebuf[BUFFER_SIZE];
const char *fmt = NULL;
+ char *rebuf;
+ size_t rebuf_size;
assert(pattern);
- assert(strlen(pattern) < sizeof(rebuf) - 2);
if (pattern[0] == '\0')
{
log_error(LOG_LEVEL_FATAL,
"Invalid anchoring in compile_pattern %d", anchoring);
}
+ rebuf_size = strlen(pattern) + strlen(fmt);
+ rebuf = malloc_or_die(rebuf_size);
+ *regex = zalloc_or_die(sizeof(**regex));
- *regex = zalloc(sizeof(**regex));
- if (NULL == *regex)
- {
- free_pattern_spec(url);
- return JB_ERR_MEMORY;
- }
-
- snprintf(rebuf, sizeof(rebuf), fmt, pattern);
+ snprintf(rebuf, rebuf_size, fmt, pattern);
errcode = regcomp(*regex, rebuf, (REG_EXTENDED|REG_NOSUB|REG_ICASE));
if (errcode)
{
- size_t errlen = regerror(errcode, *regex, rebuf, sizeof(rebuf));
- if (errlen > (sizeof(rebuf) - (size_t)1))
+ size_t errlen = regerror(errcode, *regex, rebuf, rebuf_size);
+ if (errlen > (rebuf_size - (size_t)1))
{
- errlen = sizeof(rebuf) - (size_t)1;
+ errlen = rebuf_size - (size_t)1;
}
rebuf[errlen] = '\0';
log_error(LOG_LEVEL_ERROR, "error compiling %s from %s: %s",
pattern, url->spec, rebuf);
free_pattern_spec(url);
+ freez(rebuf);
return JB_ERR_PARSE;
}
+ freez(rebuf);
return JB_ERR_OK;
static jb_err compile_url_pattern(struct pattern_spec *url, char *buf)
{
char *p;
+ const size_t prefix_length = 18;
+
+#ifdef FEATURE_PCRE_HOST_PATTERNS
+ if (strncmpic(buf, "PCRE-HOST-PATTERN:", prefix_length) == 0)
+ {
+ url->pattern.url_spec.host_regex_type = PCRE_HOST_PATTERN;
+ /* Overwrite the "PCRE-HOST-PATTERN:" prefix */
+ memmove(buf, buf+prefix_length, strlen(buf+prefix_length)+1);
+ }
+ else
+ {
+ url->pattern.url_spec.host_regex_type = VANILLA_HOST_PATTERN;
+ }
+#else
+ if (strncmpic(buf, "PCRE-HOST-PATTERN:", prefix_length) == 0)
+ {
+ log_error(LOG_LEVEL_ERROR,
+ "PCRE-HOST-PATTERN detected while Privoxy has been compiled "
+ "without FEATURE_PCRE_HOST_PATTERNS: %s",
+ buf);
+ /* Overwrite the "PCRE-HOST-PATTERN:" prefix */
+ memmove(buf, buf+prefix_length, strlen(buf+prefix_length)+1);
+ /*
+ * The pattern will probably not work as expected.
+ * We don't simply return JB_ERR_PARSE here so the
+ * regression tests can be loaded with and without
+ * FEATURE_PCRE_HOST_PATTERNS.
+ */
+ }
+#endif
p = strchr(buf, '/');
if (NULL != p)
if (buf[0] != '\0')
{
- return compile_host_pattern(url, buf);
+#ifdef FEATURE_PCRE_HOST_PATTERNS
+ if (url->pattern.url_spec.host_regex_type == PCRE_HOST_PATTERN)
+ {
+ return compile_pcre_host_pattern(url, buf);
+ }
+ else
+#endif
+ {
+ return compile_vanilla_host_pattern(url, buf);
+ }
}
return JB_ERR_OK;
}
-#ifdef FEATURE_EXTENDED_HOST_PATTERNS
+#ifdef FEATURE_PCRE_HOST_PATTERNS
/*********************************************************************
*
- * Function : compile_host_pattern
+ * Function : compile_pcre_host_pattern
*
- * Description : Parses and compiles a host pattern.
+ * Description : Parses and compiles a pcre host pattern.
*
* Parameters :
* 1 : url = Target pattern_spec to be filled in.
* JB_ERR_PARSE - Cannot parse regex
*
*********************************************************************/
-static jb_err compile_host_pattern(struct pattern_spec *url, const char *host_pattern)
+static jb_err compile_pcre_host_pattern(struct pattern_spec *url, const char *host_pattern)
{
return compile_pattern(host_pattern, RIGHT_ANCHORED_HOST, url, &url->pattern.url_spec.host_regex);
}
+#endif /* def FEATURE_PCRE_HOST_PATTERNS */
-#else
/*********************************************************************
*
- * Function : compile_host_pattern
+ * Function : compile_vanilla_host_pattern
*
* Description : Parses and "compiles" an old-school host pattern.
*
* JB_ERR_PARSE - Cannot parse regex
*
*********************************************************************/
-static jb_err compile_host_pattern(struct pattern_spec *url, const char *host_pattern)
+static jb_err compile_vanilla_host_pattern(struct pattern_spec *url, const char *host_pattern)
{
char *v[150];
size_t size;
*/
if (host_pattern[strlen(host_pattern) - 1] == '.')
{
- url->unanchored |= ANCHOR_RIGHT;
+ url->pattern.url_spec.unanchored |= ANCHOR_RIGHT;
}
if (host_pattern[0] == '.')
{
- url->unanchored |= ANCHOR_LEFT;
+ url->pattern.url_spec.unanchored |= ANCHOR_LEFT;
}
/*
* Split domain into components
*/
- url->dbuffer = strdup_or_die(host_pattern);
+ url->pattern.url_spec.dbuffer = strdup_or_die(host_pattern);
/*
* Map to lower case
*/
- for (p = url->dbuffer; *p ; p++)
+ for (p = url->pattern.url_spec.dbuffer; *p ; p++)
{
*p = (char)privoxy_tolower(*p);
}
/*
* Split the domain name into components
*/
- url->dcount = ssplit(url->dbuffer, ".", v, SZ(v));
+ url->pattern.url_spec.dcount = ssplit(url->pattern.url_spec.dbuffer, ".", v, SZ(v));
- if (url->dcount < 0)
+ if (url->pattern.url_spec.dcount < 0)
{
free_pattern_spec(url);
return JB_ERR_PARSE;
}
- else if (url->dcount != 0)
+ else if (url->pattern.url_spec.dcount != 0)
{
/*
* Save a copy of the pointers in dvec
*/
- size = (size_t)url->dcount * sizeof(*url->dvec);
+ size = (size_t)url->pattern.url_spec.dcount * sizeof(*url->pattern.url_spec.dvec);
- url->dvec = malloc_or_die(size);
+ url->pattern.url_spec.dvec = malloc_or_die(size);
- memcpy(url->dvec, v, size);
+ memcpy(url->pattern.url_spec.dvec, v, size);
}
/*
* else dcount == 0 in which case we needn't do anything,
* Function : domain_match
*
* Description : Domain-wise Compare fqdn's. Governed by the bimap in
- * pattern->unachored, the comparison is un-, left-,
+ * p.pattern->unachored, the comparison is un-, left-,
* right-anchored, or both.
* The individual domain names are compared with
* simplematch().
*
* Parameters :
- * 1 : pattern = a domain that may contain a '*' as a wildcard.
+ * 1 : p = a domain that may contain a '*' as a wildcard.
* 2 : fqdn = domain name against which the patterns are compared.
*
* Returns : 0 => domains are equivalent, else no match.
*
*********************************************************************/
-static int domain_match(const struct pattern_spec *pattern, const struct http_request *fqdn)
+static int domain_match(const struct pattern_spec *p, const struct http_request *fqdn)
{
char **pv, **fv; /* vectors */
int plen, flen;
- int unanchored = pattern->unanchored & (ANCHOR_RIGHT | ANCHOR_LEFT);
+ int unanchored = p->pattern.url_spec.unanchored & (ANCHOR_RIGHT | ANCHOR_LEFT);
- plen = pattern->dcount;
+ plen = p->pattern.url_spec.dcount;
flen = fqdn->dcount;
if (flen < plen)
return 1;
}
- pv = pattern->dvec;
+ pv = p->pattern.url_spec.dvec;
fv = fqdn->dvec;
if (unanchored == ANCHOR_LEFT)
}
}
-#endif /* def FEATURE_EXTENDED_HOST_PATTERNS */
/*********************************************************************
const unsigned flag;
} tag_pattern[] = {
{ "TAG:", 4, PATTERN_SPEC_TAG_PATTERN},
+#ifdef FEATURE_CLIENT_TAGS
+ { "CLIENT-TAG:", 11, PATTERN_SPEC_CLIENT_TAG_PATTERN},
+#endif
{ "NO-REQUEST-TAG:", 15, PATTERN_SPEC_NO_REQUEST_TAG_PATTERN},
{ "NO-RESPONSE-TAG:", 16, PATTERN_SPEC_NO_RESPONSE_TAG_PATTERN}
};
if (pattern == NULL) return;
freez(pattern->spec);
-#ifdef FEATURE_EXTENDED_HOST_PATTERNS
+
+ if (!(pattern->flags & PATTERN_SPEC_URL_PATTERN))
+ {
+ if (pattern->pattern.tag_regex)
+ {
+ regfree(pattern->pattern.tag_regex);
+ freez(pattern->pattern.tag_regex);
+ }
+ return;
+ }
+
+#ifdef FEATURE_PCRE_HOST_PATTERNS
if (pattern->pattern.url_spec.host_regex)
{
regfree(pattern->pattern.url_spec.host_regex);
freez(pattern->pattern.url_spec.host_regex);
}
-#else
+#endif /* def FEATURE_PCRE_HOST_PATTERNS */
freez(pattern->pattern.url_spec.dbuffer);
freez(pattern->pattern.url_spec.dvec);
pattern->pattern.url_spec.dcount = 0;
-#endif /* ndef FEATURE_EXTENDED_HOST_PATTERNS */
freez(pattern->pattern.url_spec.port_list);
if (pattern->pattern.url_spec.preg)
{
regfree(pattern->pattern.url_spec.preg);
freez(pattern->pattern.url_spec.preg);
}
- if (pattern->pattern.tag_regex)
- {
- regfree(pattern->pattern.tag_regex);
- freez(pattern->pattern.tag_regex);
- }
}
static int host_matches(const struct http_request *http,
const struct pattern_spec *pattern)
{
-#ifdef FEATURE_EXTENDED_HOST_PATTERNS
- return ((NULL == pattern->pattern.url_spec.host_regex)
- || (0 == regexec(pattern->pattern.url_spec.host_regex, http->host, 0, NULL, 0)));
-#else
- return ((NULL == pattern->pattern.url_spec.dbuffer) || (0 == domain_match(pattern, http)));
+ assert(http->host != NULL);
+#ifdef FEATURE_PCRE_HOST_PATTERNS
+ if (pattern->pattern.url_spec.host_regex_type == PCRE_HOST_PATTERN)
+ {
+ return ((NULL == pattern->pattern.url_spec.host_regex)
+ || (0 == regexec(pattern->pattern.url_spec.host_regex,
+ http->host, 0, NULL, 0)));
+ }
#endif
+ return ((NULL == pattern->pattern.url_spec.dbuffer) || (0 == domain_match(pattern, http)));
}
*
* Function : parse_forwarder_address
*
- * Description : Parse out the host and port from a forwarder address.
+ * Description : Parse out the username, password, host and port from
+ * a forwarder address.
*
* Parameters :
* 1 : address = The forwarder address to parse.
* 2 : hostname = Used to return the hostname. NULL on error.
* 3 : port = Used to return the port. Untouched if no port
* is specified.
+ * 4 : username = Used to return the username if any.
+ * 5 : password = Used to return the password if any.
*
* Returns : JB_ERR_OK on success
* JB_ERR_MEMORY on out of memory
* JB_ERR_PARSE on malformed address.
*
*********************************************************************/
-jb_err parse_forwarder_address(char *address, char **hostname, int *port)
+jb_err parse_forwarder_address(char *address, char **hostname, int *port,
+ char **username, char **password)
{
- char *p = address;
+ char *p;
+ char *tmp;
+
+ tmp = *hostname = strdup_or_die(address);
- if ((*address == '[') && (NULL == strchr(address, ']')))
+ /* Parse username and password */
+ if (username && password && (NULL != (p = strchr(*hostname, '@'))))
+ {
+ *p++ = '\0';
+ *username = strdup_or_die(*hostname);
+ *hostname = strdup_or_die(p);
+
+ if (NULL != (p = strchr(*username, ':')))
+ {
+ *p++ = '\0';
+ *password = strdup_or_die(p);
+ }
+ freez(tmp);
+ }
+
+ /* Parse hostname and port */
+ p = *hostname;
+ if ((*p == '[') && (NULL == strchr(p, ']')))
{
/* XXX: Should do some more validity checks here. */
return JB_ERR_PARSE;
}
- *hostname = strdup_or_die(address);
-
if ((**hostname == '[') && (NULL != (p = strchr(*hostname, ']'))))
{
*p++ = '\0';