starter: Add new bison/flex based parser for ipsec.conf

The parser simply returns key/value pairs of all sections, it already
resolves also= and allows overriding options in all included sections
(not only %default), options set in included section can also be cleared
again (key=).
It provides other improvements too, like quoted strings (with escape
sequences), unlimited includes and better whitespace/comment handling.

src/starter/.gitignore

@@ -1,4 +1,7 @@
 starter
+parser/lexer.c
+parser/parser.[ch]
+parser/parser.output
 lexer.c
 parser.h
 parser.c

src/starter/Android.mk

@@ -3,11 +3,12 @@ include $(CLEAR_VARS)
 
 # copy-n-paste from Makefile.am (update for LEX/YACC)
 starter_SOURCES := \
-parser.c lexer.c ipsec-parser.h netkey.c args.h netkey.h \
-starterstroke.c confread.c \
-starterstroke.h confread.h args.c \
-keywords.c files.h keywords.h cmp.c starter.c cmp.h invokecharon.c \
-invokecharon.h klips.c klips.h
+starter.c files.h \
+parser/parser.c parser/lexer.c parser/conf_parser.c parser/conf_parser.h \
+parser.c lexer.c ipsec-parser.h args.c args.h \
+confread.c confread.h keywords.c keywords.h cmp.c cmp.h \
+invokecharon.c invokecharon.h starterstroke.c starterstroke.h \
+netkey.c netkey.h klips.c klips.h
 
 LOCAL_SRC_FILES := $(filter %.c,$(starter_SOURCES))
 
@@ -16,6 +17,7 @@ LOCAL_SRC_FILES := $(filter %.c,$(starter_SOURCES))
 LOCAL_C_INCLUDES += \
 	$(strongswan_PATH)/src/libhydra \
 	$(strongswan_PATH)/src/libstrongswan \
+	$(strongswan_PATH)/src/starter \
 	$(strongswan_PATH)/src/stroke
 
 LOCAL_CFLAGS := $(strongswan_CFLAGS) -DSTART_CHARON \

src/starter/Makefile.am

@@ -1,15 +1,17 @@
 ipsec_PROGRAMS = starter
 starter_SOURCES = \
-parser.y lexer.l ipsec-parser.h netkey.c args.h netkey.h \
-starterstroke.c confread.c \
-starterstroke.h confread.h args.c \
-keywords.c files.h keywords.h cmp.c starter.c cmp.h invokecharon.c \
-invokecharon.h klips.c klips.h
+starter.c files.h \
+parser/parser.y parser/lexer.l parser/conf_parser.c parser/conf_parser.h \
+parser.y lexer.l ipsec-parser.h args.c args.h \
+confread.c confread.h keywords.c keywords.h cmp.c cmp.h \
+invokecharon.c invokecharon.h starterstroke.c starterstroke.h \
+netkey.c netkey.h klips.c klips.h
 
 AM_CPPFLAGS = \
 	-I${linux_headers} \
 	-I$(top_srcdir)/src/libstrongswan \
 	-I$(top_srcdir)/src/libhydra \
+	-I$(top_srcdir)/src/starter \
 	-I$(top_srcdir)/src/stroke \
 	-DIPSEC_DIR=\"${ipsecdir}\" \
 	-DIPSEC_CONFDIR=\"${sysconfdir}\" \
@@ -23,10 +25,14 @@ AM_CPPFLAGS = \
 
 AM_YFLAGS = -v -d
 
-starter_LDADD = $(top_builddir)/src/libstrongswan/libstrongswan.la $(top_builddir)/src/libhydra/libhydra.la $(SOCKLIB) $(PTHREADLIB)
+starter_LDADD = \
+	$(top_builddir)/src/libstrongswan/libstrongswan.la \
+	$(top_builddir)/src/libhydra/libhydra.la \
+	$(SOCKLIB) $(PTHREADLIB)
+
 EXTRA_DIST = keywords.txt ipsec.conf Android.mk
 MAINTAINERCLEANFILES = keywords.c
-BUILT_SOURCES = parser.h
+BUILT_SOURCES = keywords.c parser.h parser/parser.h
 
 if USE_CHARON
   AM_CPPFLAGS += -DSTART_CHARON

src/starter/parser/conf_parser.c

@@ -0,0 +1,655 @@
+/*
+ * Copyright (C) 2013-2014 Tobias Brunner
+ * Hochschule fuer Technik Rapperswil
+ *
+ * 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 your
+ * option) any later version.  See <http://www.fsf.org/copyleft/gpl.txt>.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * for more details.
+ */
+
+#include "conf_parser.h"
+
+#include <collections/array.h>
+#include <collections/hashtable.h>
+
+/**
+ * Provided by the generated parser
+ */
+bool conf_parser_parse_file(conf_parser_t *this, char *file);
+
+typedef struct private_conf_parser_t private_conf_parser_t;
+typedef struct section_t section_t;
+
+/**
+ * Private data
+ */
+struct private_conf_parser_t {
+
+	/**
+	 * Public interface
+	 */
+	conf_parser_t public;
+
+	/**
+	 * Path to config file
+	 */
+	char *file;
+
+	/**
+	 * TRUE while parsing the config file
+	 */
+	bool parsing;
+
+	/**
+	 * Hashtable for ca sections, as section_t
+	 */
+	hashtable_t *cas;
+
+	/**
+	 * Hashtable for conn sections, as section_t
+	 */
+	hashtable_t *conns;
+
+	/**
+	 * Array to keep track of the order of conn sections, as section_t
+	 */
+	array_t *conns_order;
+
+	/**
+	 * config setup section
+	 */
+	section_t *config_setup;
+
+	/**
+	 * Pointer to the current section (the one added last) during parsing
+	 */
+	section_t *current_section;
+
+	/**
+	 * Refcount for this parser instance (also used by dictionaries)
+	 */
+	refcount_t ref;
+};
+
+typedef struct {
+	/** Key of the setting */
+	char *key;
+	/** Value of the setting */
+	char *value;
+} setting_t;
+
+int setting_find(const void *a, const void *b)
+{
+	const char *key = a;
+	const setting_t *setting = b;
+	return strcmp(key, setting->key);
+}
+
+int setting_sort(const void *a, const void *b, void *user)
+{
+	const setting_t *sa = a, *sb = b;
+	return strcmp(sa->key, sb->key);
+}
+
+static void setting_destroy(setting_t *this)
+{
+	free(this->key);
+	free(this->value);
+	free(this);
+}
+
+struct section_t {
+	/** Name of the section */
+	char *name;
+	/** Sorted array of settings, as setting_t */
+	array_t *settings;
+	/** Array of also= settings (in reversed order, i.e. most important to least
+	 * important), as setting_t */
+	array_t *also;
+	/** Array of linearized parent objects derived from also= settings, in their
+	 * order of importance (most to least, i.e. %default) as section_t
+	 * NULL if not yet determined */
+	array_t *parents;
+};
+
+static section_t *section_create(char *name)
+{
+	section_t *this;
+
+	INIT(this,
+		.name = name,
+	);
+	return this;
+}
+
+static void section_destroy(section_t *this)
+{
+	array_destroy_function(this->settings, (void*)setting_destroy, NULL);
+	array_destroy_function(this->also, (void*)setting_destroy, NULL);
+	array_destroy(this->parents);
+	free(this->name);
+	free(this);
+}
+
+typedef struct {
+	/** Public interface */
+	dictionary_t public;
+	/** Parser object */
+	private_conf_parser_t *parser;
+	/** Section object */
+	section_t *section;
+} section_dictionary_t;
+
+typedef struct {
+	/** Public interface */
+	enumerator_t public;
+	/** Current settings enumerator */
+	enumerator_t *settings;
+	/** Enumerator into parent list */
+	enumerator_t *parents;
+	/** Hashtable to keep track of already enumerated settings */
+	hashtable_t *seen;
+} dictionary_enumerator_t;
+
+METHOD(enumerator_t, dictionary_enumerate, bool,
+	dictionary_enumerator_t *this, char **key, char **value)
+{
+	setting_t *setting;
+	section_t *parent;
+
+	while (TRUE)
+	{
+		if (this->settings &&
+			this->settings->enumerate(this->settings, &setting))
+		{
+			if (this->seen->get(this->seen, setting->key))
+			{
+				continue;
+			}
+			this->seen->put(this->seen, setting->key, setting->key);
+			if (!setting->value)
+			{
+				continue;
+			}
+			if (key)
+			{
+				*key = setting->key;
+			}
+			if (value)
+			{
+				*value = setting->value;
+			}
+			return TRUE;
+		}
+		DESTROY_IF(this->settings);
+		this->settings = NULL;
+		if (this->parents &&
+			this->parents->enumerate(this->parents, &parent))
+		{
+			if (parent->settings)
+			{
+				this->settings = array_create_enumerator(parent->settings);
+			}
+			continue;
+		}
+		DESTROY_IF(this->parents);
+		this->parents = NULL;
+		break;
+	}
+	return FALSE;
+}
+
+METHOD(enumerator_t, dictionary_enumerator_destroy, void,
+	dictionary_enumerator_t *this)
+{
+	DESTROY_IF(this->settings);
+	DESTROY_IF(this->parents);
+	this->seen->destroy(this->seen);
+	free(this);
+}
+
+METHOD(dictionary_t, dictionary_create_enumerator, enumerator_t*,
+	section_dictionary_t *this)
+{
+	dictionary_enumerator_t *enumerator;
+
+	INIT(enumerator,
+		.public = {
+			.enumerate = (void*)_dictionary_enumerate,
+			.destroy = _dictionary_enumerator_destroy,
+		},
+		.seen = hashtable_create(hashtable_hash_str, hashtable_equals_str, 8),
+	);
+	if (this->section->settings)
+	{
+		enumerator->settings = array_create_enumerator(this->section->settings);
+	}
+	if (this->section->parents)
+	{
+		enumerator->parents = array_create_enumerator(this->section->parents);
+	}
+	return &enumerator->public;
+}
+
+METHOD(dictionary_t, dictionary_get, void*,
+	section_dictionary_t *this, const void *key)
+{
+	enumerator_t *parents;
+	section_t *section;
+	setting_t *setting;
+	char *value = NULL;
+
+	section = this->section;
+	if (array_bsearch(section->settings, key, setting_find, &setting) != -1)
+	{
+		return setting->value;
+	}
+	parents = array_create_enumerator(section->parents);
+	while (parents->enumerate(parents, &section))
+	{
+		if (array_bsearch(section->settings, key, setting_find, &setting) != -1)
+		{
+			value = setting->value;
+			break;
+		}
+	}
+	parents->destroy(parents);
+	return value;
+}
+
+METHOD(dictionary_t, dictionary_destroy, void,
+	section_dictionary_t *this)
+{
+	this->parser->public.destroy(&this->parser->public);
+	free(this);
+}
+
+static dictionary_t *section_dictionary_create(private_conf_parser_t *parser,
+											   section_t *section)
+{
+	section_dictionary_t *this;
+
+	INIT(this,
+		.public = {
+			.create_enumerator = _dictionary_create_enumerator,
+			.get = _dictionary_get,
+			.destroy = _dictionary_destroy,
+		},
+		.parser = parser,
+		.section = section,
+	);
+
+	ref_get(&parser->ref);
+
+	return &this->public;
+}
+
+static bool conn_filter(void *unused, section_t **section, char **name)
+{
+	*name = (*section)->name;
+	return TRUE;
+}
+
+static bool ca_filter(void *unused, void *key, char **name, section_t **section)
+{
+	*name = (*section)->name;
+	return TRUE;
+}
+
+METHOD(conf_parser_t, get_sections, enumerator_t*,
+	private_conf_parser_t *this, conf_parser_section_t type)
+{
+	switch (type)
+	{
+		case CONF_PARSER_CONN:
+			return enumerator_create_filter(
+						array_create_enumerator(this->conns_order),
+						(void*)conn_filter, NULL, NULL);
+		case CONF_PARSER_CA:
+			return enumerator_create_filter(
+						this->cas->create_enumerator(this->cas),
+						(void*)ca_filter, NULL, NULL);
+		case CONF_PARSER_CONFIG_SETUP:
+		default:
+			return enumerator_create_empty();
+	}
+}
+
+METHOD(conf_parser_t, get_section, dictionary_t*,
+	private_conf_parser_t *this, conf_parser_section_t type, char *name)
+{
+	section_t *section = NULL;
+
+	switch (type)
+	{
+		case CONF_PARSER_CONFIG_SETUP:
+			section = this->config_setup;
+			break;
+		case CONF_PARSER_CONN:
+			section = this->conns->get(this->conns, name);
+			break;
+		case CONF_PARSER_CA:
+			section = this->cas->get(this->cas, name);
+			break;
+		default:
+			break;
+	}
+	return section ? section_dictionary_create(this, section) : NULL;
+}
+
+METHOD(conf_parser_t, add_section, bool,
+	private_conf_parser_t *this, conf_parser_section_t type, char *name)
+{
+	hashtable_t *sections = this->conns;
+	array_t *order = this->conns_order;
+	section_t *section = NULL;
+	bool exists = FALSE;
+
+	if (!this->parsing)
+	{
+		free(name);
+		return exists;
+	}
+	switch (type)
+	{
+		case CONF_PARSER_CONFIG_SETUP:
+			section = this->config_setup;
+			/* we don't expect a name, but just in case */
+			free(name);
+			break;
+		case CONF_PARSER_CA:
+			sections = this->cas;
+			order = NULL;
+			/* fall-through */
+		case CONF_PARSER_CONN:
+			section = sections->get(sections, name);
+			if (!section)
+			{
+				section = section_create(name);
+				sections->put(sections, name, section);
+				if (order)
+				{
+					array_insert(order, ARRAY_TAIL, section);
+				}
+			}
+			else
+			{
+				exists = TRUE;
+				free(name);
+			}
+			break;
+
+	}
+	this->current_section = section;
+	return exists;
+}
+
+METHOD(conf_parser_t, add_setting, void,
+	private_conf_parser_t *this, char *key, char *value)
+{
+	section_t *section = this->current_section;
+	setting_t *setting;
+
+	if (!this->parsing || !this->current_section)
+	{
+		free(key);
+		free(value);
+		return;
+	}
+	if (streq(key, "also"))
+	{
+		if (!value || !strlen(value) || streq(value, "%default"))
+		{	/* we require a name, but all sections inherit from %default */
+			free(key);
+			free(value);
+			return;
+		}
+		INIT(setting,
+			.key = key,
+			.value = value,
+		);
+		array_insert_create(&section->also, ARRAY_HEAD, setting);
+		return;
+	}
+	if (array_bsearch(section->settings, key, setting_find, &setting) == -1)
+	{
+		INIT(setting,
+			.key = key,
+			.value = value,
+		);
+		array_insert_create(&section->settings, ARRAY_TAIL, setting);
+		array_sort(section->settings, setting_sort, NULL);
+	}
+	else
+	{
+		free(setting->value);
+		setting->value = value;
+		free(key);
+	}
+}
+
+/**
+ * Check if the given section is contained in the given array.  The search
+ * starts at the given index.
+ */
+static bool is_contained_in(array_t *arr, section_t *section)
+{
+	section_t *current;
+	int i;
+
+	for (i = 0; i < array_count(arr); i++)
+	{
+		array_get(arr, i, &current);
+		if (streq(section->name, current->name))
+		{
+			return TRUE;
+		}
+	}
+	return FALSE;
+}
+
+/**
+ * This algorithm to linearize sections uses a bottom-first depth-first
+ * semantic, with an additional elimination step that removes all but the
+ * last occurrence of each section.
+ *
+ * Consider this configuration:
+ *
+ *   conn A
+ *   conn B
+ *     also=A
+ *   conn C
+ *     also=A
+ *   conn D
+ *     also=C
+ *     also=B
+ *
+ * The linearization would yield D B A C A, which gets reduced to D B C A.
+ *
+ * Ambiguous configurations are handled pragmatically.
+ *
+ * Consider the following configuration:
+ *
+ *   conn A
+ *   conn B
+ *   conn C
+ *     also=A
+ *     also=B
+ *   conn D
+ *     also=B
+ *     also=A
+ *   conn E
+ *     also=C
+ *     also=D
+ *
+ * It is ambiguous because D and C include the same two sections but in
+ * a different order.
+ *
+ * The linearization would yield E D A B C B A which gets reduced to E D C B A.
+ */
+static bool resolve_also_single(hashtable_t *sections,
+							section_t *section, section_t *def, array_t *stack)
+{
+	enumerator_t *enumerator;
+	array_t *parents;
+	section_t *parent, *grandparent;
+	setting_t *also;
+	bool success = TRUE;
+	int i;
+
+	array_insert(stack, ARRAY_HEAD, section);
+	parents = array_create(0, 0);
+	enumerator = array_create_enumerator(section->also);
+	while (enumerator->enumerate(enumerator, &also))
+	{
+		parent = sections->get(sections, also->value);
+		if (!parent || is_contained_in(stack, parent))
+		{
+			if (!parent)
+			{
+				DBG1(DBG_CFG, "section '%s' referenced in section '%s' not "
+					 "found", also->value, section->name);
+			}
+			else
+			{
+				DBG1(DBG_CFG, "section '%s' referenced in section '%s' causes "
+					 "a loop", parent->name, section->name);
+			}
+			array_remove_at(section->also, enumerator);
+			setting_destroy(also);
+			success = FALSE;
+			continue;
+		}
+		if (!parent->parents)
+		{
+			if (!resolve_also_single(sections, parent, def, stack))
+			{
+				success = FALSE;
+				continue;
+			}
+		}
+		/* add the grandparents and the parent to the list */
+		array_insert(parents, ARRAY_TAIL, parent);
+		for (i = 0; i < array_count(parent->parents); i++)
+		{
+			array_get(parent->parents, i, &grandparent);
+			array_insert(parents, ARRAY_TAIL, grandparent);
+		}
+	}
+	enumerator->destroy(enumerator);
+	array_remove(stack, ARRAY_HEAD, NULL);
+
+	if (success && def && !array_count(parents))
+	{
+		array_insert(parents, ARRAY_TAIL, def);
+	}
+
+	while (success && array_remove(parents, ARRAY_HEAD, &parent))
+	{
+		if (!is_contained_in(parents, parent))
+		{	/* last occurrence of this section */
+			array_insert_create(&section->parents, ARRAY_TAIL, parent);
+		}
+	}
+	array_destroy(parents);
+	return success;
+}
+
+/**
+ * Resolve also= statements. The functions returns TRUE if everything is fine,
+ * or FALSE if either a referenced section does not exist, or if the section
+ * inheritance can't be determined properly (e.g. if there are loops or if a
+ * section inherits from multiple sections - perhaps over several levels - in
+ * an ambiguous way).
+ */
+static bool resolve_also(hashtable_t *sections)
+{
+	enumerator_t *enumerator;
+	section_t *def, *section;
+	array_t *stack;
+	bool success = TRUE;
+
+	stack = array_create(0, 0);
+
+	def = sections->get(sections, "%default");
+	if (def)
+	{	/* the default section is the only one with an empty parents list */
+		def->parents = array_create(0, 0);
+	}
+	enumerator = sections->create_enumerator(sections);
+	while (enumerator->enumerate(enumerator, NULL, &section))
+	{
+		if (section->parents)
+		{	/* already determined */
+			continue;
+		}
+		success = resolve_also_single(sections, section, def, stack) && success;
+	}
+	enumerator->destroy(enumerator);
+	array_destroy(stack);
+	return success;
+}
+
+METHOD(conf_parser_t, parse, bool,
+	private_conf_parser_t *this)
+{
+	bool success;
+
+	if (!this->file)
+	{	/* no file, lets assume this is OK */
+		return TRUE;
+	}
+	this->parsing = TRUE;
+	success = conf_parser_parse_file(&this->public, this->file);
+	this->parsing = FALSE;
+	return success && resolve_also(this->conns) && resolve_also(this->cas);
+}
+
+METHOD(conf_parser_t, destroy, void,
+	private_conf_parser_t *this)
+{
+	if (ref_put(&this->ref))
+	{
+		this->cas->destroy_function(this->cas, (void*)section_destroy);
+		this->conns->destroy_function(this->conns, (void*)section_destroy);
+		section_destroy(this->config_setup);
+		array_destroy(this->conns_order);
+		free(this->file);
+		free(this);
+	}
+}
+
+/*
+ * Described in header
+ */
+conf_parser_t *conf_parser_create(const char *file)
+{
+	private_conf_parser_t *this;
+
+	INIT(this,
+		.public = {
+			.parse = _parse,
+			.get_sections = _get_sections,
+			.get_section = _get_section,
+			.add_section = _add_section,
+			.add_setting = _add_setting,
+			.destroy = _destroy,
+		},
+		.file = strdupnull(file),
+		.cas = hashtable_create(hashtable_hash_str,
+								hashtable_equals_str, 8),
+		.conns = hashtable_create(hashtable_hash_str,
+								  hashtable_equals_str, 8),
+		.conns_order = array_create(0, 0),
+		.config_setup = section_create(NULL),
+		.ref = 1,
+	);
+
+	return &this->public;
+}

src/starter/parser/conf_parser.h

@@ -0,0 +1,120 @@
+/*
+ * Copyright (C) 2013-2014 Tobias Brunner
+ * Hochschule fuer Technik Rapperswil
+ *
+ * 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 your
+ * option) any later version.  See <http://www.fsf.org/copyleft/gpl.txt>.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * for more details.
+ */
+
+/**
+ * @defgroup conf_parser conf_parser
+ * @{ @ingroup starter
+ */
+
+#ifndef CONF_PARSER_H_
+#define CONF_PARSER_H_
+
+#include <library.h>
+#include <collections/dictionary.h>
+
+typedef enum conf_parser_section_t conf_parser_section_t;
+typedef struct conf_parser_t conf_parser_t;
+
+/**
+ * Type of section
+ */
+enum conf_parser_section_t {
+	/**
+	 * config setup
+	 */
+	CONF_PARSER_CONFIG_SETUP,
+
+	/**
+	 * conn <name>
+	 */
+	CONF_PARSER_CONN,
+
+	/**
+	 * ca <name>
+	 */
+	CONF_PARSER_CA,
+};
+
+/**
+ * Parser for ipsec.conf
+ */
+struct conf_parser_t {
+
+	/**
+	 * Parse the config file.
+	 *
+	 * @return		TRUE if config file was parsed successfully
+	 */
+	bool (*parse)(conf_parser_t *this);
+
+	/**
+	 * Get the names of all sections of the given type.
+	 *
+	 * @note Returns an empty enumerator for the config setup section.
+	 *
+	 * @return		enumerator over char*
+	 */
+	enumerator_t *(*get_sections)(conf_parser_t *this,
+								  conf_parser_section_t type);
+
+	/**
+	 * Get the section with the given type and name.
+	 *
+	 * @note The name is ignored for the config setup section.
+	 *
+	 * @return		dictionary with settings
+	 */
+	dictionary_t *(*get_section)(conf_parser_t *this,
+								 conf_parser_section_t type, char *name);
+
+	/**
+	 * Add a section while parsing.
+	 *
+	 * @note This method can only be called while parsing the config file.
+	 *
+	 * @param type	type of section to add
+	 * @param name	name of the section, if applicable (gets adopted)
+	 * @return		TRUE if the section already existed (settings get added)
+	 */
+	bool (*add_section)(conf_parser_t *this, conf_parser_section_t type,
+						char *name);
+
+	/**
+	 * Add a key/value pair to the latest section.
+	 *
+	 * @note This method can only be called while parsing the config file.
+	 *
+	 * @param name	key string (gets adopted)
+	 * @param value	optional value string (gets adopted), if no value is
+	 * 				specified the key is set empty
+	 */
+	void (*add_setting)(conf_parser_t *this, char *key, char *value);
+
+
+	/**
+	 * Destroy a conf_parser_t instance.
+	 */
+	void (*destroy)(conf_parser_t *this);
+};
+
+/**
+ * Create a conf_parser_t instance.
+ *
+ * @param file		ipsec.conf file to parse (gets copied)
+ * @return			conf_parser_t instance
+ */
+conf_parser_t *conf_parser_create(const char *file);
+
+#endif /** CONF_PARSER_H_ @}*/

src/starter/parser/lexer.l

@@ -0,0 +1,205 @@
+%{
+/*
+ * Copyright (C) 2013-2014 Tobias Brunner
+ * Hochschule fuer Technik Rapperswil
+ *
+ * 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 your
+ * option) any later version.  See <http://www.fsf.org/copyleft/gpl.txt>.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * for more details.
+ */
+
+#include <utils/parser_helper.h>
+#include <parser/conf_parser.h>
+
+#include "parser.h"
+
+bool conf_parser_open_next_file(parser_helper_t *ctx);
+
+static void include_files(parser_helper_t *ctx);
+
+%}
+%option debug
+%option warn
+
+/* use start conditions stack */
+%option stack
+
+/* do not declare unneded functions */
+%option noinput noyywrap
+
+/* don't use global variables, and interact properly with bison */
+%option reentrant bison-bridge
+
+/* maintain the line number */
+%option yylineno
+
+/* don't generate a default rule */
+%option nodefault
+
+/* prefix function/variable declarations */
+%option prefix="conf_parser_"
+/* don't change the name of the output file otherwise autotools has issues */
+%option outfile="lex.yy.c"
+
+/* type of our extra data */
+%option extra-type="parser_helper_t*"
+
+/* state used to scan include file patterns */
+%x inc
+/* state used to scan quoted strings */
+%x str
+
+%%
+
+^[\t ]*"version"[^\n]*$	/* eat legacy version delcaration */
+^[\t ]+					return SPACES;
+[\t ]+					/* eat other whitespace */
+[\t ]*#[^\n]*			/* eat comments */
+
+\n						return NEWLINE;
+
+"="						return EQ;
+^"config setup"			return CONFIG_SETUP;
+^"conn"					return CONN;
+^"ca"					return CA;
+
+"include"[\t ]+/[^=]	{
+	yyextra->string_init(yyextra);
+	yy_push_state(inc, yyscanner);
+}
+
+"\""					{
+	yyextra->string_init(yyextra);
+	yy_push_state(str, yyscanner);
+}
+
+[^\"#= \t\n]+			{
+	yylval->s = strdup(yytext);
+	return STRING;
+}
+
+<inc>{
+	/* we allow all characters except # and spaces, they can be escaped */
+	<<EOF>>				|
+	[#\n\t ]			{
+		if (*yytext)
+		{
+			switch (yytext[0])
+			{
+				case '\n':
+					/* put the newline back to fix the line numbers */
+					unput('\n');
+					yy_set_bol(0);
+					break;
+				case '#':
+					/* comments are parsed outside of this start condition */
+					unput(yytext[0]);
+					break;
+			}
+		}
+		include_files(yyextra);
+		yy_pop_state(yyscanner);
+	}
+	"\""				{	/* string include */
+		yy_push_state(str, yyscanner);
+	}
+	\\					{
+		yyextra->string_add(yyextra, yytext);
+	}
+	\\["#} ]			{
+		yyextra->string_add(yyextra, yytext+1);
+	}
+	[^"\\#\n\t ]+ {
+		yyextra->string_add(yyextra, yytext);
+	}
+}
+
+<str>{
+	"\""				|
+	<<EOF>>				|
+	\n					|
+	\\					{
+		if (!streq(yytext, "\""))
+		{
+			if (streq(yytext, "\n"))
+			{	/* put the newline back to fix the line numbers */
+				unput('\n');
+				yy_set_bol(0);
+			}
+			PARSER_DBG1(yyextra, "unterminated string detected");
+		}
+		if (yy_top_state(yyscanner) == inc)
+		{	/* string include */
+			include_files(yyextra);
+			yy_pop_state(yyscanner);
+			yy_pop_state(yyscanner);
+		}
+		else
+		{
+			yy_pop_state(yyscanner);
+			yylval->s = yyextra->string_get(yyextra);
+			return STRING;
+		}
+	}
+	\\n     yyextra->string_add(yyextra, "\n");
+	\\r     yyextra->string_add(yyextra, "\r");
+	\\t     yyextra->string_add(yyextra, "\t");
+	\\b     yyextra->string_add(yyextra, "\b");
+	\\f     yyextra->string_add(yyextra, "\f");
+	\\(.|\n)			{
+		yyextra->string_add(yyextra, yytext+1);
+	}
+	[^\\\n"]+			{
+		yyextra->string_add(yyextra, yytext);
+	}
+}
+
+<<EOF>>					{
+	conf_parser_pop_buffer_state(yyscanner);
+	if (!conf_parser_open_next_file(yyextra) && !YY_CURRENT_BUFFER)
+	{
+		yyterminate();
+	}
+}
+
+%%
+
+/**
+ * Open the next file, if any is queued and readable, otherwise returns FALSE.
+ */
+bool conf_parser_open_next_file(parser_helper_t *ctx)
+{
+	FILE *file;
+
+	file = ctx->file_next(ctx);
+	if (!file)
+	{
+		return FALSE;
+	}
+
+	conf_parser_set_in(file, ctx->scanner);
+	conf_parser_push_buffer_state(
+			conf_parser__create_buffer(file, YY_BUF_SIZE,
+									   ctx->scanner), ctx->scanner);
+	return TRUE;
+}
+
+/**
+ * Assumes that the file pattern to include is currently stored as string on
+ * the helper object.
+ */
+static void include_files(parser_helper_t *ctx)
+{
+	char *pattern = ctx->string_get(ctx);
+
+	ctx->file_include(ctx, pattern);
+	free(pattern);
+
+	conf_parser_open_next_file(ctx);
+}

src/starter/parser/parser.y

@@ -0,0 +1,254 @@
+%{
+/*
+ * Copyright (C) 2013-2014 Tobias Brunner
+ * Hochschule fuer Technik Rapperswil
+ *
+ * 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 your
+ * option) any later version.  See <http://www.fsf.org/copyleft/gpl.txt>.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
+ * or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
+ * for more details.
+ */
+
+#define _GNU_SOURCE /* for asprintf() */
+#include <stdio.h>
+
+#include <utils/parser_helper.h>
+#include <settings/settings_types.h>
+#include <parser/conf_parser.h>
+
+#include "parser.h"
+
+#define YYDEBUG 1
+
+/**
+ * Defined by the lexer
+ */
+int conf_parser_lex(YYSTYPE *lvalp, void *scanner);
+int conf_parser_lex_init_extra(parser_helper_t *extra, void *scanner);
+int conf_parser_lex_destroy(void *scanner);
+int conf_parser_set_in(FILE *in, void *scanner);
+void conf_parser_set_debug(int debug, void *scanner);
+char *conf_parser_get_text(void *scanner);
+int conf_parser_get_leng(void *scanner);
+int conf_parser_get_lineno(void *scanner);
+/* Custom functions in lexer */
+bool conf_parser_open_next_file(parser_helper_t *ctx);
+
+/**
+ * Forward declaration
+ */
+static void conf_parser_error(parser_helper_t *ctx, const char *s);
+
+/**
+ * Make sure to call lexer with the proper context
+ */
+#undef yylex
+static int yylex(YYSTYPE *lvalp, parser_helper_t *ctx)
+{
+	return conf_parser_lex(lvalp, ctx->scanner);
+}
+
+%}
+%debug
+
+/* generate verbose error messages */
+%error-verbose
+/* generate a reentrant parser */
+%define api.pure
+/* prefix function/variable declarations */
+%name-prefix "conf_parser_"
+
+/* interact properly with the reentrant lexer */
+%lex-param {parser_helper_t *ctx}
+%parse-param {parser_helper_t *ctx}
+
+/* types for terminal symbols... */
+%union {
+	char *s;
+	conf_parser_section_t t;
+}
+%token <s> STRING
+%token EQ SPACES NEWLINE CONFIG_SETUP CONN CA
+
+/* ...and other symbols */
+%type <t> section_type
+%type <s> section_name value
+
+/* make the equal sign left associative */
+%left EQ
+
+/* properly destroy STRING tokens, which are strdup()ed, on errors */
+%destructor { free($$); } STRING section_name value
+
+/* there are two shift/reduce conflicts because we allow empty lines (and lines
+ * with spaces) within settings and anywhere else (i.e. in the beginning) */
+//%expect 2
+
+%%
+
+/**
+ * ipsec.conf grammar rules
+ */
+statements:
+	/* empty */
+	| statements NEWLINE
+	| statements statement
+	;
+
+statement:
+	section
+	| SPACES setting
+	;
+
+section:
+	section_type section_name
+	{
+		if ($1 != CONF_PARSER_CONFIG_SETUP && (!$2 || !strlen($2)))
+		{
+			PARSER_DBG1(ctx, "section name missing");
+			free($2);
+			YYERROR;
+		}
+		conf_parser_t *parser = (conf_parser_t*)ctx->context;
+		parser->add_section(parser, $1, $2);
+	}
+	;
+
+section_type:
+	CONFIG_SETUP
+	{
+		$$ = CONF_PARSER_CONFIG_SETUP;
+	}
+	|
+	CONN
+	{
+		$$ = CONF_PARSER_CONN;
+	}
+	|
+	CA
+	{
+		$$ = CONF_PARSER_CA;
+	}
+	;
+
+section_name:
+	/* empty */
+	{
+		$$ = NULL;
+	}
+	| STRING
+	{
+		$$ = $1;
+	}
+	;
+
+setting:
+	/* empty */
+	|
+	STRING EQ value
+	{
+		if (!strlen($1))
+		{
+			PARSER_DBG1(ctx, "setting name can't be empty");
+			free($1);
+			free($3);
+			YYERROR;
+		}
+		conf_parser_t *parser = (conf_parser_t*)ctx->context;
+		parser->add_setting(parser, $1, $value);
+	}
+	|
+	STRING EQ
+	{
+		if (!strlen($1))
+		{
+			PARSER_DBG1(ctx, "setting name can't be empty");
+			free($1);
+			YYERROR;
+		}
+		conf_parser_t *parser = (conf_parser_t*)ctx->context;
+		parser->add_setting(parser, $1, NULL);
+	}
+	|
+	STRING
+	{
+		PARSER_DBG1(ctx, "missing value for setting '%s'", $1);
+		free($1);
+		YYERROR;
+	}
+	;
+
+value:
+	STRING
+	| value STRING
+	{	/* just put a single space between them, use strings for more */
+		if (asprintf(&$$, "%s %s", $1, $2) < 0)
+		{
+			free($1);
+			free($2);
+			YYERROR;
+		}
+		free($1);
+		free($2);
+	}
+	;
+
+%%
+
+/**
+ * Referenced by the generated parser
+ */
+static void conf_parser_error(parser_helper_t *ctx, const char *s)
+{
+	char *text = conf_parser_get_text(ctx->scanner);
+	int len = conf_parser_get_leng(ctx->scanner);
+
+	if (len && text[len-1] == '\n')
+	{	/* cut off newline at the end to avoid muti-line log messages */
+		len--;
+	}
+	PARSER_DBG1(ctx, "%s [%.*s]", s, (int)len, text);
+}
+
+/**
+ * Parse the given file
+ */
+bool conf_parser_parse_file(conf_parser_t *this, char *name)
+{
+	parser_helper_t *helper;
+	bool success = FALSE;
+
+	helper = parser_helper_create(this);
+	helper->get_lineno = conf_parser_get_lineno;
+	if (conf_parser_lex_init_extra(helper, &helper->scanner) != 0)
+	{
+		helper->destroy(helper);
+		return FALSE;
+	}
+	helper->file_include(helper, name);
+	if (!conf_parser_open_next_file(helper))
+	{
+		DBG1(DBG_CFG, "failed to open config file '%s'", name);
+	}
+	else
+	{
+		if (getenv("DEBUG_CONF_PARSER"))
+		{
+			yydebug = 1;
+			conf_parser_set_debug(1, helper->scanner);
+		}
+		success = yyparse(helper) == 0;
+		if (!success)
+		{
+			DBG1(DBG_CFG, "invalid config file '%s'", name);
+		}
+	}
+	conf_parser_lex_destroy(helper->scanner);
+	helper->destroy(helper);
+	return success;
+}