resolver.h

一个开源SIP协议栈

/* $Id: resolver.h 974 2007-02-19 01:13:53Z bennylp $ */
/* 
 * Copyright (C) 2003-2007 Benny Prijono <[e-mail]>
 *
 * 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.
 *
 * 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.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, write to the Free Software
 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA 
 */
#ifndef __PJLIB_UTIL_RESOLVER_H__
#define __PJLIB_UTIL_RESOLVER_H__

/**
 * @file resolver.h
 * @brief Asynchronous DNS resolver
 */
#include <pjlib-util/dns.h>


PJ_BEGIN_DECL


/**
 * @defgroup PJ_DNS_RESOLVER DNS Asynchronous/Caching Resolution Engine
 * @ingroup PJ_DNS
 * @{
 *
 * This module manages the host/server resolution by performing asynchronous
 * DNS queries and caching the results in the cache. It uses PJLIB-UTIL 
 * low-level DNS parsing functions (see @ref PJ_DNS) and currently supports
 * several types of DNS resource records such as A record (typical query with
 * gethostbyname()) and SRV record.
 *
 * \section PJ_DNS_RESOLVER_FEATURES Features
 *
 * \subsection PJ_DNS_RESOLVER_FEATURES_ASYNC Asynchronous Query and Query Aggregation
 * 
 * The DNS queries are performed asychronously, with timeout setting 
 * configured on per resolver instance basis. Application can issue multiple
 * asynchronous queries simultaneously. Subsequent queries to the same resource
 * (name and DNS resource type) while existing query is still pending will be
 * merged into one query, so that only one DNS request packet is issued.
 * 
 * \subsection PJ_DNS_RESOLVER_FEATURES_RETRANSMISSION Query Retransmission
 *
 * Asynchronous query will be retransmitted if no response is received
 * within the preconfigured time. Once maximum retransmission count is
 * exceeded and no response is received, the query will time out and the
 * callback will be called when error status.
 *
 * \subsection PJ_DNS_RESOLVER_FEATURES_CACHING Response Caching with TTL
 *
 * The resolver instance caches the results returned by nameservers, to
 * enhance the performance by minimizing the message round-trip to the server.
 * The TTL of the cached resposne is calculated from minimum TTL value found 
 * across all resource record (RR) TTL in the response and further more it can
 * be limited to some preconfigured maximum TTL in the resolver. 
 *
 * Response caching can be  disabled by setting the maximum TTL value of the 
 * resolver to zero.
 *
 * \subsection PJ_DNS_RESOLVER_FEATURES_PARALLEL Parallel and Backup Name Servers
 *
 * When the resolver is configured with multiple nameservers, initially the
 * queries will be issued to multiple name servers simultaneously to probe
 * which servers are not active. Once the probing stage is done, subsequent 
 * queries will be directed to only one ACTIVE server which provides the best
 * response time.
 *
 * Name servers are probed periodically to see which nameservers are active
 * and which are down. This probing is done when a query is sent, thus no
 * timer is needed to maintain this. Also probing will be done in parallel
 * so that there would be no additional delay for the query.
 *
 *
 * \subsection PJ_DNS_RESOLVER_FEATURES_REC Supported Resource Records
 *
 * The low-level DNS parsing utility (see @ref PJ_DNS) supports parsing of
 * the following DNS resource records (RR):
 *  - DNS A record
 *  - DNS SRV record
 *  - DNS PTR record
 *  - DNS NS record
 *  - DNS CNAME record
 *
 * For other types of record, application can parse the raw resource 
 * record data (rdata) from the parsed DNS packet (#pj_dns_parsed_packet).
 *
 *
 * \section PJ_DNS_RESOLVER_USING Using the Resolver
 *
 * To use the resolver, application first creates the resolver instance by
 * calling #pj_dns_resolver_create(). If application already has its own
 * timer and ioqueue instances, it can instruct the resolver to use these
 * instances so that application does not need to poll the resolver 
 * periodically to process events. If application does not specify the
 * timer and ioqueue instance for the resolver, an internal timer and
 * ioqueue will be created by the resolver. And since the resolver does not
 * create it's own thread, application MUST poll the resolver periodically
 * by calling #pj_dns_resolver_handle_events() to allow events (network and 
 * timer) to be processed.
 *
 * Next, application MUST configure the nameservers to be used by the
 * resolver, by calling #pj_dns_resolver_set_ns().
 *
 * Application performs asynchronous query by submitting the query with
 * #pj_dns_resolver_start_query(). Once the query completes (either 
 * successfully or times out), the callback will be called.
 *
 * Application can cancel a pending query by calling #pj_dns_resolver_cancel_query().
 *
 * Resolver must be destroyed by calling #pj_dns_resolver_destroy() to
 * release all resources back to the system.
 *
 *
 * \section PJ_DNS_RESOLVER_LIMITATIONS Resolver Limitations
 *
 * Current implementation mainly suffers from a growing memory problem,
 * which mainly is caused by the response caching. Although there is only
 * one cache entry per {query, name} combination, these cache entry will
 * never get deleted since there is no timer is created to invalidate these
 * entries. So the more unique names being queried by application, there more
 * enties will be created in the response cache.
 *
 * Note that a single response entry will occupy about 600-700 bytes of 
 * pool memory (the PJ_DNS_RESOLVER_RES_BUF_SIZE value plus internal
 * structure). 
 *
 * Application can work around this problem by doing one of these:
 *  - disable caching by setting PJ_DNS_RESOLVER_MAX_TTL and 
 *    PJ_DNS_RESOLVER_INVALID_TTL to zero.
 *  - periodically query #pj_dns_resolver_get_cached_count() and destroy-
 *    recreate the resolver to recycle the memory used by the resolver.
 *
 * Note that future improvement may solve this problem by introducing 
 * expiration timer to the cached entries.
 *
 *
 * \section PJ_DNS_RESOLVER_REFERENCE Reference
 *
 * The PJLIB-UTIL resolver was built from the information in the following
 * standards:
 *  - <A HREF="http://www.faqs.org/rfcs/rfc1035.html">
 *    RFC 1035: "Domain names - implementation and specification"</A>
 *  - <A HREF="http://www.faqs.org/rfcs/rfc2782.html">
 *    RFC 2782: "A DNS RR for specifying the location of services (DNS SRV)"
 *    </A>
 */



/**
 * Opaque data type for DNS resolver object.
 */
typedef struct pj_dns_resolver pj_dns_resolver;

/**
 * Opaque data type for asynchronous DNS query object.
 */
typedef struct pj_dns_async_query pj_dns_async_query;

/**
 * Type of asynchronous callback which will be called when the asynchronous
 * query completes.
 *
 * @param user_data	The user data set by application when creating the
 *			asynchronous query.
 * @param status	Status of the DNS resolution.
 * @param response	The response packet received from the server. This
 *			argument may be NULL when status is not PJ_SUCCESS.
 */
typedef void pj_dns_callback(void *user_data,
			     pj_status_t status,
			     pj_dns_parsed_packet *response);


/**
 * This structure describes resolver settings.
 */
typedef struct pj_dns_settings
{
    unsigned	options;	/**< Options flags.			    */
    unsigned	qretr_delay;	/**< Query retransmit delay in msec.	    */
    unsigned	qretr_count;	/**< Query maximum retransmission count.    */
    unsigned	cache_max_ttl;	/**< Maximum TTL for cached responses. If the
				     value is zero, caching is disabled.    */
} pj_dns_settings;


/**
 * Create DNS resolver instance. After the resolver is created, application
 * MUST configure the nameservers with #pj_dns_resolver_set_ns().
 *
 * When creating the resolver, application may specify both timer heap
 * and ioqueue instance, so that it doesn't need to poll the resolver
 * periodically.
 *