socket.c

bind 9.3结合mysql数据库

 */
static int
doio_send(isc_socket_t *sock, isc_socketevent_t *dev) {
	int cc;
	struct iovec iov[MAXSCATTERGATHER_SEND];
	size_t write_count;
	struct msghdr msghdr;
	char addrbuf[ISC_SOCKADDR_FORMATSIZE];
	int attempts = 0;
	int send_errno;
	char strbuf[ISC_STRERRORSIZE];

	build_msghdr_send(sock, dev, &msghdr, iov, &write_count);

 resend:
	cc = sendmsg(sock->fd, &msghdr, 0);
	send_errno = errno;

	/*
	 * Check for error or block condition.
	 */
	if (cc < 0) {
		if (send_errno == EINTR && ++attempts < NRETRIES)
			goto resend;

		if (SOFT_ERROR(send_errno))
			return (DOIO_SOFT);

#define SOFT_OR_HARD(_system, _isc) \
	if (send_errno == _system) { \
		if (sock->connected) { \
			dev->result = _isc; \
			return (DOIO_HARD); \
		} \
		return (DOIO_SOFT); \
	}
#define ALWAYS_HARD(_system, _isc) \
	if (send_errno == _system) { \
		dev->result = _isc; \
		return (DOIO_HARD); \
	}

		SOFT_OR_HARD(ECONNREFUSED, ISC_R_CONNREFUSED);
		ALWAYS_HARD(EACCES, ISC_R_NOPERM);
		ALWAYS_HARD(EAFNOSUPPORT, ISC_R_ADDRNOTAVAIL);
		ALWAYS_HARD(EADDRNOTAVAIL, ISC_R_ADDRNOTAVAIL);
		ALWAYS_HARD(EHOSTUNREACH, ISC_R_HOSTUNREACH);
#ifdef EHOSTDOWN
		ALWAYS_HARD(EHOSTDOWN, ISC_R_HOSTUNREACH);
#endif
		ALWAYS_HARD(ENETUNREACH, ISC_R_NETUNREACH);
		ALWAYS_HARD(ENOBUFS, ISC_R_NORESOURCES);
		ALWAYS_HARD(EPERM, ISC_R_HOSTUNREACH);
		ALWAYS_HARD(EPIPE, ISC_R_NOTCONNECTED);
		ALWAYS_HARD(ECONNRESET, ISC_R_CONNECTIONRESET);

#undef SOFT_OR_HARD
#undef ALWAYS_HARD

		/*
		 * The other error types depend on whether or not the
		 * socket is UDP or TCP.  If it is UDP, some errors
		 * that we expect to be fatal under TCP are merely
		 * annoying, and are really soft errors.
		 *
		 * However, these soft errors are still returned as
		 * a status.
		 */
		isc_sockaddr_format(&dev->address, addrbuf, sizeof(addrbuf));
		isc__strerror(send_errno, strbuf, sizeof(strbuf));
		UNEXPECTED_ERROR(__FILE__, __LINE__, "internal_send: %s: %s",
				 addrbuf, strbuf);
		dev->result = isc__errno2result(send_errno);
		return (DOIO_HARD);
	}

	if (cc == 0)
		UNEXPECTED_ERROR(__FILE__, __LINE__,
				 "internal_send: send() %s 0",
				 isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL,
						ISC_MSG_RETURNED, "returned"));

	/*
	 * If we write less than we expected, update counters, poke.
	 */
	dev->n += cc;
	if ((size_t)cc != write_count)
		return (DOIO_SOFT);

	/*
	 * Exactly what we wanted to write.  We're done with this
	 * entry.  Post its completion event.
	 */
	dev->result = ISC_R_SUCCESS;
	return (DOIO_SUCCESS);
}

/*
 * Kill.
 *
 * Caller must ensure that the socket is not locked and no external
 * references exist.
 */
static void
destroy(isc_socket_t **sockp) {
	isc_socket_t *sock = *sockp;
	isc_socketmgr_t *manager = sock->manager;

	socket_log(sock, NULL, CREATION, isc_msgcat, ISC_MSGSET_SOCKET,
		   ISC_MSG_DESTROYING, "destroying");

	INSIST(ISC_LIST_EMPTY(sock->accept_list));
	INSIST(ISC_LIST_EMPTY(sock->recv_list));
	INSIST(ISC_LIST_EMPTY(sock->send_list));
	INSIST(sock->connect_ev == NULL);
	REQUIRE(sock->fd >= 0 && sock->fd < (int)FD_SETSIZE);

	LOCK(&manager->lock);

	/*
	 * No one has this socket open, so the watcher doesn't have to be
	 * poked, and the socket doesn't have to be locked.
	 */
	manager->fds[sock->fd] = NULL;
	manager->fdstate[sock->fd] = CLOSE_PENDING;
	select_poke(manager, sock->fd, SELECT_POKE_CLOSE);
	ISC_LIST_UNLINK(manager->socklist, sock, link);

#ifdef ISC_PLATFORM_USETHREADS
	if (ISC_LIST_EMPTY(manager->socklist))
		SIGNAL(&manager->shutdown_ok);
#endif /* ISC_PLATFORM_USETHREADS */

	/*
	 * XXX should reset manager->maxfd here
	 */

	UNLOCK(&manager->lock);

	free_socket(sockp);
}

static isc_result_t
allocate_socket(isc_socketmgr_t *manager, isc_sockettype_t type,
		isc_socket_t **socketp)
{
	isc_socket_t *sock;
	isc_result_t ret;
	ISC_SOCKADDR_LEN_T cmsgbuflen;

	sock = isc_mem_get(manager->mctx, sizeof(*sock));

	if (sock == NULL)
		return (ISC_R_NOMEMORY);

	ret = ISC_R_UNEXPECTED;

	sock->magic = 0;
	sock->references = 0;

	sock->manager = manager;
	sock->type = type;
	sock->fd = -1;

	ISC_LINK_INIT(sock, link);

	sock->recvcmsgbuf = NULL;
	sock->sendcmsgbuf = NULL;

	/*
	 * set up cmsg buffers
	 */
	cmsgbuflen = 0;
#if defined(USE_CMSG) && defined(ISC_PLATFORM_HAVEIPV6)
	cmsgbuflen = cmsg_space(sizeof(struct in6_pktinfo));
#endif
#if defined(USE_CMSG) && defined(SO_TIMESTAMP)
	cmsgbuflen += cmsg_space(sizeof(struct timeval));
#endif
	sock->recvcmsgbuflen = cmsgbuflen;
	if (sock->recvcmsgbuflen != 0) {
		sock->recvcmsgbuf = isc_mem_get(manager->mctx, cmsgbuflen);
		if (sock->recvcmsgbuf == NULL)
			goto error;
	}

	cmsgbuflen = 0;
#if defined(USE_CMSG) && defined(ISC_PLATFORM_HAVEIPV6)
	cmsgbuflen = cmsg_space(sizeof(struct in6_pktinfo));
#endif
	sock->sendcmsgbuflen = cmsgbuflen;
	if (sock->sendcmsgbuflen != 0) {
		sock->sendcmsgbuf = isc_mem_get(manager->mctx, cmsgbuflen);
		if (sock->sendcmsgbuf == NULL)
			goto error;
	}

	/*
	 * set up list of readers and writers to be initially empty
	 */
	ISC_LIST_INIT(sock->recv_list);
	ISC_LIST_INIT(sock->send_list);
	ISC_LIST_INIT(sock->accept_list);
	sock->connect_ev = NULL;
	sock->pending_recv = 0;
	sock->pending_send = 0;
	sock->pending_accept = 0;
	sock->listener = 0;
	sock->connected = 0;
	sock->connecting = 0;
	sock->bound = 0;

	/*
	 * initialize the lock
	 */
	if (isc_mutex_init(&sock->lock) != ISC_R_SUCCESS) {
		sock->magic = 0;
		UNEXPECTED_ERROR(__FILE__, __LINE__,
				 "isc_mutex_init() %s",
				 isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL,
						ISC_MSG_FAILED, "failed"));
		ret = ISC_R_UNEXPECTED;
		goto error;
	}

	/*
	 * Initialize readable and writable events
	 */
	ISC_EVENT_INIT(&sock->readable_ev, sizeof(intev_t),
		       ISC_EVENTATTR_NOPURGE, NULL, ISC_SOCKEVENT_INTR,
		       NULL, sock, sock, NULL, NULL);
	ISC_EVENT_INIT(&sock->writable_ev, sizeof(intev_t),
		       ISC_EVENTATTR_NOPURGE, NULL, ISC_SOCKEVENT_INTW,
		       NULL, sock, sock, NULL, NULL);

	sock->magic = SOCKET_MAGIC;
	*socketp = sock;

	return (ISC_R_SUCCESS);

 error:
	if (sock->recvcmsgbuf != NULL)
		isc_mem_put(manager->mctx, sock->recvcmsgbuf,
			    sock->recvcmsgbuflen);
	if (sock->sendcmsgbuf != NULL)
		isc_mem_put(manager->mctx, sock->sendcmsgbuf,
			    sock->sendcmsgbuflen);
	isc_mem_put(manager->mctx, sock, sizeof(*sock));

	return (ret);
}

/*
 * This event requires that the various lists be empty, that the reference
 * count be 1, and that the magic number is valid.  The other socket bits,
 * like the lock, must be initialized as well.  The fd associated must be
 * marked as closed, by setting it to -1 on close, or this routine will
 * also close the socket.
 */
static void
free_socket(isc_socket_t **socketp) {
	isc_socket_t *sock = *socketp;

	INSIST(sock->references == 0);
	INSIST(VALID_SOCKET(sock));
	INSIST(!sock->connecting);
	INSIST(!sock->pending_recv);
	INSIST(!sock->pending_send);
	INSIST(!sock->pending_accept);
	INSIST(ISC_LIST_EMPTY(sock->recv_list));
	INSIST(ISC_LIST_EMPTY(sock->send_list));
	INSIST(ISC_LIST_EMPTY(sock->accept_list));
	INSIST(!ISC_LINK_LINKED(sock, link));

	if (sock->recvcmsgbuf != NULL)
		isc_mem_put(sock->manager->mctx, sock->recvcmsgbuf,
			    sock->recvcmsgbuflen);
	if (sock->sendcmsgbuf != NULL)
		isc_mem_put(sock->manager->mctx, sock->sendcmsgbuf,
			    sock->sendcmsgbuflen);

	sock->magic = 0;

	DESTROYLOCK(&sock->lock);

	isc_mem_put(sock->manager->mctx, sock, sizeof(*sock));

	*socketp = NULL;
}

/*
 * Create a new 'type' socket managed by 'manager'.  Events
 * will be posted to 'task' and when dispatched 'action' will be
 * called with 'arg' as the arg value.  The new socket is returned
 * in 'socketp'.
 */
isc_result_t
isc_socket_create(isc_socketmgr_t *manager, int pf, isc_sockettype_t type,
		  isc_socket_t **socketp)
{
	isc_socket_t *sock = NULL;
	isc_result_t ret;
#if defined(USE_CMSG) || defined(SO_BSDCOMPAT)
	int on = 1;
#endif
	char strbuf[ISC_STRERRORSIZE];

	REQUIRE(VALID_MANAGER(manager));
	REQUIRE(socketp != NULL && *socketp == NULL);

	ret = allocate_socket(manager, type, &sock);
	if (ret != ISC_R_SUCCESS)
		return (ret);

	sock->pf = pf;
	switch (type) {
	case isc_sockettype_udp:
		sock->fd = socket(pf, SOCK_DGRAM, IPPROTO_UDP);
		break;
	case isc_sockettype_tcp:
		sock->fd = socket(pf, SOCK_STREAM, IPPROTO_TCP);
		break;
	}

#ifdef F_DUPFD
        /*
         * Leave a space for stdio to work in.
         */
        if (sock->fd >= 0 && sock->fd < 20) {
                int new, tmp;
                new = fcntl(sock->fd, F_DUPFD, 20);
                tmp = errno;
                (void)close(sock->fd);
                errno = tmp;
                sock->fd = new;
        }
#endif

	if (sock->fd >= (int)FD_SETSIZE) {
		(void)close(sock->fd);
		isc_log_iwrite(isc_lctx, ISC_LOGCATEGORY_GENERAL,
			      ISC_LOGMODULE_SOCKET, ISC_LOG_ERROR,
			       isc_msgcat, ISC_MSGSET_SOCKET,
			       ISC_MSG_TOOMANYFDS,
			       "%s: too many open file descriptors", "socket");
		free_socket(&sock);
		return (ISC_R_NORESOURCES);
	}
	
	if (sock->fd < 0) {
		free_socket(&sock);

		switch (errno) {
		case EMFILE:
		case ENFILE:
		case ENOBUFS:
			return (ISC_R_NORESOURCES);

		case EPROTONOSUPPORT:
		case EPFNOSUPPORT:
		case EAFNOSUPPORT:
		/*
		 * Linux 2.2 (and maybe others) return EINVAL instead of
		 * EAFNOSUPPORT.
		 */
		case EINVAL:
			return (ISC_R_FAMILYNOSUPPORT);

		default:
			isc__strerror(errno, strbuf, sizeof(strbuf));
			UNEXPECTED_ERROR(__FILE__, __LINE__,
					 "socket() %s: %s",
					 isc_msgcat_get(isc_msgcat,
							ISC_MSGSET_GENERAL,
							ISC_MSG_FAILED,
							"failed"),
					 strbuf);
			return (ISC_R_UNEXPECTED);
		}
	}

	if (make_nonblock(sock->fd) != ISC_R_SUCCESS) {
		(void)close(sock->fd);
		free_socket(&sock);
		return (ISC_R_UNEXPECTED);
	}

#ifdef SO_BSDCOMPAT
	if (setsockopt(sock->fd, SOL_SOCKET, SO_BSDCOMPAT,
		       (void *)&on, sizeof(on)) < 0) {
		isc__strerror(errno, strbuf, sizeof(strbuf));
		UNEXPECTED_ERROR(__FILE__, __LINE__,
				 "setsockopt(%d, SO_BSDCOMPAT) %s: %s",
				 sock->fd,
				 isc_msgcat_get(isc_msgcat, ISC_MSGSET_GENERAL,
						ISC_MSG_FAILED, "failed"),
				 strbuf);
		/* Press on... */
	}
#endif

#if defined(USE_CMSG)
	if (type == isc_sockettype_udp) {

#if defined(SO_TIMESTAMP)
		if (setsockopt(sock->fd, SOL_SOCKET, SO_TIMESTAMP,
			       (void *)&on, sizeof(on)) < 0
		    && errno != ENOPROTOOPT) {
			isc__strerror(errno, strbuf, sizeof(strbuf));
			UNEXPECTED_ERROR(__FILE__, __LINE__,
					 "setsockopt(%d, SO_TIMESTAMP) %s: %s",
					 sock->fd, 
					 isc_msgcat_get(isc_msgcat,
							ISC_MSGSET_GENERAL,
							ISC_MSG_FAILED,
							"failed"),
					 strbuf);
			/* Press on... */
		}
#endif /* SO_TIMESTAMP */

#if defined(ISC_PLATFORM_HAVEIPV6)
		if (pf == AF_INET6 && sock->recvcmsgbuflen == 0) {
			/*
			 * Warn explicitly because this anomaly can be hidden
			 * in usual operation (and unexpectedly appear later).
			 */
			UNEXPECTED_ERROR(__FILE__, __LINE__,
					 "No buffer available to receive "
					 "IPv6 destination");
		}
#ifdef IPV6_RECVPKTINFO
		/* 2292bis */
		if ((pf == AF_INET6)
		    && (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_RECVPKTINFO,
				   (void *)&on, sizeof(on)) < 0)) {
			isc__strerror(errno, strbuf, sizeof(strbuf));
			UNEXPECTED_ERROR(__FILE__, __LINE__,
					 "setsockopt(%d, IPV6_RECVPKTINFO) "
					 "%s: %s", sock->fd,
					 isc_msgcat_get(isc_msgcat,
							ISC_MSGSET_GENERAL,
							ISC_MSG_FAILED,
							"failed"),
					 strbuf);
		}
#else
		/* 2292 */
		if ((pf == AF_INET6)
		    && (setsockopt(sock->fd, IPPROTO_IPV6, IPV6_PKTINFO,
				   (void *)&on, sizeof(on)) < 0)) {
			isc__strerror(errno, strbuf, sizeof(strbuf));
			UNEXPECTED_ERROR(__FILE__, __LINE__,
					 "setsockopt(%d, IPV6_PKTINFO) %s: %s",
					 sock->fd,
					 isc_msgcat_get(isc_msgcat,
							ISC_MSGSET_GENERAL,
							ISC_MSG_FAILED,
							"failed"),
					 strbuf);
		}
#endif /* IPV6_RECVPKTINFO */
#ifdef IPV6_USE_MIN_MTU        /*2292bis, not too common yet*/
		/* use minimum MTU */
		if (pf == AF_INET6) {
			(void)setsockopt(sock->fd, IPPROTO_IPV6,
					 IPV6_USE_MIN_MTU,
					 (void *)&on, sizeof(on));
		}
#endif
#endif /* ISC_PLATFORM_HAVEIPV6 */

	}
#endif /* USE_CMSG */

	sock->references = 1;
	*socketp = sock;

	LOCK(&manager->lock);

	/*
	 * Note we don't have to lock the socket like we normally would because
	 * there are no external references to it yet.
	 */

	manager->fds[sock->fd] = sock;
	manager->fdstate[sock->fd] = MANAGED;
	ISC_LIST_APPEND(manager->socklist, sock, link);
	if (manager->maxfd < sock->fd)
		manager->maxfd = sock->fd;

	UNLOCK(&manager->lock);

	socket_log(sock, NULL, CREATION, isc_msgcat, ISC_MSGSET_SOCKET,
		   ISC_MSG_CREATED, "created");

	return (ISC_R_SUCCESS);
}

/*
 * Attach to a socket.  Caller must explicitly detach when it is done.
 */
void
isc_socket_attach(isc_socket_t *sock, isc_socket_t **socketp) {
	REQUIRE(VALID_SOCKET(sock));
	REQUIRE(socketp != NULL && *socketp == NULL);

	LOCK(&sock->lock);
	sock->references++;
	UNLOCK(&sock->lock);

	*socketp = sock;
}

/*
 * Dereference a socket.  If this is the last reference to it, clean things
 * up by destroying the socket.
 */
void
isc_socket_detach(isc_socket_t **socketp) {
	isc_socket_t *sock;
	isc_boolean_t kill_socket = ISC_FALSE;

	REQUIRE(socketp != NULL);