tcp_subr.c

嵌入式操作系统ECOS的网络开发包

tcp_newtcpcb(inp)
	struct inpcb *inp;
{
	register struct tcpcb *tp;

	tp = malloc(sizeof(*tp), M_PCB, M_NOWAIT);
	if (tp == NULL)
		return ((struct tcpcb *)0);
	bzero((char *) tp, sizeof(struct tcpcb));
	LIST_INIT(&tp->segq);
	tp->t_maxseg = tp->t_maxopd = tcp_mssdflt;

#ifdef TCP_SACK
	tp->sack_disable = tcp_do_sack ? 0 : 1;
#endif
	tp->t_flags = tcp_do_rfc1323 ? (TF_REQ_SCALE|TF_REQ_TSTMP) : 0;
	tp->t_inpcb = inp;
	/*
	 * Init srtt to TCPTV_SRTTBASE (0), so we can tell that we have no
	 * rtt estimate.  Set rttvar so that srtt + 2 * rttvar gives
	 * reasonable initial retransmit time.
	 */
	tp->t_srtt = TCPTV_SRTTBASE;
	tp->t_rttvar = tcp_rttdflt * PR_SLOWHZ << (TCP_RTTVAR_SHIFT + 2 - 1);
	tp->t_rttmin = TCPTV_MIN;
	TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
	    TCPTV_MIN, TCPTV_REXMTMAX);
	tp->snd_cwnd = TCP_MAXWIN << TCP_MAX_WINSHIFT;
	tp->snd_ssthresh = TCP_MAXWIN << TCP_MAX_WINSHIFT;
#ifdef INET6
	/*
	 * If we want to use tp->pf for a quick-n-easy way to determine
	 * the outbound dgram type, we cannot make this decision
	 * until a connection is established!  Bzero() sets pf to zero, and
	 * that's the way we want it, unless, of course, it's an AF_INET
	 * socket...
	 */
	if ((inp->inp_flags & INP_IPV6) == 0)
		tp->pf = PF_INET;  /* If AF_INET socket, we can't do v6 from it. */
#else
	tp->pf = PF_INET;
#endif

#ifdef INET6
	if (inp->inp_flags & INP_IPV6) 
		inp->inp_ipv6.ip6_hlim = ip6_defhlim;
	else
#endif /* INET6 */
		inp->inp_ip.ip_ttl = ip_defttl;

	inp->inp_ppcb = (caddr_t)tp;
	return (tp);
}

/*
 * Drop a TCP connection, reporting
 * the specified error.  If connection is synchronized,
 * then send a RST to peer.
 */
struct tcpcb *
tcp_drop(tp, errno)
	register struct tcpcb *tp;
	int errno;
{
	struct socket *so = tp->t_inpcb->inp_socket;

	if (TCPS_HAVERCVDSYN(tp->t_state)) {
		tp->t_state = TCPS_CLOSED;
		(void) tcp_output(tp);
		tcpstat.tcps_drops++;
	} else
		tcpstat.tcps_conndrops++;
	if (errno == ETIMEDOUT && tp->t_softerror)
		errno = tp->t_softerror;
	so->so_error = errno;
	return (tcp_close(tp));
}

/*
 * Close a TCP control block:
 *	discard all space held by the tcp
 *	discard internet protocol block
 *	wake up any sleepers
 */
struct tcpcb *
tcp_close(tp)
	register struct tcpcb *tp;
{
	register struct ipqent *qe;
	struct inpcb *inp = tp->t_inpcb;
	struct socket *so = inp->inp_socket;
#ifdef TCP_SACK
	struct sackhole *p, *q;
#endif
#ifdef RTV_RTT
	register struct rtentry *rt;
#ifdef INET6
	register int bound_to_specific = 0;  /* I.e. non-default */

	/*
	 * This code checks the nature of the route for this connection.
	 * Normally this is done by two simple checks in the next
	 * INET/INET6 ifdef block, but because of two possible lower layers,
	 * that check is done here.
	 *
	 * Perhaps should be doing this only for a RTF_HOST route.
	 */
	rt = inp->inp_route.ro_rt;  /* Same for route or route6. */
	if (tp->pf == PF_INET6) {
		if (rt)
			bound_to_specific =
			    !(IN6_IS_ADDR_UNSPECIFIED(&
			    ((struct sockaddr_in6 *)rt_key(rt))->sin6_addr));
	} else {
		if (rt)
			bound_to_specific =
			    (((struct sockaddr_in *)rt_key(rt))->
			    sin_addr.s_addr != INADDR_ANY);
	}
#endif /* INET6 */

	/*
	 * If we sent enough data to get some meaningful characteristics,
	 * save them in the routing entry.  'Enough' is arbitrarily 
	 * defined as the sendpipesize (default 4K) * 16.  This would
	 * give us 16 rtt samples assuming we only get one sample per
	 * window (the usual case on a long haul net).  16 samples is
	 * enough for the srtt filter to converge to within 5% of the correct
	 * value; fewer samples and we could save a very bogus rtt.
	 *
	 * Don't update the default route's characteristics and don't
	 * update anything that the user "locked".
	 */
#ifdef INET6
	/*
	 * Note that rt and bound_to_specific are set above.
	 */
	if (SEQ_LT(tp->iss + so->so_snd.sb_hiwat * 16, tp->snd_max) &&
	    rt && bound_to_specific) {
#else /* INET6 */
	if (SEQ_LT(tp->iss + so->so_snd.sb_hiwat * 16, tp->snd_max) &&
	    (rt = inp->inp_route.ro_rt) &&
	    satosin(rt_key(rt))->sin_addr.s_addr != INADDR_ANY) {
#endif /* INET6 */
		register u_long i = 0;

		if ((rt->rt_rmx.rmx_locks & RTV_RTT) == 0) {
			i = tp->t_srtt *
			    (RTM_RTTUNIT / (PR_SLOWHZ * TCP_RTT_SCALE));
			if (rt->rt_rmx.rmx_rtt && i)
				/*
				 * filter this update to half the old & half
				 * the new values, converting scale.
				 * See route.h and tcp_var.h for a
				 * description of the scaling constants.
				 */
				rt->rt_rmx.rmx_rtt =
				    (rt->rt_rmx.rmx_rtt + i) / 2;
			else
				rt->rt_rmx.rmx_rtt = i;
		}
		if ((rt->rt_rmx.rmx_locks & RTV_RTTVAR) == 0) {
			i = tp->t_rttvar *
			    (RTM_RTTUNIT / (PR_SLOWHZ * TCP_RTTVAR_SCALE));
			if (rt->rt_rmx.rmx_rttvar && i)
				rt->rt_rmx.rmx_rttvar =
				    (rt->rt_rmx.rmx_rttvar + i) / 2;
			else
				rt->rt_rmx.rmx_rttvar = i;
		}
		/*
		 * update the pipelimit (ssthresh) if it has been updated
		 * already or if a pipesize was specified & the threshhold
		 * got below half the pipesize.  I.e., wait for bad news
		 * before we start updating, then update on both good
		 * and bad news.
		 */
		if (((rt->rt_rmx.rmx_locks & RTV_SSTHRESH) == 0 &&
		    (i = tp->snd_ssthresh) && rt->rt_rmx.rmx_ssthresh) ||
		    i < (rt->rt_rmx.rmx_sendpipe / 2)) {
			/*
			 * convert the limit from user data bytes to
			 * packets then to packet data bytes.
			 */
			i = (i + tp->t_maxseg / 2) / tp->t_maxseg;
			if (i < 2)
				i = 2;
#ifdef INET6
			if (tp->pf == PF_INET6)
				i *= (u_long)(tp->t_maxseg + sizeof (struct tcphdr)
				    + sizeof(struct ip6_hdr));
			else
#endif /* INET6 */
				i *= (u_long)(tp->t_maxseg +
				    sizeof (struct tcpiphdr));

			if (rt->rt_rmx.rmx_ssthresh)
				rt->rt_rmx.rmx_ssthresh =
				    (rt->rt_rmx.rmx_ssthresh + i) / 2;
			else
				rt->rt_rmx.rmx_ssthresh = i;
		}
	}
#endif /* RTV_RTT */

	/* free the reassembly queue, if any */
#ifdef INET6
	/* Reassembling TCP segments in v6 might be sufficiently different
	 * to merit two codepaths to free the reasssembly queue.
	 * If an undecided TCP socket, then the IPv4 codepath will be used 
	 * because it won't matter much anyway.
	 */
	if (tp->pf == AF_INET6) {
		while ((qe = tp->segq.lh_first) != NULL) {
			LIST_REMOVE(qe, ipqe_q);
			m_freem(qe->ipqe_m);
			FREE(qe, M_IPQ);
		}
	} else
#endif /* INET6 */
		while ((qe = tp->segq.lh_first) != NULL) {
			LIST_REMOVE(qe, ipqe_q);
			m_freem(qe->ipqe_m);
			FREE(qe, M_IPQ);
		}
#ifdef TCP_SACK
	/* Free SACK holes. */
	q = p = tp->snd_holes;
	while (p != 0) {
		q = p->next;
		free(p, M_PCB);
		p = q;
	}
#endif
	if (tp->t_template)
		(void) m_free(tp->t_template);
	free(tp, M_PCB);
	inp->inp_ppcb = 0;
	soisdisconnected(so);
	in_pcbdetach(inp);
	tcpstat.tcps_closed++;
	return ((struct tcpcb *)0);
}

void
tcp_drain()
{

}

/*
 * Notify a tcp user of an asynchronous error;
 * store error as soft error, but wake up user
 * (for now, won't do anything until can select for soft error).
 */
void
tcp_notify(inp, error)
	struct inpcb *inp;
	int error;
{
	register struct tcpcb *tp = (struct tcpcb *)inp->inp_ppcb;
	register struct socket *so = inp->inp_socket;

	/*
	 * Ignore some errors if we are hooked up.
	 * If connection hasn't completed, has retransmitted several times,
	 * and receives a second error, give up now.  This is better
	 * than waiting a long time to establish a connection that
	 * can never complete.
	 */
	if (tp->t_state == TCPS_ESTABLISHED &&
	     (error == EHOSTUNREACH || error == ENETUNREACH ||
	      error == EHOSTDOWN)) {
		return;
	} else if (TCPS_HAVEESTABLISHED(tp->t_state) == 0 &&
	    tp->t_rxtshift > 3 && tp->t_softerror)
		so->so_error = error;
	else 
		tp->t_softerror = error;
	wakeup((caddr_t) &so->so_timeo);
	sorwakeup(so);
	sowwakeup(so);
}

#if defined(INET6) && !defined(TCP6)
void
tcp6_ctlinput(cmd, sa, d)
	int cmd;
	struct sockaddr *sa;
	void *d;
{
	(void)tcp_ctlinput(cmd, sa, NULL);	/*XXX*/
}
#endif

void *
tcp_ctlinput(cmd, sa, v)
	int cmd;
	struct sockaddr *sa;
	register void *v;
{
	register struct ip *ip = v;
	register struct tcphdr *th;
	extern int inetctlerrmap[];
	void (*notify) __P((struct inpcb *, int)) = tcp_notify;
	int errno;

	if ((unsigned)cmd >= PRC_NCMDS)
		return NULL;
	errno = inetctlerrmap[cmd];
	if (cmd == PRC_QUENCH)
		notify = tcp_quench;
	else if (PRC_IS_REDIRECT(cmd))
		notify = in_rtchange, ip = 0;
	else if (cmd == PRC_HOSTDEAD)
		ip = 0;
	else if (errno == 0)
		return NULL;

#ifdef INET6
	if (sa->sa_family == AF_INET6) {
		if (ip) {
			struct ip6_hdr *ipv6 = (struct ip6_hdr *)ip;

			th = (struct tcphdr *)(ipv6 + 1);
#if 0 /*XXX*/
			in6_pcbnotify(&tcbtable, sa, th->th_dport,
			    &ipv6->ip6_src, th->th_sport, cmd, notify);
#endif
		} else {
#if 0 /*XXX*/
			in6_pcbnotify(&tcbtable, sa, 0,
			    (struct in6_addr *)&in6addr_any, 0, cmd, notify);
#endif
		}
	} else
#endif /* INET6 */
	{
		if (ip) {
			th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
			in_pcbnotify(&tcbtable, sa, th->th_dport, ip->ip_src,
			    th->th_sport, errno, notify);
		} else
			in_pcbnotifyall(&tcbtable, sa, errno, notify);
	}
	return NULL;
}

/*
 * When a source quench is received, close congestion window
 * to one segment.  We will gradually open it again as we proceed.
 */
void
tcp_quench(inp, errno)
	struct inpcb *inp;
	int errno;
{
	struct tcpcb *tp = intotcpcb(inp);

	if (tp)
		tp->snd_cwnd = tp->t_maxseg;
}

#ifdef TCP_SIGNATURE
int
tcp_signature_tdb_attach()
{
	return (0);
}

int
tcp_signature_tdb_init(tdbp, xsp, ii)
	struct tdb *tdbp;
	struct xformsw *xsp;
	struct ipsecinit *ii;
{
	char *c;
#define isdigit(c)	  (((c) >= '0') && ((c) <= '9'))
#define isalpha(c)	( (((c) >= 'A') && ((c) <= 'Z')) || \
			  (((c) >= 'a') && ((c) <= 'z')) )

	if ((ii->ii_authkeylen < 1) || (ii->ii_authkeylen > 80))
		return (EINVAL);

	c = (char *)ii->ii_authkey;

	while (c < (char *)ii->ii_authkey + ii->ii_authkeylen - 1) {
		if (isdigit(*c)) {
			if (*(c + 1) == ' ')
				return (EINVAL);
		} else {
			if (!isalpha(*c))
				return (EINVAL);
		}

		c++;
	}

	if (!isdigit(*c) && !isalpha(*c))
		return (EINVAL);

	tdbp->tdb_amxkey = malloc(ii->ii_authkeylen, M_XDATA, M_DONTWAIT);
	if (tdbp->tdb_amxkey == NULL)
		return (ENOMEM);
	bcopy(ii->ii_authkey, tdbp->tdb_amxkey, ii->ii_authkeylen);
	tdbp->tdb_amxkeylen = ii->ii_authkeylen;

	return (0);
}

int
tcp_signature_tdb_zeroize(tdbp)
	struct tdb *tdbp;
{
	if (tdbp->tdb_amxkey) {
		bzero(tdbp->tdb_amxkey, tdbp->tdb_amxkeylen);
		free(tdbp->tdb_amxkey, M_XDATA);
		tdbp->tdb_amxkey = NULL;
	}

	return (0);
}

struct mbuf *
tcp_signature_tdb_input(m, tdbp)
	struct mbuf *m;
	struct tdb *tdbp;
{
	return (0);
}

int
tcp_signature_tdb_output(m, tdbp, mp)
	struct mbuf *m;
	struct tdb *tdbp;
	struct mbuf **mp;
{
	return (EINVAL);
}

int
tcp_signature_apply(fstate, data, len)
	caddr_t fstate;
	caddr_t data;
	unsigned int len;
{
	MD5Update((MD5_CTX *)fstate, (char *)data, len);
	return 0;
}
#endif /* TCP_SIGNATURE */