tcp_subr.c

来自「基于组件方式开发操作系统的OSKIT源代码」· C语言 代码 · 共 609 行 · 第 1/2 页

	 * If we got enough samples through the srtt filter,
	 * save the rtt and rttvar in the routing entry.
	 * 'Enough' is arbitrarily defined as the 16 samples.
	 * 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".
	 */
	if (tp->t_rttupdated >= 16 &&
	    (rt = inp->inp_route.ro_rt) &&
	    ((struct sockaddr_in *)rt_key(rt))->sin_addr.s_addr != INADDR_ANY) {
		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;
			tcpstat.tcps_cachedrtt++;
		}
		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;
			tcpstat.tcps_cachedrttvar++;
		}
		/*
		 * 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) != 0) && 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;
			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;
			tcpstat.tcps_cachedssthresh++;
		}
	}
#endif /* RTV_RTT */
	/* free the reassembly queue, if any */
	t = tp->seg_next;
	while (t != (struct tcpiphdr *)tp) {
		t = (struct tcpiphdr *)t->ti_next;
		m = REASS_MBUF((struct tcpiphdr *)t->ti_prev);
		remque(t->ti_prev);
		m_freem(m);
	}
	if (tp->t_template)
		(void) m_free(dtom(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 (tp->t_state < TCPS_ESTABLISHED && 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);
}

void
tcp_ctlinput(cmd, sa, ip)
	int cmd;
	struct sockaddr *sa;
	register struct ip *ip;
{
	register struct tcphdr *th;
	void (*notify) __P((struct inpcb *, int)) = tcp_notify;

	if (cmd == PRC_QUENCH)
		notify = tcp_quench;
#if 1
	else if (cmd == PRC_MSGSIZE)
		notify = tcp_mtudisc;
#endif
	else if (!PRC_IS_REDIRECT(cmd) &&
		 ((unsigned)cmd > PRC_NCMDS || inetctlerrmap[cmd] == 0))
		return;
	if (ip) {
		th = (struct tcphdr *)((caddr_t)ip + (ip->ip_hl << 2));
		in_pcbnotify(&tcb, sa, th->th_dport, ip->ip_src, th->th_sport,
			cmd, notify);
	} else
		in_pcbnotify(&tcb, sa, 0, zeroin_addr, 0, cmd, notify);
}

/*
 * 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;
}

#if 1
/*
 * When `need fragmentation' ICMP is received, update our idea of the MSS
 * based on the new value in the route.  Also nudge TCP to send something,
 * since we know the packet we just sent was dropped.
 * This duplicates some code in the tcp_mss() function in tcp_input.c.
 */
void
tcp_mtudisc(inp, errno)
	struct inpcb *inp;
	int errno;
{
	struct tcpcb *tp = intotcpcb(inp);
	struct rtentry *rt;
	struct rmxp_tao *taop;
	struct socket *so = inp->inp_socket;
	int offered;
	int mss;

	if (tp) {
		rt = tcp_rtlookup(inp);
		if (!rt || !rt->rt_rmx.rmx_mtu) {
			tp->t_maxopd = tp->t_maxseg = tcp_mssdflt;
			return;
		}
		taop = rmx_taop(rt->rt_rmx);
		offered = taop->tao_mssopt;
		mss = rt->rt_rmx.rmx_mtu - sizeof(struct tcpiphdr);
		if (offered)
			mss = min(mss, offered);
		/*
		 * XXX - The above conditional probably violates the TCP
		 * spec.  The problem is that, since we don't know the
		 * other end's MSS, we are supposed to use a conservative
		 * default.  But, if we do that, then MTU discovery will
		 * never actually take place, because the conservative
		 * default is much less than the MTUs typically seen
		 * on the Internet today.  For the moment, we'll sweep
		 * this under the carpet.
		 *
		 * The conservative default might not actually be a problem
		 * if the only case this occurs is when sending an initial
		 * SYN with options and data to a host we've never talked
		 * to before.  Then, they will reply with an MSS value which
		 * will get recorded and the new parameters should get
		 * recomputed.  For Further Study.
		 */
		if (tp->t_maxopd <= mss)
			return;
		tp->t_maxopd = mss;

		if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP &&
		    (tp->t_flags & TF_RCVD_TSTMP) == TF_RCVD_TSTMP)
			mss -= TCPOLEN_TSTAMP_APPA;
		if ((tp->t_flags & (TF_REQ_CC|TF_NOOPT)) == TF_REQ_CC &&
		    (tp->t_flags & TF_RCVD_CC) == TF_RCVD_CC)
			mss -= TCPOLEN_CC_APPA;
#if	(MCLBYTES & (MCLBYTES - 1)) == 0
		if (mss > MCLBYTES)
			mss &= ~(MCLBYTES-1);
#else
		if (mss > MCLBYTES)
			mss = mss / MCLBYTES * MCLBYTES;
#endif
		if (so->so_snd.sb_hiwat < mss)
			mss = so->so_snd.sb_hiwat;

		tp->t_maxseg = mss;

		tcpstat.tcps_mturesent++;
		tp->t_rtt = 0;
		tp->snd_nxt = tp->snd_una;
		tcp_output(tp);
	}
}
#endif

/*
 * Look-up the routing entry to the peer of this inpcb.  If no route
 * is found and it cannot be allocated the return NULL.  This routine
 * is called by TCP routines that access the rmx structure and by tcp_mss
 * to get the interface MTU.
 */
struct rtentry *
tcp_rtlookup(inp)
	struct inpcb *inp;
{
	struct route *ro;
	struct rtentry *rt;

	ro = &inp->inp_route;
	rt = ro->ro_rt;
	if (rt == NULL || !(rt->rt_flags & RTF_UP)) {
		/* No route yet, so try to acquire one */
		if (inp->inp_faddr.s_addr != INADDR_ANY) {
			ro->ro_dst.sa_family = AF_INET;
			ro->ro_dst.sa_len = sizeof(ro->ro_dst);
			((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
				inp->inp_faddr;
			rtalloc(ro);
			rt = ro->ro_rt;
		}
	}
	return rt;
}

/*
 * Return a pointer to the cached information about the remote host.
 * The cached information is stored in the protocol specific part of
 * the route metrics.
 */
struct rmxp_tao *
tcp_gettaocache(inp)
	struct inpcb *inp;
{
	struct rtentry *rt = tcp_rtlookup(inp);

	/* Make sure this is a host route and is up. */
	if (rt == NULL ||
	    (rt->rt_flags & (RTF_UP|RTF_HOST)) != (RTF_UP|RTF_HOST))
		return NULL;

	return rmx_taop(rt->rt_rmx);
}

/*
 * Clear all the TAO cache entries, called from tcp_init.
 *
 * XXX
 * This routine is just an empty one, because we assume that the routing
 * routing tables are initialized at the same time when TCP, so there is
 * nothing in the cache left over.
 */
static void
tcp_cleartaocache(void)
{ }