tcp_input.c

嵌入式操作系统ECOS的网络开发包

		m->m_pkthdr.tdbi = NULL;
	}
#endif /* IPSEC */
#ifdef INET6
	/*
	 * Before we do ANYTHING, we have to figure out if it's TCP/IPv6 or
	 * TCP/IPv4.
 	 */
	is_ipv6 = mtod(m, struct ip *)->ip_v == 6;
#endif /* INET6 */

	/*
	 * Get IP and TCP header together in first mbuf.
	 * Note: IP leaves IP header in first mbuf.
	 */
#ifndef INET6
	ti = mtod(m, struct tcpiphdr *);
#else /* INET6 */
	if (!is_ipv6)
#endif /* INET6 */
	if (iphlen > sizeof (struct ip)) {
#if 0	/*XXX*/
		ip_stripoptions(m, (struct mbuf *)0);
#else
#ifdef __ECOS
		diag_printf("extension headers are not allowed\n");
#else
		printf("extension headers are not allowed\n");
#endif
		m_freem(m);
		return;
#endif
	}
	if (m->m_len < iphlen + sizeof(struct tcphdr)) {
		if ((m = m_pullup2(m, iphlen + sizeof(struct tcphdr))) == 0) {
			tcpstat.tcps_rcvshort++;
			return;
		}
#ifndef INET6
		ti = mtod(m, struct tcpiphdr *);
#endif /* INET6 */
	}

	tlen = m->m_pkthdr.len - iphlen;

#ifdef INET6
	/*
	 * After that, do initial segment processing which is still very
	 * dependent on what IP version you're using.
	 */

	if (is_ipv6) {
#ifdef DIAGNOSTIC
	  if (iphlen < sizeof(struct ip6_hdr)) {
	    m_freem(m);
	    return;
	  }
#endif /* DIAGNOSTIC */

	  /* strip off any options */
	  if (iphlen > sizeof(struct ip6_hdr)) {
#if 0 /*XXX*/
	    ipv6_stripoptions(m, iphlen);
#else
#ifdef __ECOS
		diag_printf("extension headers are not allowed\n");
#else
		printf("extension headers are not allowed\n");
#endif
		m_freem(m);
		return;
#endif
	    iphlen = sizeof(struct ip6_hdr);
	  }

	  ti = NULL;
	  ipv6 = mtod(m, struct ip6_hdr *);

	  if (in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), tlen)) {
	    tcpstat.tcps_rcvbadsum++;
	    goto drop;
	  } /* endif in6_cksum */
	} else {
	  ti = mtod(m, struct tcpiphdr *);
#endif /* INET6 */

	/*
	 * Checksum extended TCP header and data.
	 */
#ifndef INET6
	tlen = ((struct ip *)ti)->ip_len;
#endif /* INET6 */
	len = sizeof (struct ip) + tlen;
	bzero(ti->ti_x1, sizeof ti->ti_x1);
	ti->ti_len = (u_int16_t)tlen;
	HTONS(ti->ti_len);
	if ((ti->ti_sum = in_cksum(m, len)) != 0) {
		tcpstat.tcps_rcvbadsum++;
		goto drop;
	}
#ifdef INET6
	}
#endif /* INET6 */
#endif /* TUBA_INCLUDE */

	th = (struct tcphdr *)(mtod(m, caddr_t) + iphlen);

	/*
	 * Check that TCP offset makes sense,
	 * pull out TCP options and adjust length.		XXX
	 */
	off = th->th_off << 2;
	if (off < sizeof (struct tcphdr) || off > tlen) {
		tcpstat.tcps_rcvbadoff++;
		goto drop;
	}
	tlen -= off;
	if (off > sizeof (struct tcphdr)) {
		if (m->m_len < iphlen + off) {
			if ((m = m_pullup2(m, iphlen + off)) == 0) {
				tcpstat.tcps_rcvshort++;
				return;
			}
#ifdef INET6
			if (is_ipv6)
			  ipv6 = mtod(m, struct ip6_hdr *);
			else
#endif /* INET6 */
			ti = mtod(m, struct tcpiphdr *);
			th = (struct tcphdr *)(mtod(m, caddr_t) + iphlen);
		}
		optlen = off - sizeof (struct tcphdr);
		optp = mtod(m, caddr_t) + iphlen + sizeof(struct tcphdr);
		/* 
		 * Do quick retrieval of timestamp options ("options
		 * prediction?").  If timestamp is the only option and it's
		 * formatted as recommended in RFC 1323 appendix A, we
		 * quickly get the values now and not bother calling
		 * tcp_dooptions(), etc.
		 */
		if ((optlen == TCPOLEN_TSTAMP_APPA ||
		     (optlen > TCPOLEN_TSTAMP_APPA &&
			optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
		     *(u_int32_t *)optp == htonl(TCPOPT_TSTAMP_HDR) &&
		     (th->th_flags & TH_SYN) == 0) {
			ts_present = 1;
			ts_val = ntohl(*(u_int32_t *)(optp + 4));
			ts_ecr = ntohl(*(u_int32_t *)(optp + 8));
			optp = NULL;	/* we've parsed the options */
		}
	}
	tiflags = th->th_flags;

	/*
	 * Convert TCP protocol specific fields to host format.
	 */
	NTOHL(th->th_seq);
	NTOHL(th->th_ack);
	NTOHS(th->th_win);
	NTOHS(th->th_urp);

	/*
	 * Locate pcb for segment.
	 */
findpcb:
#ifdef INET6
	if (is_ipv6) {
	  inp = in6_pcbhashlookup(&tcbtable, &ipv6->ip6_src, th->th_sport,
				 &ipv6->ip6_dst, th->th_dport);
	} else
#endif /* INET6 */
	inp = in_pcbhashlookup(&tcbtable, ti->ti_src, ti->ti_sport,
	    ti->ti_dst, ti->ti_dport);
	if (inp == 0) {
		++tcpstat.tcps_pcbhashmiss;
#ifdef INET6
		if (is_ipv6)
			inp = in_pcblookup(&tcbtable, &ipv6->ip6_src,
			    th->th_sport, &ipv6->ip6_dst, th->th_dport,
			    INPLOOKUP_WILDCARD | INPLOOKUP_IPV6);
		else
#endif /* INET6 */
		inp = in_pcblookup(&tcbtable, &ti->ti_src, ti->ti_sport,
		    &ti->ti_dst, ti->ti_dport, INPLOOKUP_WILDCARD);
		/*
		 * If the state is CLOSED (i.e., TCB does not exist) then
		 * all data in the incoming segment is discarded.
		 * If the TCB exists but is in CLOSED state, it is embryonic,
		 * but should either do a listen or a connect soon.
		 */
		if (inp == 0) {
			++tcpstat.tcps_noport;
			goto dropwithreset;
		}
	}

	tp = intotcpcb(inp);
	if (tp == 0)
		goto dropwithreset;
	if (tp->t_state == TCPS_CLOSED)
		goto drop;
	
	/* Unscale the window into a 32-bit value. */
	if ((tiflags & TH_SYN) == 0)
		tiwin = th->th_win << tp->snd_scale;
	else
		tiwin = th->th_win;

	so = inp->inp_socket;
	if (so->so_options & (SO_DEBUG|SO_ACCEPTCONN)) {
		if (so->so_options & SO_DEBUG) {
			ostate = tp->t_state;
#ifdef INET6
			if (is_ipv6)
			  tcp_saveti6 = *(mtod(m, struct tcpipv6hdr *));
			else
#endif /* INET6 */
			tcp_saveti = *ti;
		}
		if (so->so_options & SO_ACCEPTCONN) {
			struct socket *so1;

			so1 = sonewconn(so, 0);
			if (so1 == NULL) {
				tcpdropoldhalfopen(tp, th->th_dport);
				so1 = sonewconn(so, 0);
				if (so1 == NULL)
					goto drop;
			}
			so = so1;
			/*
			 * This is ugly, but ....
			 *
			 * Mark socket as temporary until we're
			 * committed to keeping it.  The code at
			 * ``drop'' and ``dropwithreset'' check the
			 * flag dropsocket to see if the temporary
			 * socket created here should be discarded.
			 * We mark the socket as discardable until
			 * we're committed to it below in TCPS_LISTEN.
			 */
			dropsocket++;
#ifdef IPSEC
			/* 
			 * We need to copy the required security levels
			 * from the old pcb.
			 */
			{
			  struct inpcb *newinp = (struct inpcb *)so->so_pcb;
			  bcopy(inp->inp_seclevel, newinp->inp_seclevel,
				sizeof(inp->inp_seclevel));
			  newinp->inp_secrequire = inp->inp_secrequire;
			}
#endif /* IPSEC */
#ifdef INET6
			/*
			 * inp still has the OLD in_pcb stuff, set the
			 * v6-related flags on the new guy, too.   This is
			 * done particularly for the case where an AF_INET6
			 * socket is bound only to a port, and a v4 connection
			 * comes in on that port.
			 * we also copy the flowinfo from the original pcb 
			 * to the new one.
			 */
			{
			  int flags = inp->inp_flags;
			  struct inpcb *oldinpcb = inp;
			  
			  inp = (struct inpcb *)so->so_pcb;
			  inp->inp_flags |= (flags & (INP_IPV6 | INP_IPV6_UNDEC
						      | INP_IPV6_MAPPED));
			  if ((inp->inp_flags & INP_IPV6) &&
			      !(inp->inp_flags & INP_IPV6_MAPPED)) {
			    inp->inp_ipv6.ip6_hlim = 
			      oldinpcb->inp_ipv6.ip6_hlim;
			    inp->inp_ipv6.ip6_flow = 
			      oldinpcb->inp_ipv6.ip6_flow;
			  }
			}
#else /* INET6 */
			inp = (struct inpcb *)so->so_pcb;
#endif /* INET6 */
			inp->inp_lport = th->th_dport;
#ifdef INET6
			if (is_ipv6) {
			  inp->inp_laddr6 = ipv6->ip6_dst;
			  inp->inp_fflowinfo = htonl(0x0fffffff) & 
			    ipv6->ip6_flow;
			  
			  /*inp->inp_options = ip6_srcroute();*/ /* soon. */
			  /* still need to tweak outbound options
			     processing to include this mbuf in
			     the right place and put the correct
			     NextHdr values in the right places.
			     XXX  rja */
			} else {
			  if (inp->inp_flags & INP_IPV6) {/* v4 to v6 socket */
			    CREATE_IPV6_MAPPED(inp->inp_laddr6,
			      ti->ti_dst.s_addr);
			  } else {
#endif /* INET6 */
			    inp->inp_laddr = ti->ti_dst;
			    inp->inp_options = ip_srcroute();
#ifdef INET6
			  }
			}
#endif /* INET6 */
			in_pcbrehash(inp);
			tp = intotcpcb(inp);
			tp->t_state = TCPS_LISTEN;

			/* Compute proper scaling value from buffer space
			 */
			while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
			   TCP_MAXWIN << tp->request_r_scale < so->so_rcv.sb_hiwat)
				tp->request_r_scale++;
		}
	}

#ifdef IPSEC
	/* Check if this socket requires security for incoming packets */
	if ((inp->inp_seclevel[SL_AUTH] >= IPSEC_LEVEL_REQUIRE &&
	     !(m->m_flags & M_AUTH)) ||
	    (inp->inp_seclevel[SL_ESP_TRANS] >= IPSEC_LEVEL_REQUIRE &&
	     !(m->m_flags & M_CONF))) {
#ifdef notyet
#ifdef INET6
		if (is_ipv6)
			icmp6_error(m, ICMPV6_BLAH, ICMPV6_BLAH, 0);
		else
#endif /* INET6 */
		icmp_error(m, ICMP_BLAH, ICMP_BLAH, 0, 0);
#endif /* notyet */
		tcpstat.tcps_rcvnosec++;
		goto drop;
	}
	/* Use tdb_bind_out for this inp's outbound communication */
	if (tdb)
		tdb_add_inp(tdb, inp);
#endif /*IPSEC */

	/*
	 * Segment received on connection.
	 * Reset idle time and keep-alive timer.
	 */
	tp->t_idle = 0;
	if (tp->t_state != TCPS_SYN_RECEIVED)
		tp->t_timer[TCPT_KEEP] = tcp_keepidle;

#ifdef TCP_SACK
	if (!tp->sack_disable)
		tcp_del_sackholes(tp, th); /* Delete stale SACK holes */
#endif /* TCP_SACK */

	/*
	 * Process options if not in LISTEN state,
	 * else do it below (after getting remote address).
	 */
	if (optp && tp->t_state != TCPS_LISTEN)
		tcp_dooptions(tp, optp, optlen, th,
			&ts_present, &ts_val, &ts_ecr);

#ifdef TCP_SACK
	if (!tp->sack_disable) {
		tp->rcv_laststart = th->th_seq; /* last rec'vd segment*/
		tp->rcv_lastend = th->th_seq + tlen;
	}
#endif /* TCP_SACK */
	/* 
	 * Header prediction: check for the two common cases
	 * of a uni-directional data xfer.  If the packet has
	 * no control flags, is in-sequence, the window didn't
	 * change and we're not retransmitting, it's a
	 * candidate.  If the length is zero and the ack moved
	 * forward, we're the sender side of the xfer.  Just
	 * free the data acked & wake any higher level process
	 * that was blocked waiting for space.  If the length
	 * is non-zero and the ack didn't move, we're the
	 * receiver side.  If we're getting packets in-order
	 * (the reassembly queue is empty), add the data to
	 * the socket buffer and note that we need a delayed ack.
	 */
	if (tp->t_state == TCPS_ESTABLISHED &&
	    (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK &&
	    (!ts_present || TSTMP_GEQ(ts_val, tp->ts_recent)) &&
	    th->th_seq == tp->rcv_nxt &&
	    tiwin && tiwin == tp->snd_wnd &&
	    tp->snd_nxt == tp->snd_max) {

		/* 
		 * If last ACK falls within this segment's sequence numbers,
		 *  record the timestamp.
		 * Fix from Braden, see Stevens p. 870
		 */
		if (ts_present && SEQ_LEQ(th->th_seq, tp->last_ack_sent)) {
			tp->ts_recent_age = tcp_now;
			tp->ts_recent = ts_val;
		}

		if (tlen == 0) {
			if (SEQ_GT(th->th_ack, tp->snd_una) &&
			    SEQ_LEQ(th->th_ack, tp->snd_max) &&
			    tp->snd_cwnd >= tp->snd_wnd &&
			    tp->t_dupacks == 0) {
				/*
				 * this is a pure ack for outstanding data.
				 */
				++tcpstat.tcps_predack;
				if (ts_present)
					tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
				else if (tp->t_rtt &&
					    SEQ_GT(th->th_ack, tp->t_rtseq))
					tcp_xmit_timer(tp, tp->t_rtt);
				acked = th->th_ack - tp->snd_una;
				tcpstat.tcps_rcvackpack++;
				tcpstat.tcps_rcvackbyte += acked;
				ND6_HINT(tp);
				sbdrop(&so->so_snd, acked);
				tp->snd_una = th->th_ack;
#if defined(TCP_SACK) || defined(TCP_NEWRENO)
				/* 
				 * We want snd_last to track snd_una so
				 * as to avoid sequence wraparound problems
				 * for very large transfers.
				 */
				tp->snd_last = tp->snd_una;
#endif /* TCP_SACK or TCP_NEWRENO */
#if defined(TCP_SACK) && defined(TCP_FACK)
				tp->snd_fack = tp->snd_una;
				tp->retran_data = 0;
#endif /* TCP_FACK */
				m_freem(m);

				/*
				 * If all outstanding data are acked, stop
				 * retransmit timer, otherwise restart timer
				 * using current (possibly backed-off) value.
				 * If process is waiting for space,
				 * wakeup/selwakeup/signal.  If data
				 * are ready to send, let tcp_output
				 * decide between more output or persist.
				 */
				if (tp->snd_una == tp->snd_max)
					tp->t_timer[TCPT_REXMT] = 0;
				else if (tp->t_timer[TCPT_PERSIST] == 0)
					tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;

				if (sb_notify(&so->so_snd))
					sowwakeup(so);
				if (so->so_snd.sb_cc)
					(void) tcp_output(tp);
				return;
			}
		} else if (th->th_ack == tp->snd_una &&
		    tp->segq.lh_first == NULL &&
		    tlen <= sbspace(&so->so_rcv)) {
			/*
			 * This is a pure, in-sequence data packet
			 * with nothing on the reassembly queue and