tcp_subr.c

来自「eCos操作系统源码」· C语言 代码 · 共 1,397 行 · 第 1/3 页

	if (cmd == PRC_QUENCH)
		notify = tcp_quench;
	else if (icmp_may_rst && (cmd == PRC_UNREACH_ADMIN_PROHIB ||
		cmd == PRC_UNREACH_PORT) && ip)
		notify = tcp_drop_syn_sent;
	else if (cmd == PRC_MSGSIZE)
		notify = tcp_mtudisc;
	else if (PRC_IS_REDIRECT(cmd)) {
		ip = 0;
		notify = in_rtchange;
	} else if (cmd == PRC_HOSTDEAD)
		ip = 0;
	else if ((unsigned)cmd > PRC_NCMDS || inetctlerrmap[cmd] == 0)
		return;
	if (ip) {
		s = splnet();
		th = (struct tcphdr *)((caddr_t)ip 
				       + (IP_VHL_HL(ip->ip_vhl) << 2));
		inp = in_pcblookup_hash(&tcbinfo, faddr, th->th_dport,
		    ip->ip_src, th->th_sport, 0, NULL);
		if (inp != NULL && inp->inp_socket != NULL) {
			icmp_seq = htonl(th->th_seq);
			tp = intotcpcb(inp);
			if (SEQ_GEQ(icmp_seq, tp->snd_una) &&
			    SEQ_LT(icmp_seq, tp->snd_max))
				(*notify)(inp, inetctlerrmap[cmd]);
		}
		splx(s);
	} else
		in_pcbnotifyall(&tcb, faddr, inetctlerrmap[cmd], notify);
}

#ifdef INET6
void
tcp6_ctlinput(cmd, sa, d)
	int cmd;
	struct sockaddr *sa;
	void *d;
{
	struct tcphdr th;
	void (*notify) __P((struct inpcb *, int)) = tcp_notify;
	struct ip6_hdr *ip6;
	struct mbuf *m;
	struct ip6ctlparam *ip6cp = NULL;
	const struct sockaddr_in6 *sa6_src = NULL;
	int off;
	struct tcp_portonly {
		u_int16_t th_sport;
		u_int16_t th_dport;
	} *thp;

	if (sa->sa_family != AF_INET6 ||
	    sa->sa_len != sizeof(struct sockaddr_in6))
		return;

	if (cmd == PRC_QUENCH)
		notify = tcp_quench;
	else if (cmd == PRC_MSGSIZE)
		notify = tcp_mtudisc;
	else if (!PRC_IS_REDIRECT(cmd) &&
		 ((unsigned)cmd > PRC_NCMDS || inet6ctlerrmap[cmd] == 0))
		return;

	/* if the parameter is from icmp6, decode it. */
	if (d != NULL) {
		ip6cp = (struct ip6ctlparam *)d;
		m = ip6cp->ip6c_m;
		ip6 = ip6cp->ip6c_ip6;
		off = ip6cp->ip6c_off;
		sa6_src = ip6cp->ip6c_src;
	} else {
		m = NULL;
		ip6 = NULL;
		off = 0;	/* fool gcc */
		sa6_src = &sa6_any;
	}

	if (ip6) {
		/*
		 * XXX: We assume that when IPV6 is non NULL,
		 * M and OFF are valid.
		 */

		/* check if we can safely examine src and dst ports */
		if (m->m_pkthdr.len < off + sizeof(*thp))
			return;

		bzero(&th, sizeof(th));
		m_copydata(m, off, sizeof(*thp), (caddr_t)&th);

		in6_pcbnotify(&tcb, sa, th.th_dport,
		    (struct sockaddr *)ip6cp->ip6c_src,
		    th.th_sport, cmd, NULL, notify);
	} else
		in6_pcbnotify(&tcb, sa, 0, (struct sockaddr *)sa6_src,
			      0, cmd, NULL, notify);
}
#endif /* INET6 */

/*
 * Following is where TCP initial sequence number generation occurs.
 *
 * There are two places where we must use initial sequence numbers:
 * 1.  In SYN-ACK packets.
 * 2.  In SYN packets.
 *
 * The ISNs in SYN-ACK packets have no monotonicity requirement, 
 * and should be as unpredictable as possible to avoid the possibility
 * of spoofing and/or connection hijacking.  To satisfy this
 * requirement, SYN-ACK ISNs are generated via the arc4random()
 * function.  If exact RFC 1948 compliance is requested via sysctl,
 * these ISNs will be generated just like those in SYN packets.
 *
 * The ISNs in SYN packets must be monotonic; TIME_WAIT recycling
 * depends on this property.  In addition, these ISNs should be
 * unguessable so as to prevent connection hijacking.  To satisfy
 * the requirements of this situation, the algorithm outlined in
 * RFC 1948 is used to generate sequence numbers.
 *
 * For more information on the theory of operation, please see
 * RFC 1948.
 *
 * Implementation details:
 *
 * Time is based off the system timer, and is corrected so that it
 * increases by one megabyte per second.  This allows for proper
 * recycling on high speed LANs while still leaving over an hour
 * before rollover.
 *
 * Two sysctls control the generation of ISNs:
 *
 * net.inet.tcp.isn_reseed_interval controls the number of seconds
 * between seeding of isn_secret.  This is normally set to zero,
 * as reseeding should not be necessary.
 *
 * net.inet.tcp.strict_rfc1948 controls whether RFC 1948 is followed
 * strictly.  When strict compliance is requested, reseeding is
 * disabled and SYN-ACKs will be generated in the same manner as
 * SYNs.  Strict mode is disabled by default.
 *
 */

#define ISN_BYTES_PER_SECOND 1048576

u_char isn_secret[32];
int isn_last_reseed;
MD5_CTX isn_ctx;

tcp_seq
tcp_new_isn(tp)
	struct tcpcb *tp;
{
	u_int32_t md5_buffer[4];
	tcp_seq new_isn;

	/* Use arc4random for SYN-ACKs when not in exact RFC1948 mode. */
	if (((tp->t_state == TCPS_LISTEN) || (tp->t_state == TCPS_TIME_WAIT))
	   && tcp_strict_rfc1948 == 0)
		return arc4random();

	/* Seed if this is the first use, reseed if requested. */
	if ((isn_last_reseed == 0) ||
	    ((tcp_strict_rfc1948 == 0) && (tcp_isn_reseed_interval > 0) &&
	     (((u_int)isn_last_reseed + (u_int)tcp_isn_reseed_interval*hz)
		< (u_int)ticks))) {
		read_random_unlimited(&isn_secret, sizeof(isn_secret));
		isn_last_reseed = ticks;
	}
		
	/* Compute the md5 hash and return the ISN. */
	MD5Init(&isn_ctx);
	MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_fport, sizeof(u_short));
	MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_lport, sizeof(u_short));
#ifdef INET6
	if ((tp->t_inpcb->inp_vflag & INP_IPV6) != 0) {
		MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_faddr,
			  sizeof(struct in6_addr));
		MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->in6p_laddr,
			  sizeof(struct in6_addr));
	} else
#endif
	{
		MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_faddr,
			  sizeof(struct in_addr));
		MD5Update(&isn_ctx, (u_char *) &tp->t_inpcb->inp_laddr,
			  sizeof(struct in_addr));
	}
	MD5Update(&isn_ctx, (u_char *) &isn_secret, sizeof(isn_secret));
	MD5Final((u_char *) &md5_buffer, &isn_ctx);
	new_isn = (tcp_seq) md5_buffer[0];
	new_isn += ticks * (ISN_BYTES_PER_SECOND / hz);
	return new_isn;
}

/*
 * 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;
}

/*
 * When a specific ICMP unreachable message is received and the
 * connection state is SYN-SENT, drop the connection.  This behavior
 * is controlled by the icmp_may_rst sysctl.
 */
void
tcp_drop_syn_sent(inp, _errno)
	struct inpcb *inp;
	int _errno;
{
	struct tcpcb *tp = intotcpcb(inp);

	if (tp && tp->t_state == TCPS_SYN_SENT)
		tcp_drop(tp, _errno);
}

/*
 * 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;
#ifdef INET6
	int isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV6) != 0;
#endif /* INET6 */

	if (tp) {
#ifdef INET6
		if (isipv6)
			rt = tcp_rtlookup6(inp);
		else
#endif /* INET6 */
		rt = tcp_rtlookup(inp);
		if (!rt || !rt->rt_rmx.rmx_mtu) {
			tp->t_maxopd = tp->t_maxseg =
#ifdef INET6
				isipv6 ? tcp_v6mssdflt :
#endif /* INET6 */
				tcp_mssdflt;
			return;
		}
		taop = rmx_taop(rt->rt_rmx);
		offered = taop->tao_mssopt;
		mss = rt->rt_rmx.rmx_mtu -
#ifdef INET6
			(isipv6 ?
			 sizeof(struct ip6_hdr) + sizeof(struct tcphdr) :
#endif /* INET6 */
			 sizeof(struct tcpiphdr)
#ifdef INET6
			 )
#endif /* INET6 */
			;

		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_rtttime = 0;
		tp->snd_nxt = tp->snd_una;
		tcp_output(tp);
	}
}

/*
 * 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(struct sockaddr_in);
			((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
				inp->inp_faddr;
			rtalloc(ro);
			rt = ro->ro_rt;
		}
	}
	return rt;
}

#ifdef INET6
struct rtentry *
tcp_rtlookup6(inp)
	struct inpcb *inp;
{
#ifdef NEW_STRUCT_ROUTE
	struct route *ro6;
#else
	struct route_in6 *ro6;
#endif
	struct rtentry *rt;

	ro6 = &inp->in6p_route;
	rt = ro6->ro_rt;
	if (rt == NULL || !(rt->rt_flags & RTF_UP)) {
		/* No route yet, so try to acquire one */
		if (!IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) {
			struct sockaddr_in6 *dst6;

			dst6 = (struct sockaddr_in6 *)&ro6->ro_dst;
			dst6->sin6_family = AF_INET6;
			dst6->sin6_len = sizeof(*dst6);
			dst6->sin6_addr = inp->in6p_faddr;
			rtalloc((struct route *)ro6);
			rt = ro6->ro_rt;
		}
	}
	return rt;
}
#endif /* INET6 */

#ifdef IPSEC
/* compute ESP/AH header size for TCP, including outer IP header. */
size_t
ipsec_hdrsiz_tcp(tp)
	struct tcpcb *tp;
{
	struct inpcb *inp;
	struct mbuf *m;
	size_t hdrsiz;
	struct ip *ip;
#ifdef INET6
	struct ip6_hdr *ip6;
#endif /* INET6 */
	struct tcphdr *th;

	if ((tp == NULL) || ((inp = tp->t_inpcb) == NULL))
		return 0;
	MGETHDR(m, M_DONTWAIT, MT_DATA);
	if (!m)
		return 0;

#ifdef INET6
	if ((inp->inp_vflag & INP_IPV6) != 0) {
		ip6 = mtod(m, struct ip6_hdr *);
		th = (struct tcphdr *)(ip6 + 1);
		m->m_pkthdr.len = m->m_len =
			sizeof(struct ip6_hdr) + sizeof(struct tcphdr);
		tcp_fillheaders(tp, ip6, th);
		hdrsiz = ipsec6_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp);
	} else
#endif /* INET6 */
      {
	ip = mtod(m, struct ip *);
	th = (struct tcphdr *)(ip + 1);
	m->m_pkthdr.len = m->m_len = sizeof(struct tcpiphdr);
	tcp_fillheaders(tp, ip, th);
	hdrsiz = ipsec4_hdrsiz(m, IPSEC_DIR_OUTBOUND, inp);
      }

	m_free(m);
	return hdrsiz;
}
#endif /*IPSEC*/

/*
 * 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;

#ifdef INET6
	if ((inp->inp_vflag & INP_IPV6) != 0)
		rt = tcp_rtlookup6(inp);
	else
#endif /* INET6 */
	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()
{
}