X-Git-Url: https://git.subgeniuskitty.com/unix-history/.git/blobdiff_plain/ee954ff170ffcebec3b59cf47a38de446bdaa376..224f3a720c8acd42674bb9f2a4864989c7ae75f0:/usr/src/sys/netinet/tcp_input.c

diff --git a/usr/src/sys/netinet/tcp_input.c b/usr/src/sys/netinet/tcp_input.c
index e13bcd3233..2e010262e3 100644
--- a/usr/src/sys/netinet/tcp_input.c
+++ b/usr/src/sys/netinet/tcp_input.c
@@ -1,28 +1,34 @@
-/*	tcp_input.c	1.92	83/04/03	*/
+/*
+ * Copyright (c) 1982 Regents of the University of California.
+ * All rights reserved.  The Berkeley software License Agreement
+ * specifies the terms and conditions for redistribution.
+ *
+ *	@(#)tcp_input.c	6.15 (Berkeley) %G%
+ */
 
-#include "../h/param.h"
-#include "../h/systm.h"
-#include "../h/mbuf.h"
-#include "../h/protosw.h"
-#include "../h/socket.h"
-#include "../h/socketvar.h"
-#include "../h/errno.h"
+#include "param.h"
+#include "systm.h"
+#include "mbuf.h"
+#include "protosw.h"
+#include "socket.h"
+#include "socketvar.h"
+#include "errno.h"
 
 #include "../net/if.h"
 #include "../net/route.h"
 
-#include "../netinet/in.h"
-#include "../netinet/in_pcb.h"
-#include "../netinet/in_systm.h"
-#include "../netinet/ip.h"
-#include "../netinet/ip_var.h"
-#include "../netinet/tcp.h"
-#include "../netinet/tcp_fsm.h"
-#include "../netinet/tcp_seq.h"
-#include "../netinet/tcp_timer.h"
-#include "../netinet/tcp_var.h"
-#include "../netinet/tcpip.h"
-#include "../netinet/tcp_debug.h"
+#include "in.h"
+#include "in_pcb.h"
+#include "in_systm.h"
+#include "ip.h"
+#include "ip_var.h"
+#include "tcp.h"
+#include "tcp_fsm.h"
+#include "tcp_seq.h"
+#include "tcp_timer.h"
+#include "tcp_var.h"
+#include "tcpip.h"
+#include "tcp_debug.h"
 
 int	tcpprintfs = 0;
 int	tcpcksum = 1;
@@ -30,6 +36,126 @@ struct	tcpiphdr tcp_saveti;
 extern	tcpnodelack;
 
 struct	tcpcb *tcp_newtcpcb();
+
+/*
+ * Insert segment ti into reassembly queue of tcp with
+ * control block tp.  Return TH_FIN if reassembly now includes
+ * a segment with FIN.  The macro form does the common case inline
+ * (segment is the next to be received on an established connection,
+ * and the queue is empty), avoiding linkage into and removal
+ * from the queue and repetition of various conversions.
+ */
+#define	TCP_REASS(tp, ti, m, so, flags) { \
+	if ((ti)->ti_seq == (tp)->rcv_nxt && \
+	    (tp)->seg_next == (struct tcpiphdr *)(tp) && \
+	    (tp)->t_state == TCPS_ESTABLISHED) { \
+		(tp)->rcv_nxt += (ti)->ti_len; \
+		flags = (ti)->ti_flags & TH_FIN; \
+		sbappend(&(so)->so_rcv, (m)); \
+		sorwakeup(so); \
+	} else \
+		(flags) = tcp_reass((tp), (ti)); \
+}
+
+tcp_reass(tp, ti)
+	register struct tcpcb *tp;
+	register struct tcpiphdr *ti;
+{
+	register struct tcpiphdr *q;
+	struct socket *so = tp->t_inpcb->inp_socket;
+	struct mbuf *m;
+	int flags;
+
+	/*
+	 * Call with ti==0 after become established to
+	 * force pre-ESTABLISHED data up to user socket.
+	 */
+	if (ti == 0)
+		goto present;
+
+	/*
+	 * Find a segment which begins after this one does.
+	 */
+	for (q = tp->seg_next; q != (struct tcpiphdr *)tp;
+	    q = (struct tcpiphdr *)q->ti_next)
+		if (SEQ_GT(q->ti_seq, ti->ti_seq))
+			break;
+
+	/*
+	 * If there is a preceding segment, it may provide some of
+	 * our data already.  If so, drop the data from the incoming
+	 * segment.  If it provides all of our data, drop us.
+	 */
+	if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) {
+		register int i;
+		q = (struct tcpiphdr *)q->ti_prev;
+		/* conversion to int (in i) handles seq wraparound */
+		i = q->ti_seq + q->ti_len - ti->ti_seq;
+		if (i > 0) {
+			if (i >= ti->ti_len)
+				goto drop;
+			m_adj(dtom(ti), i);
+			ti->ti_len -= i;
+			ti->ti_seq += i;
+		}
+		q = (struct tcpiphdr *)(q->ti_next);
+	}
+
+	/*
+	 * While we overlap succeeding segments trim them or,
+	 * if they are completely covered, dequeue them.
+	 */
+	while (q != (struct tcpiphdr *)tp) {
+		register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
+		if (i <= 0)
+			break;
+		if (i < q->ti_len) {
+			q->ti_seq += i;
+			q->ti_len -= i;
+			m_adj(dtom(q), i);
+			break;
+		}
+		q = (struct tcpiphdr *)q->ti_next;
+		m = dtom(q->ti_prev);
+		remque(q->ti_prev);
+		m_freem(m);
+	}
+
+	/*
+	 * Stick new segment in its place.
+	 */
+	insque(ti, q->ti_prev);
+
+present:
+	/*
+	 * Present data to user, advancing rcv_nxt through
+	 * completed sequence space.
+	 */
+	if (TCPS_HAVERCVDSYN(tp->t_state) == 0)
+		return (0);
+	ti = tp->seg_next;
+	if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt)
+		return (0);
+	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
+		return (0);
+	do {
+		tp->rcv_nxt += ti->ti_len;
+		flags = ti->ti_flags & TH_FIN;
+		remque(ti);
+		m = dtom(ti);
+		ti = (struct tcpiphdr *)ti->ti_next;
+		if (so->so_state & SS_CANTRCVMORE)
+			m_freem(m);
+		else
+			sbappend(&so->so_rcv, m);
+	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
+	sorwakeup(so);
+	return (flags);
+drop:
+	m_freem(dtom(ti));
+	return (0);
+}
+
 /*
  * TCP input routine, follows pages 65-76 of the
  * protocol specification dated September, 1981 very closely.
@@ -98,11 +224,13 @@ tcp_input(m0)
 	tlen -= off;
 	ti->ti_len = tlen;
 	if (off > sizeof (struct tcphdr)) {
-		if ((m = m_pullup(m, sizeof (struct ip) + off)) == 0) {
-			tcpstat.tcps_hdrops++;
-			return;
+		if (m->m_len < sizeof(struct ip) + off) {
+			if ((m = m_pullup(m, sizeof (struct ip) + off)) == 0) {
+				tcpstat.tcps_hdrops++;
+				return;
+			}
+			ti = mtod(m, struct tcpiphdr *);
 		}
-		ti = mtod(m, struct tcpiphdr *);
 		om = m_get(M_DONTWAIT, MT_DATA);
 		if (om == 0)
 			goto drop;
@@ -171,6 +299,7 @@ tcp_input(m0)
 		inp = (struct inpcb *)so->so_pcb;
 		inp->inp_laddr = ti->ti_dst;
 		inp->inp_lport = ti->ti_dport;
+		inp->inp_options = ip_srcroute();
 		tp = intotcpcb(inp);
 		tp->t_state = TCPS_LISTEN;
 	}
@@ -183,20 +312,24 @@ tcp_input(m0)
 	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
 
 	/*
-	 * Process options.
+	 * Process options if not in LISTEN state,
+	 * else do it below (after getting remote address).
 	 */
-	if (om) {
-		tcp_dooptions(tp, om);
+	if (om && tp->t_state != TCPS_LISTEN) {
+		tcp_dooptions(tp, om, ti);
 		om = 0;
 	}
 
 	/*
 	 * Calculate amount of space in receive window,
 	 * and then do TCP input processing.
+	 * Receive window is amount of space in rcv queue,
+	 * but not less than advertised window.
 	 */
 	tp->rcv_wnd = sbspace(&so->so_rcv);
 	if (tp->rcv_wnd < 0)
 		tp->rcv_wnd = 0;
+	tp->rcv_wnd = MAX(tp->rcv_wnd, (short)(tp->rcv_adv - tp->rcv_nxt));
 
 	switch (tp->t_state) {
 
@@ -204,6 +337,7 @@ tcp_input(m0)
 	 * If the state is LISTEN then ignore segment if it contains an RST.
 	 * If the segment contains an ACK then it is bad and send a RST.
 	 * If it does not contain a SYN then it is not interesting; drop it.
+	 * Don't bother responding if the destination was a broadcast.
 	 * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
 	 * tp->iss, and send a segment:
 	 *     <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
@@ -222,6 +356,8 @@ tcp_input(m0)
 			goto dropwithreset;
 		if ((tiflags & TH_SYN) == 0)
 			goto drop;
+		if (in_broadcast(ti->ti_dst))
+			goto drop;
 		am = m_get(M_DONTWAIT, MT_SONAME);
 		if (am == NULL)
 			goto drop;
@@ -242,10 +378,14 @@ tcp_input(m0)
 		tp->t_template = tcp_template(tp);
 		if (tp->t_template == 0) {
 			in_pcbdisconnect(inp);
-			inp->inp_laddr = laddr;
+			dropsocket = 0;		/* socket is already gone */
 			tp = 0;
 			goto drop;
 		}
+		if (om) {
+			tcp_dooptions(tp, om, ti);
+			om = 0;
+		}
 		tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
 		tp->irs = ti->ti_seq;
 		tcp_sendseqinit(tp);
@@ -270,8 +410,7 @@ tcp_input(m0)
 	 */
 	case TCPS_SYN_SENT:
 		if ((tiflags & TH_ACK) &&
-/* this should be SEQ_LT; is SEQ_LEQ for BBN vax TCP only */
-		    (SEQ_LT(ti->ti_ack, tp->iss) ||
+		    (SEQ_LEQ(ti->ti_ack, tp->iss) ||
 		     SEQ_GT(ti->ti_ack, tp->snd_max)))
 			goto dropwithreset;
 		if (tiflags & TH_RST) {
@@ -291,6 +430,7 @@ tcp_input(m0)
 		if (SEQ_GT(tp->snd_una, tp->iss)) {
 			soisconnected(so);
 			tp->t_state = TCPS_ESTABLISHED;
+			tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp));
 			(void) tcp_reass(tp, (struct tcpiphdr *)0);
 		} else
 			tp->t_state = TCPS_SYN_RECEIVED;
@@ -313,6 +453,16 @@ trimthenstep6:
 		goto step6;
 	}
 
+	/*
+	 * If data is received on a connection after the
+	 * user processes are gone, then RST the other end.
+	 */
+	if ((so->so_state & SS_NOFDREF) && tp->t_state > TCPS_CLOSE_WAIT &&
+	    ti->ti_len) {
+		tp = tcp_close(tp);
+		goto dropwithreset;
+	}
+
 	/*
 	 * States other than LISTEN or SYN_SENT.
 	 * First check that at least some bytes of segment are within 
@@ -376,16 +526,6 @@ trimthenstep6:
 		}
 	}
 
-	/*
-	 * If data is received on a connection after the
-	 * user processes are gone, then RST the other end.
-	 */
-	if ((so->so_state & SS_NOFDREF) && tp->t_state > TCPS_CLOSE_WAIT &&
-	    ti->ti_len) {
-		tp = tcp_close(tp);
-		goto dropwithreset;
-	}
-
 	/*
 	 * If the RST bit is set examine the state:
 	 *    SYN_RECEIVED STATE:
@@ -451,6 +591,7 @@ trimthenstep6:
 		tp->t_timer[TCPT_REXMT] = 0;
 		soisconnected(so);
 		tp->t_state = TCPS_ESTABLISHED;
+		tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp));
 		(void) tcp_reass(tp, (struct tcpiphdr *)0);
 		tp->snd_wl1 = ti->ti_seq - 1;
 		/* fall into ... */
@@ -497,12 +638,16 @@ trimthenstep6:
 		else {
 			TCPT_RANGESET(tp->t_timer[TCPT_REXMT],
 			    tcp_beta * tp->t_srtt, TCPTV_MIN, TCPTV_MAX);
-			tp->t_rtt = 1;
 			tp->t_rxtshift = 0;
 		}
+		/*
+		 * When new data is acked, open the congestion window a bit.
+		 */
+		if (acked > 0)
+			tp->snd_cwnd = MIN(11 * tp->snd_cwnd / 10, 65535);
 		if (acked > so->so_snd.sb_cc) {
-			sbdrop(&so->so_snd, so->so_snd.sb_cc);
 			tp->snd_wnd -= so->so_snd.sb_cc;
+			sbdrop(&so->so_snd, so->so_snd.sb_cc);
 		} else {
 			sbdrop(&so->so_snd, acked);
 			tp->snd_wnd -= acked;
@@ -526,9 +671,14 @@ trimthenstep6:
 				/*
 				 * If we can't receive any more
 				 * data, then closing user can proceed.
+				 * Starting the timer is contrary to the
+				 * specification, but if we don't get a FIN
+				 * we'll hang forever.
 				 */
-				if (so->so_state & SS_CANTRCVMORE)
+				if (so->so_state & SS_CANTRCVMORE) {
 					soisdisconnected(so);
+					tp->t_timer[TCPT_2MSL] = TCPTV_MAXIDLE;
+				}
 				tp->t_state = TCPS_FIN_WAIT_2;
 			}
 			break;
@@ -581,8 +731,6 @@ step6:
 		tp->snd_wnd = ti->ti_win;
 		tp->snd_wl1 = ti->ti_seq;
 		tp->snd_wl2 = ti->ti_ack;
-		if (tp->snd_wnd != 0)
-			tp->t_timer[TCPT_PERSIST] = 0;
 	}
 
 	/*
@@ -590,6 +738,18 @@ step6:
 	 */
 	if ((tiflags & TH_URG) && ti->ti_urp &&
 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
+		/*
+		 * This is a kludge, but if we receive accept
+		 * random urgent pointers, we'll crash in
+		 * soreceive.  It's hard to imagine someone
+		 * actually wanting to send this much urgent data.
+		 */
+		if (ti->ti_urp + (unsigned) so->so_rcv.sb_cc > 32767) {
+			ti->ti_urp = 0;			/* XXX */
+			tiflags &= ~TH_URG;		/* XXX */
+			ti->ti_flags &= ~TH_URG;	/* XXX */
+			goto badurp;			/* XXX */
+		}
 		/*
 		 * If this segment advances the known urgent pointer,
 		 * then mark the data stream.  This should not happen
@@ -604,7 +764,7 @@ step6:
 			if (so->so_oobmark == 0)
 				so->so_state |= SS_RCVATMARK;
 			sohasoutofband(so);
-			tp->t_oobflags &= ~TCPOOB_HAVEDATA;
+			tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA);
 		}
 		/*
 		 * Remove out of band data so doesn't get presented to user.
@@ -615,6 +775,7 @@ step6:
 		if (ti->ti_urp <= ti->ti_len)
 			tcp_pulloutofband(so, ti);
 	}
+badurp:							/* XXX */
 
 	/*
 	 * Process the segment text, merging it into the TCP sequencing queue,
@@ -626,7 +787,7 @@ step6:
 	 */
 	if ((ti->ti_len || (tiflags&TH_FIN)) &&
 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
-		tiflags = tcp_reass(tp, ti);
+		TCP_REASS(tp, ti, m, so, tiflags);
 		if (tcpnodelack == 0)
 			tp->t_flags |= TF_DELACK;
 		else
@@ -715,8 +876,9 @@ dropwithreset:
 	/*
 	 * Generate a RST, dropping incoming segment.
 	 * Make ACK acceptable to originator of segment.
+	 * Don't bother to respond if destination was broadcast.
 	 */
-	if (tiflags & TH_RST)
+	if ((tiflags & TH_RST) || in_broadcast(ti->ti_dst))
 		goto drop;
 	if (tiflags & TH_ACK)
 		tcp_respond(tp, ti, (tcp_seq)0, ti->ti_ack, TH_RST);
@@ -746,9 +908,10 @@ drop:
 	return;
 }
 
-tcp_dooptions(tp, om)
+tcp_dooptions(tp, om, ti)
 	struct tcpcb *tp;
 	struct mbuf *om;
+	struct tcpiphdr *ti;
 {
 	register u_char *cp;
 	int opt, optlen, cnt;
@@ -761,8 +924,11 @@ tcp_dooptions(tp, om)
 			break;
 		if (opt == TCPOPT_NOP)
 			optlen = 1;
-		else
+		else {
 			optlen = cp[1];
+			if (optlen <= 0)
+				break;
+		}
 		switch (opt) {
 
 		default:
@@ -771,8 +937,11 @@ tcp_dooptions(tp, om)
 		case TCPOPT_MAXSEG:
 			if (optlen != 4)
 				continue;
+			if (!(ti->ti_flags & TH_SYN))
+				continue;
 			tp->t_maxseg = *(u_short *)(cp + 2);
 			tp->t_maxseg = ntohs((u_short)tp->t_maxseg);
+			tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp));
 			break;
 		}
 	}
@@ -813,105 +982,50 @@ tcp_pulloutofband(so, ti)
 }
 
 /*
- * Insert segment ti into reassembly queue of tcp with
- * control block tp.  Return TH_FIN if reassembly now includes
- * a segment with FIN.
+ *  Determine a reasonable value for maxseg size.
+ *  If the route is known, use one that can be handled
+ *  on the given interface without forcing IP to fragment.
+ *  If bigger than a page (CLBYTES), round down to nearest pagesize
+ *  to utilize pagesize mbufs.
+ *  If interface pointer is unavailable, or the destination isn't local,
+ *  use a conservative size (512 or the default IP max size, but no more
+ *  than the mtu of the interface through which we route),
+ *  as we can't discover anything about intervening gateways or networks.
+ *
+ *  This is ugly, and doesn't belong at this level, but has to happen somehow.
  */
-tcp_reass(tp, ti)
+tcp_mss(tp)
 	register struct tcpcb *tp;
-	register struct tcpiphdr *ti;
 {
-	register struct tcpiphdr *q;
-	struct socket *so = tp->t_inpcb->inp_socket;
-	struct mbuf *m;
-	int flags;
-
-	/*
-	 * Call with ti==0 after become established to
-	 * force pre-ESTABLISHED data up to user socket.
-	 */
-	if (ti == 0)
-		goto present;
-
-	/*
-	 * Find a segment which begins after this one does.
-	 */
-	for (q = tp->seg_next; q != (struct tcpiphdr *)tp;
-	    q = (struct tcpiphdr *)q->ti_next)
-		if (SEQ_GT(q->ti_seq, ti->ti_seq))
-			break;
-
-	/*
-	 * If there is a preceding segment, it may provide some of
-	 * our data already.  If so, drop the data from the incoming
-	 * segment.  If it provides all of our data, drop us.
-	 */
-	if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) {
-		register int i;
-		q = (struct tcpiphdr *)q->ti_prev;
-		/* conversion to int (in i) handles seq wraparound */
-		i = q->ti_seq + q->ti_len - ti->ti_seq;
-		if (i > 0) {
-			if (i >= ti->ti_len)
-				goto drop;
-			m_adj(dtom(ti), i);
-			ti->ti_len -= i;
-			ti->ti_seq += i;
-		}
-		q = (struct tcpiphdr *)(q->ti_next);
-	}
+	struct route *ro;
+	struct ifnet *ifp;
+	int mss;
+	struct inpcb *inp;
 
-	/*
-	 * While we overlap succeeding segments trim them or,
-	 * if they are completely covered, dequeue them.
-	 */
-	while (q != (struct tcpiphdr *)tp) {
-		register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
-		if (i <= 0)
-			break;
-		if (i < q->ti_len) {
-			q->ti_seq += i;
-			q->ti_len -= i;
-			m_adj(dtom(q), i);
-			break;
+	inp = tp->t_inpcb;
+	ro = &inp->inp_route;
+	if ((ro->ro_rt == (struct rtentry *)0) ||
+	    (ifp = ro->ro_rt->rt_ifp) == (struct ifnet *)0) {
+		/* No route yet, so try to acquire one */
+		if (inp->inp_faddr.s_addr != INADDR_ANY) {
+			ro->ro_dst.sa_family = AF_INET;
+			((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
+				inp->inp_faddr;
+			rtalloc(ro);
 		}
-		q = (struct tcpiphdr *)q->ti_next;
-		m = dtom(q->ti_prev);
-		remque(q->ti_prev);
-		m_freem(m);
+		if ((ro->ro_rt == 0) || (ifp = ro->ro_rt->rt_ifp) == 0)
+			return (TCP_MSS);
 	}
 
-	/*
-	 * Stick new segment in its place.
-	 */
-	insque(ti, q->ti_prev);
-
-present:
-	/*
-	 * Present data to user, advancing rcv_nxt through
-	 * completed sequence space.
-	 */
-	if (TCPS_HAVERCVDSYN(tp->t_state) == 0)
-		return (0);
-	ti = tp->seg_next;
-	if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt)
-		return (0);
-	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
-		return (0);
-	do {
-		tp->rcv_nxt += ti->ti_len;
-		flags = ti->ti_flags & TH_FIN;
-		remque(ti);
-		m = dtom(ti);
-		ti = (struct tcpiphdr *)ti->ti_next;
-		if (so->so_state & SS_CANTRCVMORE)
-			m_freem(m);
-		else
-			sbappend(&so->so_rcv, m);
-	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
-	sorwakeup(so);
-	return (flags);
-drop:
-	m_freem(dtom(ti));
-	return (0);
+	mss = ifp->if_mtu - sizeof(struct tcpiphdr);
+#if	(CLBYTES & (CLBYTES - 1)) == 0
+	if (mss > CLBYTES)
+		mss &= ~(CLBYTES-1);
+#else
+	if (mss > CLBYTES)
+		mss = mss / CLBYTES * CLBYTES;
+#endif
+	if (in_localaddr(inp->inp_faddr))
+		return (mss);
+	return (MIN(mss, TCP_MSS));
 }