X-Git-Url: https://git.subgeniuskitty.com/unix-history/.git/blobdiff_plain/8ae6c0891d11b5aa1f590d9185be474a72d0d9ad..fcfe450e4ae7bcb3ff4d45b10c8a7c30aa1b3c16:/usr/src/sys/netinet/tcp_output.c

diff --git a/usr/src/sys/netinet/tcp_output.c b/usr/src/sys/netinet/tcp_output.c
index f75fda289c..b166b98451 100644
--- a/usr/src/sys/netinet/tcp_output.c
+++ b/usr/src/sys/netinet/tcp_output.c
@@ -1,4 +1,4 @@
-/*	tcp_output.c	4.35	82/03/20	*/
+/*	tcp_output.c	4.46	82/10/09	*/
 
 #include "../h/param.h"
 #include "../h/systm.h"
@@ -6,33 +6,28 @@
 #include "../h/protosw.h"
 #include "../h/socket.h"
 #include "../h/socketvar.h"
-#include "../net/in.h"
-#include "../net/in_pcb.h"
-#include "../net/in_systm.h"
-#include "../net/ip.h"
-#include "../net/ip_var.h"
-#include "../net/tcp.h"
+#include "../netinet/in.h"
+#include "../net/route.h"
+#include "../netinet/in_pcb.h"
+#include "../netinet/in_systm.h"
+#include "../netinet/ip.h"
+#include "../netinet/ip_var.h"
+#include "../netinet/tcp.h"
 #define	TCPOUTFLAGS
-#include "../net/tcp_fsm.h"
-#include "../net/tcp_seq.h"
-#include "../net/tcp_timer.h"
-#include "../net/tcp_var.h"
-#include "../net/tcpip.h"
-#include "../net/tcp_debug.h"
-#include "../errno.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 <errno.h>
 
 char *tcpstates[]; /* XXX */
 
 /*
- * Initial options: indicate max segment length 1/2 of space
- * allocated for receive; if TCPTRUEOOB is defined, indicate
- * willingness to do true out-of-band.
+ * Initial options.
  */
-#ifndef TCPTRUEOOB
 u_char	tcp_initopt[4] = { TCPOPT_MAXSEG, 4, 0x0, 0x0, };
-#else
-u_char	tcp_initopt[6] = { TCPOPT_MAXSEG, 4, 0x0, 0x0, TCPOPT_WILLOOB, 2 };
-#endif
 
 /*
  * Tcp output routine: figure out what should be sent and send it.
@@ -43,13 +38,13 @@ tcp_output(tp)
 	register struct socket *so = tp->t_inpcb->inp_socket;
 	register int len;
 	struct mbuf *m0;
-	int off, flags, win;
+	int off, flags, win, error;
 	register struct mbuf *m;
 	register struct tcpiphdr *ti;
 	u_char *opt;
 	unsigned optlen = 0;
+	int sendalot;
 
-COUNT(TCP_OUTPUT);
 
 	/*
 	 * Determine length of data that should be transmitted,
@@ -57,12 +52,16 @@ COUNT(TCP_OUTPUT);
 	 * If there is some data or critical controls (SYN, RST)
 	 * to send, then transmit; otherwise, investigate further.
 	 */
+again:
+	sendalot = 0;
 	off = tp->snd_nxt - tp->snd_una;
 	len = MIN(so->so_snd.sb_cc, tp->snd_wnd+tp->t_force) - off;
 	if (len < 0)
-		return (0);			/* past FIN */
-	if (len > tp->t_maxseg)
+		return (0);	/* ??? */	/* past FIN */
+	if (len > tp->t_maxseg) {
 		len = tp->t_maxseg;
+		sendalot = 1;
+	}
 
 	flags = tcp_outflags[tp->t_state];
 	if (tp->snd_nxt + len < tp->snd_una + so->so_snd.sb_cc)
@@ -92,13 +91,6 @@ COUNT(TCP_OUTPUT);
 	if (tp->t_flags&TF_ACKNOW)
 		goto send;
 
-#ifdef TCPTRUEOOB
-	/*
-	 * Send if an out of band data or ack should be transmitted.
-	 */
-	if (tp->t_oobflags&(TCPOOB_OWEACK|TCPOOB_NEEDACK)))
-		goto send;
-#endif
 
 	/*
 	 * Calculate available window in i, and also amount
@@ -111,6 +103,33 @@ COUNT(TCP_OUTPUT);
 	    ((100*(win-(tp->rcv_adv-tp->rcv_nxt))/so->so_rcv.sb_hiwat) >= 35))
 		goto send;
 
+	/*
+	 * TCP window updates are not reliable, rather a polling protocol
+	 * using ``persist'' packets is used to insure receipt of window
+	 * updates.  The three ``states'' for the output side are:
+	 *	idle			not doing retransmits or persists
+	 *	persisting		to move a zero window
+	 *	(re)transmitting	and thereby not persisting
+	 *
+	 * tp->t_timer[TCPT_PERSIST]
+	 *	is set when we are in persist state.
+	 * tp->t_force
+	 *	is set when we are called to send a persist packet.
+	 * tp->t_timer[TCPT_REXMT]
+	 *	is set when we are retransmitting
+	 * The output side is idle when both timers are zero.
+	 *
+	 * If send window is closed, there is data to transmit, and no
+	 * retransmit or persist is pending, then go to persist state,
+	 * arranging to force out a byte to get more current window information
+	 * if nothing happens soon.
+	 */
+	if (tp->snd_wnd == 0 && so->so_snd.sb_cc &&
+	    tp->t_timer[TCPT_REXMT] == 0 && tp->t_timer[TCPT_PERSIST] == 0) {
+		tp->t_rxtshift = 0;
+		tcp_setpersist(tp);
+	}
+
 	/*
 	 * No reason to send a segment, just return.
 	 */
@@ -124,7 +143,7 @@ send:
 	 */
 	MGET(m, 0);
 	if (m == 0)
-		return (0);
+		return (ENOBUFS);
 	m->m_off = MMAXOFF - sizeof (struct tcpiphdr);
 	m->m_len = sizeof (struct tcpiphdr);
 	if (len) {
@@ -156,7 +175,7 @@ send:
 			goto noopt;
 		opt = tcp_initopt;
 		optlen = sizeof (tcp_initopt);
-		*(u_short *)(opt + 2) = so->so_rcv.sb_hiwat / 2;
+		*(u_short *)(opt + 2) = MIN(so->so_rcv.sb_hiwat / 2, 1024);
 #if vax
 		*(u_short *)(opt + 2) = htons(*(u_short *)(opt + 2));
 #endif
@@ -166,50 +185,19 @@ send:
 		opt = mtod(tp->t_tcpopt, u_char *);
 		optlen = tp->t_tcpopt->m_len;
 	}
-#ifndef TCPTRUEOOB
-	if (opt)
-#else
-	if (opt || (tp->t_oobflags&(TCPOOB_OWEACK|TCPOOB_NEEDACK)))
-#endif
-	{
+	if (opt) {
 		m0 = m->m_next;
 		m->m_next = m_get(M_DONTWAIT);
 		if (m->m_next == 0) {
 			(void) m_free(m);
 			m_freem(m0);
-			return (0);
+			return (ENOBUFS);
 		}
 		m->m_next->m_next = m0;
 		m0 = m->m_next;
-		m0->m_off = MMINOFF;
 		m0->m_len = optlen;
 		bcopy((caddr_t)opt, mtod(m0, caddr_t), optlen);
 		opt = (u_char *)(mtod(m0, caddr_t) + optlen);
-#ifdef TCPTRUEOOB
-		if (tp->t_oobflags&TCPOOB_OWEACK) {
-printf("tp %x send OOBACK for %x\n", tp->t_iobseq);
-			*opt++ = TCPOPT_OOBACK;
-			*opt++ = 3;
-			*opt++ = tp->t_iobseq;
-			m0->m_len += 3;
-			tp->t_oobflags &= ~TCPOOB_OWEACK;
-			/* sender should rexmt oob to force ack repeat */
-		}
-		if (tp->t_oobflags&TCPOOB_NEEDACK) {
-printf("tp %x send OOBDATA seq %x data %x\n", tp->t_oobseq, tp->t_oobc);
-			*opt++ = TCPOPT_OOBDATA;
-			*opt++ = 8;
-			*opt++ = tp->t_oobseq;
-			*opt++ = tp->t_oobc;
-			*(tcp_seq *)opt = tp->t_oobmark - tp->snd_nxt;
-#ifdef vax
-			*(tcp_seq *)opt = htonl((unsigned)*(tcp_seq *)opt);
-#endif
-			m0->m_len += 8;
-			TCPT_RANGESET(tp->t_timer[TCPT_OOBREXMT],
-			    tcp_beta * tp->t_srtt, TCPTV_MIN, TCPTV_MAX);
-		}
-#endif
 		while (m0->m_len & 0x3) {
 			*opt++ = TCPOPT_EOL;
 			m0->m_len++;
@@ -242,7 +230,13 @@ noopt:
 		 * number wraparound.
 		 */
 		tp->snd_up = tp->snd_una;		/* drag it along */
-	/* PUSH */
+	/*
+	 * If anything to send and we can send it all, set PUSH.
+	 * (This will keep happy those implementations which only
+	 * give data to the user when a buffer fills or a PUSH comes in.
+	 */
+	if (len && off+len == so->so_snd.sb_cc)
+		ti->ti_flags |= TH_PUSH;
 
 	/*
 	 * Put TCP length in extended header, and then
@@ -257,49 +251,52 @@ noopt:
 	ti->ti_sum = in_cksum(m, sizeof (struct tcpiphdr) + (int)optlen + len);
 
 	/*
-	 * Advance snd_nxt over sequence space of this segment
-	 */
-	if (flags & (TH_SYN|TH_FIN))
-		tp->snd_nxt++;
-	tp->snd_nxt += len;
-
-	/*
-	 * If this transmission closes the window,
-	 * start persistance timer at 2 round trip times
-	 * but at least TCPTV_PERSMIN ticks.
+	 * In transmit state, time the transmission and arrange for
+	 * the retransmit.  In persist state, reset persist time for
+	 * next persist.
 	 */
-	if (TCPS_HAVERCVDSYN(tp->t_state) &&
-	    SEQ_GEQ(tp->snd_nxt, tp->snd_una+tp->snd_wnd) &&
-	    tp->t_timer[TCPT_PERSIST] == 0)
-		TCPT_RANGESET(tp->t_timer[TCPT_PERSIST],
-		    2 * tp->t_srtt, TCPTV_PERSMIN, TCPTV_MAX);
+	if (tp->t_force == 0) {
+		/*
+		 * Advance snd_nxt over sequence space of this segment.
+		 */
+		if (flags & (TH_SYN|TH_FIN))
+			tp->snd_nxt++;
+		tp->snd_nxt += len;
+		if (SEQ_GT(tp->snd_nxt, tp->snd_max))
+			tp->snd_max = tp->snd_nxt;
 
-	/*
-	 * Time this transmission if not a retransmission and
-	 * not currently timing anything.
-	 */
-	if (SEQ_GT(tp->snd_nxt, tp->snd_max) && tp->t_rtt == 0) {
-		tp->t_rtt = 1;
-		tp->t_rtseq = tp->snd_nxt - len;
-	}
+		/*
+		 * Time this transmission if not a retransmission and
+		 * not currently timing anything.
+		 */
+		if (SEQ_GT(tp->snd_nxt, tp->snd_max) && tp->t_rtt == 0) {
+			tp->t_rtt = 1;
+			tp->t_rtseq = tp->snd_nxt - len;
+		}
 
-	/*
-	 * Set retransmit timer if not currently set.
-	 * Initial value for retransmit timer to tcp_beta*tp->t_srtt.
-	 * Initialize shift counter which is used for exponential
-	 * backoff of retransmit time.
-	 */
-	if (tp->t_timer[TCPT_REXMT] == 0 && tp->snd_nxt != tp->snd_una) {
-		TCPT_RANGESET(tp->t_timer[TCPT_REXMT],
-		    tcp_beta * tp->t_srtt, TCPTV_MIN, TCPTV_MAX);
-		tp->t_rtt = 0;
-		tp->t_rxtshift = 0;
+		/*
+		 * Set retransmit timer if not currently set.
+		 * Initial value for retransmit timer to tcp_beta*tp->t_srtt.
+		 * Initialize shift counter which is used for exponential
+		 * backoff of retransmit time.
+		 */
+		if (tp->t_timer[TCPT_REXMT] == 0 &&
+		    tp->snd_nxt != tp->snd_una) {
+			TCPT_RANGESET(tp->t_timer[TCPT_REXMT],
+			    tcp_beta * tp->t_srtt, TCPTV_MIN, TCPTV_MAX);
+			tp->t_rtt = 0;
+			tp->t_rxtshift = 0;
+		}
+		tp->t_timer[TCPT_PERSIST] = 0;
+	} else {
+		if (SEQ_GT(tp->snd_una+1, tp->snd_max))
+			tp->snd_max = tp->snd_una+1;
 	}
 
 	/*
 	 * Trace.
 	 */
-	if (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)
+	if (so->so_options & SO_DEBUG)
 		tcp_trace(TA_OUTPUT, tp->t_state, tp, ti, 0);
 
 	/*
@@ -308,8 +305,9 @@ noopt:
 	 */
 	((struct ip *)ti)->ip_len = sizeof (struct tcpiphdr) + optlen + len;
 	((struct ip *)ti)->ip_ttl = TCP_TTL;
-	if (ip_output(m, tp->t_ipopt, 0) == 0)
-		return (0);
+	if (error = ip_output(m, tp->t_ipopt, (so->so_options & SO_DONTROUTE) ?
+	    &routetoif : &tp->t_inpcb->inp_route, 0))
+		return (error);
 
 	/*
 	 * Data sent (as far as we can tell).
@@ -320,7 +318,24 @@ noopt:
 	if (win > 0 && SEQ_GT(tp->rcv_nxt+win, tp->rcv_adv))
 		tp->rcv_adv = tp->rcv_nxt + win;
 	tp->t_flags &= ~(TF_ACKNOW|TF_DELACK);
-	if (SEQ_GT(tp->snd_nxt, tp->snd_max))
-		tp->snd_max = tp->snd_nxt;
-	return (1);
+	if (sendalot && tp->t_force == 0)
+		goto again;
+	return (0);
+}
+
+tcp_setpersist(tp)
+	register struct tcpcb *tp;
+{
+
+	if (tp->t_timer[TCPT_REXMT])
+		panic("tcp_output REXMT");
+	/*
+	 * Start/restart persistance timer.
+	 */
+	TCPT_RANGESET(tp->t_timer[TCPT_PERSIST],
+	    ((int)(tcp_beta * tp->t_srtt)) << tp->t_rxtshift,
+	    TCPTV_PERSMIN, TCPTV_MAX);
+	tp->t_rxtshift++;
+	if (tp->t_rxtshift >= TCP_MAXRXTSHIFT)
+		tp->t_rxtshift = 0;
 }