-/* tcp_input.c 1.60 82/03/13 */
-
-#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 "../net/in.h"
-#include "../net/in_pcb.h"
-#include "../net/in_systm.h"
+/*
+ * 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 "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/ip.h"
-#include "../net/ip_var.h"
-#include "../net/tcp.h"
-#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"
-
-#ifdef notdef
+#include "../net/route.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;
-#endif
int tcpcksum = 1;
-struct sockaddr_in tcp_in = { AF_INET };
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.
int todrop, acked;
short ostate;
struct in_addr laddr;
+ int dropsocket = 0;
-COUNT(TCP_INPUT);
/*
* Get IP and TCP header together in first mbuf.
* Note: IP leaves IP header in first mbuf.
ti->ti_next = ti->ti_prev = 0;
ti->ti_x1 = 0;
ti->ti_len = (u_short)tlen;
-#if vax
ti->ti_len = htons((u_short)ti->ti_len);
-#endif
if (ti->ti_sum = in_cksum(m, len)) {
+ if (tcpprintfs)
+ printf("tcp sum: src %x\n", ti->ti_src);
tcpstat.tcps_badsum++;
- printf("tcp cksum %x\n", ti->ti_sum);
goto drop;
}
}
*/
off = ti->ti_off << 2;
if (off < sizeof (struct tcphdr) || off > tlen) {
+ if (tcpprintfs)
+ printf("tcp off: src %x off %d\n", ti->ti_src, off);
tcpstat.tcps_badoff++;
goto drop;
}
- ti->ti_len = tlen - off;
+ tlen -= off;
+ ti->ti_len = tlen;
if (off > sizeof (struct tcphdr)) {
- if ((m = m_pullup(m, sizeof (struct ip) + off)) == 0) {
- tcpstat.tcps_hdrops++;
- goto drop;
+ 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);
+ om = m_get(M_DONTWAIT, MT_DATA);
if (om == 0)
goto drop;
- om->m_off = MMINOFF;
om->m_len = off - sizeof (struct tcphdr);
{ caddr_t op = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
bcopy(op, mtod(om, caddr_t), (unsigned)om->m_len);
tiflags = ti->ti_flags;
/*
- * drop IP header
+ * Drop TCP and IP headers.
*/
off += sizeof (struct ip);
m->m_off += off;
m->m_len -= off;
-#if vax
/*
* Convert TCP protocol specific fields to host format.
*/
ti->ti_ack = ntohl(ti->ti_ack);
ti->ti_win = ntohs(ti->ti_win);
ti->ti_urp = ntohs(ti->ti_urp);
-#endif
/*
- * Locate pcb for segment. On match, update the local
- * address stored in the block to reflect anchoring.
+ * Locate pcb for segment.
*/
inp = in_pcblookup
(&tcb, ti->ti_src, ti->ti_sport, ti->ti_dst, ti->ti_dport,
ostate = tp->t_state;
tcp_saveti = *ti;
}
+ if (so->so_options & SO_ACCEPTCONN) {
+ so = sonewconn(so);
+ if (so == 0)
+ goto drop;
+ /*
+ * 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++;
+ 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;
+ }
/*
* Segment received on connection.
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) {
* 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>
* Enter SYN_RECEIVED state, and process any other fields of this
* segment in this state.
*/
- case TCPS_LISTEN:
+ case TCPS_LISTEN: {
+ struct mbuf *am;
+ register struct sockaddr_in *sin;
+
if (tiflags & TH_RST)
goto drop;
if (tiflags & TH_ACK)
goto dropwithreset;
if ((tiflags & TH_SYN) == 0)
goto drop;
- tcp_in.sin_addr = ti->ti_src;
- tcp_in.sin_port = ti->ti_sport;
+ if (in_broadcast(ti->ti_dst))
+ goto drop;
+ am = m_get(M_DONTWAIT, MT_SONAME);
+ if (am == NULL)
+ goto drop;
+ am->m_len = sizeof (struct sockaddr_in);
+ sin = mtod(am, struct sockaddr_in *);
+ sin->sin_family = AF_INET;
+ sin->sin_addr = ti->ti_src;
+ sin->sin_port = ti->ti_sport;
laddr = inp->inp_laddr;
- if (inp->inp_laddr.s_addr == 0)
+ if (inp->inp_laddr.s_addr == INADDR_ANY)
inp->inp_laddr = ti->ti_dst;
- if (in_pcbconnect(inp, (struct sockaddr_in *)&tcp_in)) {
+ if (in_pcbconnect(inp, am)) {
inp->inp_laddr = laddr;
+ (void) m_free(am);
goto drop;
}
+ (void) m_free(am);
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;
}
- in_setsockaddr(inp);
+ 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);
tcp_rcvseqinit(tp);
tp->t_state = TCPS_SYN_RECEIVED;
tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
+ dropsocket = 0; /* committed to socket */
goto trimthenstep6;
+ }
/*
* If the state is SYN_SENT:
*/
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) {
if (tiflags & TH_ACK)
- tcp_drop(tp, ECONNREFUSED);
+ tp = tcp_drop(tp, ECONNREFUSED);
goto drop;
}
if ((tiflags & TH_SYN) == 0)
tcp_rcvseqinit(tp);
tp->t_flags |= TF_ACKNOW;
if (SEQ_GT(tp->snd_una, tp->iss)) {
- if (so->so_options & SO_ACCEPTCONN)
- so->so_state |= SS_CONNAWAITING;
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;
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
tiflags &= ~TH_URG;
todrop--;
}
- if (todrop > ti->ti_len)
+ if (todrop > ti->ti_len ||
+ todrop == ti->ti_len && (tiflags&TH_FIN) == 0)
goto dropafterack;
m_adj(m, todrop);
ti->ti_seq += todrop;
*/
todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
if (todrop > 0) {
- if (todrop > ti->ti_len)
+ if (todrop >= ti->ti_len)
goto dropafterack;
m_adj(m, -todrop);
ti->ti_len -= todrop;
}
}
- /*
- * If a segment is received on a connection after the
- * user processes are gone, then RST the other end.
- */
- if (so->so_state & SS_USERGONE) {
- tcp_close(tp);
- goto dropwithreset;
- }
-
/*
* If the RST bit is set examine the state:
* SYN_RECEIVED STATE:
if (tiflags&TH_RST) switch (tp->t_state) {
case TCPS_SYN_RECEIVED:
- if (inp->inp_socket->so_options & SO_ACCEPTCONN) {
- /* a miniature tcp_close, but invisible to user */
- (void) m_free(dtom(tp->t_template));
- (void) m_free(dtom(tp));
- inp->inp_ppcb = 0;
- tp = tcp_newtcpcb(inp);
- tp->t_state = TCPS_LISTEN;
- inp->inp_faddr.s_addr = 0;
- inp->inp_fport = 0;
- inp->inp_laddr.s_addr = 0; /* not quite right */
- goto drop;
- }
- tcp_drop(tp, ECONNREFUSED);
+ tp = tcp_drop(tp, ECONNREFUSED);
goto drop;
case TCPS_ESTABLISHED:
case TCPS_FIN_WAIT_1:
case TCPS_FIN_WAIT_2:
case TCPS_CLOSE_WAIT:
- tcp_drop(tp, ECONNRESET);
+ tp = tcp_drop(tp, ECONNRESET);
goto drop;
case TCPS_CLOSING:
case TCPS_LAST_ACK:
case TCPS_TIME_WAIT:
- tcp_close(tp);
+ tp = tcp_close(tp);
goto drop;
}
* error and we send an RST and drop the connection.
*/
if (tiflags & TH_SYN) {
- tcp_drop(tp, ECONNRESET);
+ tp = tcp_drop(tp, ECONNRESET);
goto dropwithreset;
}
if (SEQ_LT(tp->snd_nxt, tp->snd_una))
tp->snd_nxt = tp->snd_una;
tp->t_timer[TCPT_REXMT] = 0;
- if (so->so_options & SO_ACCEPTCONN)
- so->so_state |= SS_CONNAWAITING;
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 ... */
tp->t_srtt =
tcp_alpha * tp->t_srtt +
(1 - tcp_alpha) * tp->t_rtt;
-/* printf("rtt %d srtt*100 now %d\n", tp->t_rtt, (int)(tp->t_srtt*100)); */
tp->t_rtt = 0;
}
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;
acked = 0;
}
- if (so->so_snd.sb_flags & SB_WAIT)
+ if ((so->so_snd.sb_flags & SB_WAIT) || so->so_snd.sb_sel)
sowwakeup(so);
tp->snd_una = ti->ti_ack;
if (SEQ_LT(tp->snd_nxt, tp->snd_una))
/*
* 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;
*/
case TCPS_LAST_ACK:
if (ourfinisacked)
- tcp_close(tp);
+ tp = tcp_close(tp);
goto drop;
/*
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;
}
/*
* Process segments with URG.
*/
- if ((tiflags & TH_URG) && TCPS_HAVERCVDFIN(tp->t_state) == 0) {
+ 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
(tp->rcv_up - tp->rcv_nxt) - 1;
if (so->so_oobmark == 0)
so->so_state |= SS_RCVATMARK;
-#ifdef TCPTRUEOOB
- if ((tp->t_flags & TF_DOOOB) == 0)
-#endif
- sohasoutofband(so);
- tp->t_oobflags &= ~TCPOOB_HAVEDATA;
+ sohasoutofband(so);
+ tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA);
}
/*
* Remove out of band data so doesn't get presented to user.
* but if two URG's are pending at once, some out-of-band
* data may creep in... ick.
*/
- if (ti->ti_urp <= ti->ti_len) {
+ if (ti->ti_urp <= ti->ti_len)
tcp_pulloutofband(so, ti);
- }
}
+badurp: /* XXX */
/*
* Process the segment text, merging it into the TCP sequencing queue,
*/
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
dropafterack:
/*
- * Generate an ACK dropping incoming segment.
- * Make ACK reflect our state.
+ * Generate an ACK dropping incoming segment if it occupies
+ * sequence space, where the ACK reflects our state.
*/
- if (tiflags & TH_RST)
+ if ((tiflags&TH_RST) ||
+ tlen == 0 && (tiflags&(TH_SYN|TH_FIN)) == 0)
goto drop;
+ if (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)
+ tcp_trace(TA_RESPOND, ostate, tp, &tcp_saveti, 0);
tcp_respond(tp, ti, tp->rcv_nxt, tp->snd_nxt, TH_ACK);
return;
dropwithreset:
- if (om)
+ if (om) {
(void) m_free(om);
+ om = 0;
+ }
/*
* 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);
else {
if (tiflags & TH_SYN)
ti->ti_len++;
- tcp_respond(tp, ti, ti->ti_seq+ti->ti_len, (tcp_seq)0, TH_RST|TH_ACK);
+ tcp_respond(tp, ti, ti->ti_seq+ti->ti_len, (tcp_seq)0,
+ TH_RST|TH_ACK);
}
+ /* destroy temporarily created socket */
+ if (dropsocket)
+ (void) soabort(so);
return;
drop:
+ if (om)
+ (void) m_free(om);
/*
* Drop space held by incoming segment and return.
*/
+ if (tp && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
+ tcp_trace(TA_DROP, ostate, tp, &tcp_saveti, 0);
m_freem(m);
+ /* destroy temporarily created socket */
+ if (dropsocket)
+ (void) soabort(so);
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;
break;
if (opt == TCPOPT_NOP)
optlen = 1;
- else
+ else {
optlen = cp[1];
+ if (optlen <= 0)
+ break;
+ }
switch (opt) {
default:
case TCPOPT_MAXSEG:
if (optlen != 4)
continue;
+ if (!(ti->ti_flags & TH_SYN))
+ continue;
tp->t_maxseg = *(u_short *)(cp + 2);
-#if vax
tp->t_maxseg = ntohs((u_short)tp->t_maxseg);
-#endif
- break;
-
-#ifdef TCPTRUEOOB
- case TCPOPT_WILLOOB:
- tp->t_flags |= TF_DOOOB;
-printf("tp %x dooob\n", tp);
- break;
-
- case TCPOPT_OOBDATA: {
- int seq;
- register struct socket *so = tp->t_inpcb->inp_socket;
- tcp_seq mark;
-
- if (optlen != 8)
- continue;
- seq = cp[2];
- if (seq < tp->t_iobseq)
- seq += 256;
-printf("oobdata cp[2] %d iobseq %d seq %d\n", cp[2], tp->t_iobseq, seq);
- if (seq - tp->t_iobseq > 128) {
-printf("bad seq\n");
- tp->t_oobflags |= TCPOOB_OWEACK;
- break;
- }
- tp->t_iobseq = cp[2];
- tp->t_iobc = cp[3];
- mark = *(tcp_seq *)(cp + 4);
-#if vax
- mark = ntohl(mark);
-#endif
- so->so_oobmark = so->so_rcv.sb_cc + (mark-tp->rcv_nxt);
- if (so->so_oobmark == 0)
- so->so_state |= SS_RCVATMARK;
-printf("take oob data %x input iobseq now %x\n", tp->t_iobc, tp->t_iobseq);
- sohasoutofband(so);
- break;
- }
-
- case TCPOPT_OOBACK: {
- int seq;
-
- if (optlen != 4)
- continue;
- if (tp->t_oobseq != cp[2]) {
-printf("wrong ack\n");
- break;
- }
-printf("take oob ack %x and cancel rexmt\n", cp[2]);
- tp->t_oobflags &= ~TCPOOB_NEEDACK;
- tp->t_timer[TCPT_OOBREXMT] = 0;
+ tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp));
break;
}
-#endif TCPTRUEOOB
- }
}
(void) m_free(om);
}
}
/*
- * 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;
-COUNT(TCP_REASS);
-
- /*
- * 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(tp), 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));
}