spp_usrreq.c

早期freebsd实现

/*
 * Copyright (c) 1984, 1985, 1986, 1987, 1993
 *	The Regents of the University of California.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)spp_usrreq.c	8.1 (Berkeley) 6/10/93
 */

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/malloc.h>
#include <sys/mbuf.h>
#include <sys/protosw.h>
#include <sys/socket.h>
#include <sys/socketvar.h>
#include <sys/errno.h>

#include <net/if.h>
#include <net/route.h>
#include <netinet/tcp_fsm.h>

#include <netns/ns.h>
#include <netns/ns_pcb.h>
#include <netns/idp.h>
#include <netns/idp_var.h>
#include <netns/ns_error.h>
#include <netns/sp.h>
#include <netns/spidp.h>
#include <netns/spp_timer.h>
#include <netns/spp_var.h>
#include <netns/spp_debug.h>

/*
 * SP protocol implementation.
 */
spp_init()
{

	spp_iss = 1; /* WRONG !! should fish it out of TODR */
}
struct spidp spp_savesi;
int traceallspps = 0;
extern int sppconsdebug;
int spp_hardnosed;
int spp_use_delack = 0;
u_short spp_newchecks[50];

/*ARGSUSED*/
spp_input(m, nsp)
	register struct mbuf *m;
	register struct nspcb *nsp;
{
	register struct sppcb *cb;
	register struct spidp *si = mtod(m, struct spidp *);
	register struct socket *so;
	short ostate;
	int dropsocket = 0;


	sppstat.spps_rcvtotal++;
	if (nsp == 0) {
		panic("No nspcb in spp_input\n");
		return;
	}

	cb = nstosppcb(nsp);
	if (cb == 0) goto bad;

	if (m->m_len < sizeof(*si)) {
		if ((m = m_pullup(m, sizeof(*si))) == 0) {
			sppstat.spps_rcvshort++;
			return;
		}
		si = mtod(m, struct spidp *);
	}
	si->si_seq = ntohs(si->si_seq);
	si->si_ack = ntohs(si->si_ack);
	si->si_alo = ntohs(si->si_alo);

	so = nsp->nsp_socket;
	if (so->so_options & SO_DEBUG || traceallspps) {
		ostate = cb->s_state;
		spp_savesi = *si;
	}
	if (so->so_options & SO_ACCEPTCONN) {
		struct sppcb *ocb = cb;

		so = sonewconn(so, 0);
		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++;
		nsp = (struct nspcb *)so->so_pcb;
		nsp->nsp_laddr = si->si_dna;
		cb = nstosppcb(nsp);
		cb->s_mtu = ocb->s_mtu;		/* preserve sockopts */
		cb->s_flags = ocb->s_flags;	/* preserve sockopts */
		cb->s_flags2 = ocb->s_flags2;	/* preserve sockopts */
		cb->s_state = TCPS_LISTEN;
	}

	/*
	 * Packet received on connection.
	 * reset idle time and keep-alive timer;
	 */
	cb->s_idle = 0;
	cb->s_timer[SPPT_KEEP] = SPPTV_KEEP;

	switch (cb->s_state) {

	case TCPS_LISTEN:{
		struct mbuf *am;
		register struct sockaddr_ns *sns;
		struct ns_addr laddr;

		/*
		 * If somebody here was carying on a conversation
		 * and went away, and his pen pal thinks he can
		 * still talk, we get the misdirected packet.
		 */
		if (spp_hardnosed && (si->si_did != 0 || si->si_seq != 0)) {
			spp_istat.gonawy++;
			goto dropwithreset;
		}
		am = m_get(M_DONTWAIT, MT_SONAME);
		if (am == NULL)
			goto drop;
		am->m_len = sizeof (struct sockaddr_ns);
		sns = mtod(am, struct sockaddr_ns *);
		sns->sns_len = sizeof(*sns);
		sns->sns_family = AF_NS;
		sns->sns_addr = si->si_sna;
		laddr = nsp->nsp_laddr;
		if (ns_nullhost(laddr))
			nsp->nsp_laddr = si->si_dna;
		if (ns_pcbconnect(nsp, am)) {
			nsp->nsp_laddr = laddr;
			(void) m_free(am);
			spp_istat.noconn++;
			goto drop;
		}
		(void) m_free(am);
		spp_template(cb);
		dropsocket = 0;		/* committed to socket */
		cb->s_did = si->si_sid;
		cb->s_rack = si->si_ack;
		cb->s_ralo = si->si_alo;
#define THREEWAYSHAKE
#ifdef THREEWAYSHAKE
		cb->s_state = TCPS_SYN_RECEIVED;
		cb->s_force = 1 + SPPT_KEEP;
		sppstat.spps_accepts++;
		cb->s_timer[SPPT_KEEP] = SPPTV_KEEP;
		}
		break;
	/*
	 * This state means that we have heard a response
	 * to our acceptance of their connection
	 * It is probably logically unnecessary in this
	 * implementation.
	 */
	 case TCPS_SYN_RECEIVED: {
		if (si->si_did!=cb->s_sid) {
			spp_istat.wrncon++;
			goto drop;
		}
#endif
		nsp->nsp_fport =  si->si_sport;
		cb->s_timer[SPPT_REXMT] = 0;
		cb->s_timer[SPPT_KEEP] = SPPTV_KEEP;
		soisconnected(so);
		cb->s_state = TCPS_ESTABLISHED;
		sppstat.spps_accepts++;
		}
		break;

	/*
	 * This state means that we have gotten a response
	 * to our attempt to establish a connection.
	 * We fill in the data from the other side,
	 * telling us which port to respond to, instead of the well-
	 * known one we might have sent to in the first place.
	 * We also require that this is a response to our
	 * connection id.
	 */
	case TCPS_SYN_SENT:
		if (si->si_did!=cb->s_sid) {
			spp_istat.notme++;
			goto drop;
		}
		sppstat.spps_connects++;
		cb->s_did = si->si_sid;
		cb->s_rack = si->si_ack;
		cb->s_ralo = si->si_alo;
		cb->s_dport = nsp->nsp_fport =  si->si_sport;
		cb->s_timer[SPPT_REXMT] = 0;
		cb->s_flags |= SF_ACKNOW;
		soisconnected(so);
		cb->s_state = TCPS_ESTABLISHED;
		/* Use roundtrip time of connection request for initial rtt */
		if (cb->s_rtt) {
			cb->s_srtt = cb->s_rtt << 3;
			cb->s_rttvar = cb->s_rtt << 1;
			SPPT_RANGESET(cb->s_rxtcur,
			    ((cb->s_srtt >> 2) + cb->s_rttvar) >> 1,
			    SPPTV_MIN, SPPTV_REXMTMAX);
			    cb->s_rtt = 0;
		}
	}
	if (so->so_options & SO_DEBUG || traceallspps)
		spp_trace(SA_INPUT, (u_char)ostate, cb, &spp_savesi, 0);

	m->m_len -= sizeof (struct idp);
	m->m_pkthdr.len -= sizeof (struct idp);
	m->m_data += sizeof (struct idp);

	if (spp_reass(cb, si)) {
		(void) m_freem(m);
	}
	if (cb->s_force || (cb->s_flags & (SF_ACKNOW|SF_WIN|SF_RXT)))
		(void) spp_output(cb, (struct mbuf *)0);
	cb->s_flags &= ~(SF_WIN|SF_RXT);
	return;

dropwithreset:
	if (dropsocket)
		(void) soabort(so);
	si->si_seq = ntohs(si->si_seq);
	si->si_ack = ntohs(si->si_ack);
	si->si_alo = ntohs(si->si_alo);
	ns_error(dtom(si), NS_ERR_NOSOCK, 0);
	if (cb->s_nspcb->nsp_socket->so_options & SO_DEBUG || traceallspps)
		spp_trace(SA_DROP, (u_char)ostate, cb, &spp_savesi, 0);
	return;

drop:
bad:
	if (cb == 0 || cb->s_nspcb->nsp_socket->so_options & SO_DEBUG ||
            traceallspps)
		spp_trace(SA_DROP, (u_char)ostate, cb, &spp_savesi, 0);
	m_freem(m);
}

int spprexmtthresh = 3;

/*
 * This is structurally similar to the tcp reassembly routine
 * but its function is somewhat different:  It merely queues
 * packets up, and suppresses duplicates.
 */
spp_reass(cb, si)
register struct sppcb *cb;
register struct spidp *si;
{
	register struct spidp_q *q;
	register struct mbuf *m;
	register struct socket *so = cb->s_nspcb->nsp_socket;
	char packetp = cb->s_flags & SF_HI;
	int incr;
	char wakeup = 0;

	if (si == SI(0))
		goto present;
	/*
	 * Update our news from them.
	 */
	if (si->si_cc & SP_SA)
		cb->s_flags |= (spp_use_delack ? SF_DELACK : SF_ACKNOW);
	if (SSEQ_GT(si->si_alo, cb->s_ralo))
		cb->s_flags |= SF_WIN;
	if (SSEQ_LEQ(si->si_ack, cb->s_rack)) {
		if ((si->si_cc & SP_SP) && cb->s_rack != (cb->s_smax + 1)) {
			sppstat.spps_rcvdupack++;
			/*
			 * If this is a completely duplicate ack
			 * and other conditions hold, we assume
			 * a packet has been dropped and retransmit
			 * it exactly as in tcp_input().
			 */
			if (si->si_ack != cb->s_rack ||
			    si->si_alo != cb->s_ralo)
				cb->s_dupacks = 0;
			else if (++cb->s_dupacks == spprexmtthresh) {
				u_short onxt = cb->s_snxt;
				int cwnd = cb->s_cwnd;

				cb->s_snxt = si->si_ack;
				cb->s_cwnd = CUNIT;
				cb->s_force = 1 + SPPT_REXMT;
				(void) spp_output(cb, (struct mbuf *)0);
				cb->s_timer[SPPT_REXMT] = cb->s_rxtcur;
				cb->s_rtt = 0;
				if (cwnd >= 4 * CUNIT)
					cb->s_cwnd = cwnd / 2;
				if (SSEQ_GT(onxt, cb->s_snxt))
					cb->s_snxt = onxt;
				return (1);
			}
		} else
			cb->s_dupacks = 0;
		goto update_window;
	}
	cb->s_dupacks = 0;
	/*
	 * If our correspondent acknowledges data we haven't sent
	 * TCP would drop the packet after acking.  We'll be a little
	 * more permissive
	 */
	if (SSEQ_GT(si->si_ack, (cb->s_smax + 1))) {
		sppstat.spps_rcvacktoomuch++;
		si->si_ack = cb->s_smax + 1;
	}
	sppstat.spps_rcvackpack++;
	/*
	 * If transmit timer is running and timed sequence
	 * number was acked, update smoothed round trip time.
	 * See discussion of algorithm in tcp_input.c
	 */
	if (cb->s_rtt && SSEQ_GT(si->si_ack, cb->s_rtseq)) {
		sppstat.spps_rttupdated++;
		if (cb->s_srtt != 0) {
			register short delta;
			delta = cb->s_rtt - (cb->s_srtt >> 3);
			if ((cb->s_srtt += delta) <= 0)
				cb->s_srtt = 1;
			if (delta < 0)
				delta = -delta;
			delta -= (cb->s_rttvar >> 2);
			if ((cb->s_rttvar += delta) <= 0)
				cb->s_rttvar = 1;
		} else {
			/*
			 * No rtt measurement yet
			 */
			cb->s_srtt = cb->s_rtt << 3;
			cb->s_rttvar = cb->s_rtt << 1;
		}
		cb->s_rtt = 0;
		cb->s_rxtshift = 0;
		SPPT_RANGESET(cb->s_rxtcur,
			((cb->s_srtt >> 2) + cb->s_rttvar) >> 1,
			SPPTV_MIN, SPPTV_REXMTMAX);
	}
	/*
	 * If all outstanding data is acked, stop retransmit
	 * timer and remember to restart (more output or persist).
	 * If there is more data to be acked, restart retransmit
	 * timer, using current (possibly backed-off) value;
	 */
	if (si->si_ack == cb->s_smax + 1) {
		cb->s_timer[SPPT_REXMT] = 0;
		cb->s_flags |= SF_RXT;
	} else if (cb->s_timer[SPPT_PERSIST] == 0)
		cb->s_timer[SPPT_REXMT] = cb->s_rxtcur;
	/*
	 * When new data is acked, open the congestion window.
	 * If the window gives us less than ssthresh packets
	 * in flight, open exponentially (maxseg at a time).
	 * Otherwise open linearly (maxseg^2 / cwnd at a time).
	 */
	incr = CUNIT;
	if (cb->s_cwnd > cb->s_ssthresh)
		incr = max(incr * incr / cb->s_cwnd, 1);
	cb->s_cwnd = min(cb->s_cwnd + incr, cb->s_cwmx);
	/*
	 * Trim Acked data from output queue.
	 */
	while ((m = so->so_snd.sb_mb) != NULL) {
		if (SSEQ_LT((mtod(m, struct spidp *))->si_seq, si->si_ack))
			sbdroprecord(&so->so_snd);
		else
			break;
	}
	sowwakeup(so);
	cb->s_rack = si->si_ack;
update_window:
	if (SSEQ_LT(cb->s_snxt, cb->s_rack))
		cb->s_snxt = cb->s_rack;
	if (SSEQ_LT(cb->s_swl1, si->si_seq) || cb->s_swl1 == si->si_seq &&
	    (SSEQ_LT(cb->s_swl2, si->si_ack) ||
	     cb->s_swl2 == si->si_ack && SSEQ_LT(cb->s_ralo, si->si_alo))) {
		/* keep track of pure window updates */
		if ((si->si_cc & SP_SP) && cb->s_swl2 == si->si_ack
		    && SSEQ_LT(cb->s_ralo, si->si_alo)) {
			sppstat.spps_rcvwinupd++;
			sppstat.spps_rcvdupack--;
		}
		cb->s_ralo = si->si_alo;
		cb->s_swl1 = si->si_seq;
		cb->s_swl2 = si->si_ack;
		cb->s_swnd = (1 + si->si_alo - si->si_ack);
		if (cb->s_swnd > cb->s_smxw)
			cb->s_smxw = cb->s_swnd;
		cb->s_flags |= SF_WIN;
	}
	/*
	 * If this packet number is higher than that which
	 * we have allocated refuse it, unless urgent
	 */
	if (SSEQ_GT(si->si_seq, cb->s_alo)) {
		if (si->si_cc & SP_SP) {
			sppstat.spps_rcvwinprobe++;
			return (1);
		} else
			sppstat.spps_rcvpackafterwin++;
		if (si->si_cc & SP_OB) {
			if (SSEQ_GT(si->si_seq, cb->s_alo + 60)) {
				ns_error(dtom(si), NS_ERR_FULLUP, 0);
				return (0);
			} /* else queue this packet; */
		} else {
			/*register struct socket *so = cb->s_nspcb->nsp_socket;
			if (so->so_state && SS_NOFDREF) {
				ns_error(dtom(si), NS_ERR_NOSOCK, 0);
				(void)spp_close(cb);
			} else
				       would crash system*/
			spp_istat.notyet++;
			ns_error(dtom(si), NS_ERR_FULLUP, 0);
			return (0);
		}
	}
	/*
	 * If this is a system packet, we don't need to
	 * queue it up, and won't update acknowledge #
	 */
	if (si->si_cc & SP_SP) {
		return (1);
	}
	/*
	 * We have already seen this packet, so drop.
	 */
	if (SSEQ_LT(si->si_seq, cb->s_ack)) {
		spp_istat.bdreas++;
		sppstat.spps_rcvduppack++;
		if (si->si_seq == cb->s_ack - 1)
			spp_istat.lstdup++;
		return (1);
	}
	/*
	 * Loop through all packets queued up to insert in
	 * appropriate sequence.
	 */
	for (q = cb->s_q.si_next; q!=&cb->s_q; q = q->si_next) {
		if (si->si_seq == SI(q)->si_seq) {
			sppstat.spps_rcvduppack++;
			return (1);
		}
		if (SSEQ_LT(si->si_seq, SI(q)->si_seq)) {
			sppstat.spps_rcvoopack++;
			break;
		}
	}
	insque(si, q->si_prev);
	/*
	 * If this packet is urgent, inform process
	 */
	if (si->si_cc & SP_OB) {
		cb->s_iobc = ((char *)si)[1 + sizeof(*si)];
		sohasoutofband(so);
		cb->s_oobflags |= SF_IOOB;
	}
present:
#define SPINC sizeof(struct sphdr)
	/*
	 * Loop through all packets queued up to update acknowledge
	 * number, and present all acknowledged data to user;
	 * If in packet interface mode, show packet headers.
	 */
	for (q = cb->s_q.si_next; q!=&cb->s_q; q = q->si_next) {
		  if (SI(q)->si_seq == cb->s_ack) {
			cb->s_ack++;
			m = dtom(q);
			if (SI(q)->si_cc & SP_OB) {
				cb->s_oobflags &= ~SF_IOOB;
				if (so->so_rcv.sb_cc)
					so->so_oobmark = so->so_rcv.sb_cc;
				else
					so->so_state |= SS_RCVATMARK;
			}
			q = q->si_prev;
			remque(q->si_next);
			wakeup = 1;
			sppstat.spps_rcvpack++;
#ifdef SF_NEWCALL
			if (cb->s_flags2 & SF_NEWCALL) {
				struct sphdr *sp = mtod(m, struct sphdr *);
				u_char dt = sp->sp_dt;
				spp_newchecks[4]++;
				if (dt != cb->s_rhdr.sp_dt) {
					struct mbuf *mm =
					   m_getclr(M_DONTWAIT, MT_CONTROL);
					spp_newchecks[0]++;
					if (mm != NULL) {
						u_short *s =
							mtod(mm, u_short *);
						cb->s_rhdr.sp_dt = dt;
						mm->m_len = 5; /*XXX*/
						s[0] = 5;
						s[1] = 1;
						*(u_char *)(&s[2]) = dt;
						sbappend(&so->so_rcv, mm);
					}
				}
				if (sp->sp_cc & SP_OB) {
					MCHTYPE(m, MT_OOBDATA);
					spp_newchecks[1]++;
					so->so_oobmark = 0;
					so->so_state &= ~SS_RCVATMARK;
				}
				if (packetp == 0) {
					m->m_data += SPINC;
					m->m_len -= SPINC;
					m->m_pkthdr.len -= SPINC;
				}
				if ((sp->sp_cc & SP_EM) || packetp) {
					sbappendrecord(&so->so_rcv, m);
					spp_newchecks[9]++;
				} else
					sbappend(&so->so_rcv, m);
			} else
#endif
			if (packetp) {
				sbappendrecord(&so->so_rcv, m);
			} else {
				cb->s_rhdr = *mtod(m, struct sphdr *);
				m->m_data += SPINC;
				m->m_len -= SPINC;
				m->m_pkthdr.len -= SPINC;
				sbappend(&so->so_rcv, m);
			}
		  } else
			break;
	}
	if (wakeup) sorwakeup(so);
	return (0);
}

spp_ctlinput(cmd, arg)
	int cmd;
	caddr_t arg;
{
	struct ns_addr *na;
	extern u_char nsctlerrmap[];
	extern spp_abort(), spp_quench();
	extern struct nspcb *idp_drop();
	struct ns_errp *errp;
	struct nspcb *nsp;
	struct sockaddr_ns *sns;
	int type;

	if (cmd < 0 || cmd > PRC_NCMDS)
		return;
	type = NS_ERR_UNREACH_HOST;

	switch (cmd) {

	case PRC_ROUTEDEAD:
		return;

	case PRC_IFDOWN:
	case PRC_HOSTDEAD:
	case PRC_HOSTUNREACH:
		sns = (struct sockaddr_ns *)arg;