tcp_input.c

来自「eCos操作系统源码」· C语言 代码 · 共 2,368 行 · 第 1/5 页

		 * random urgent pointers, we'll crash in
		 * soreceive.  It's hard to imagine someone
		 * actually wanting to send this much urgent data.
		 */
		if (th->th_urp + so->so_rcv.sb_cc > sb_max) {
			th->th_urp = 0;			/* XXX */
			tiflags &= ~TH_URG;		/* XXX */
			goto dodata;			/* XXX */
		}
		/*
		 * If this segment advances the known urgent pointer,
		 * then mark the data stream.  This should not happen
		 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
		 * a FIN has been received from the remote side. 
		 * In these states we ignore the URG.
		 *
		 * According to RFC961 (Assigned Protocols),
		 * the urgent pointer points to the last octet
		 * of urgent data.  We continue, however,
		 * to consider it to indicate the first octet
		 * of data past the urgent section as the original 
		 * spec states (in one of two places).
		 */
		if (SEQ_GT(th->th_seq+th->th_urp, tp->rcv_up)) {
			tp->rcv_up = th->th_seq + th->th_urp;
			so->so_oobmark = so->so_rcv.sb_cc +
			    (tp->rcv_up - tp->rcv_nxt) - 1;
			if (so->so_oobmark == 0)
				so->so_state |= SS_RCVATMARK;
			sohasoutofband(so);
			tp->t_oobflags &= ~(TCPOOB_HAVEDATA | TCPOOB_HADDATA);
		}
		/*
		 * Remove out of band data so doesn't get presented to user.
		 * This can happen independent of advancing the URG pointer,
		 * but if two URG's are pending at once, some out-of-band
		 * data may creep in... ick.
		 */
		if (th->th_urp <= (u_int16_t) tlen
#ifdef SO_OOBINLINE
		     && (so->so_options & SO_OOBINLINE) == 0
#endif
		     )
		        tcp_pulloutofband(so, th->th_urp, m, hdroptlen);
	} else
		/*
		 * If no out of band data is expected,
		 * pull receive urgent pointer along
		 * with the receive window.
		 */
		if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
			tp->rcv_up = tp->rcv_nxt;
dodata:							/* XXX */

	/*
	 * Process the segment text, merging it into the TCP sequencing queue,
	 * and arranging for acknowledgment of receipt if necessary.
	 * This process logically involves adjusting tp->rcv_wnd as data
	 * is presented to the user (this happens in tcp_usrreq.c,
	 * case PRU_RCVD).  If a FIN has already been received on this
	 * connection then we just ignore the text.
	 */
	if ((tlen || (tiflags & TH_FIN)) &&
	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
		if (th->th_seq == tp->rcv_nxt && tp->segq.lh_first == NULL &&
		    tp->t_state == TCPS_ESTABLISHED) {
			if (th->th_flags & TH_PUSH)
				tp->t_flags |= TF_ACKNOW;
			else
				tp->t_flags |= TF_DELACK;
			tp->rcv_nxt += tlen;
			tiflags = th->th_flags & TH_FIN;
			tcpstat.tcps_rcvpack++;
			tcpstat.tcps_rcvbyte += tlen;
			ND6_HINT(tp);
			m_adj(m, hdroptlen);
			sbappend(&so->so_rcv, m);
			sorwakeup(so);
		} else {
			m_adj(m, hdroptlen);
			tiflags = tcp_reass(tp, th, m, &tlen);
			tp->t_flags |= TF_ACKNOW;
		}
#ifdef TCP_SACK
		if (!tp->sack_disable)
			tcp_update_sack_list(tp); 
#endif 

		/* 
		 * variable len never referenced again in modern BSD,
		 * so why bother computing it ??
		 */
#if 0
		/*
		 * Note the amount of data that peer has sent into
		 * our window, in order to estimate the sender's
		 * buffer size.
		 */
		len = so->so_rcv.sb_hiwat - (tp->rcv_adv - tp->rcv_nxt);
#endif /* 0 */
	} else {
		m_freem(m);
		tiflags &= ~TH_FIN;
	}

	/*
	 * If FIN is received ACK the FIN and let the user know
	 * that the connection is closing.  Ignore a FIN received before
	 * the connection is fully established.
	 */
	if ((tiflags & TH_FIN) && TCPS_HAVEESTABLISHED(tp->t_state)) {
		if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
			socantrcvmore(so);
			tp->t_flags |= TF_ACKNOW;
			tp->rcv_nxt++;
		}
		switch (tp->t_state) {

		/*
		 * In ESTABLISHED STATE enter the CLOSE_WAIT state.
		 */
		case TCPS_ESTABLISHED:
			tp->t_state = TCPS_CLOSE_WAIT;
			break;

		/*
		 * If still in FIN_WAIT_1 STATE FIN has not been acked so
		 * enter the CLOSING state.
		 */
		case TCPS_FIN_WAIT_1:
			tp->t_state = TCPS_CLOSING;
			break;

		/*
		 * In FIN_WAIT_2 state enter the TIME_WAIT state,
		 * starting the time-wait timer, turning off the other 
		 * standard timers.
		 */
		case TCPS_FIN_WAIT_2:
			tp->t_state = TCPS_TIME_WAIT;
			tcp_canceltimers(tp);
			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
			soisdisconnected(so);
			break;

		/*
		 * In TIME_WAIT state restart the 2 MSL time_wait timer.
		 */
		case TCPS_TIME_WAIT:
			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
			break;
		}
	}
#ifdef TCPDEBUG
	if (so->so_options & SO_DEBUG) {
#ifdef INET6
		if (tp->pf == PF_INET6)
			tcp_trace(TA_INPUT, ostate, tp, (caddr_t) &tcp_saveti6, 0, tlen);
		else
#endif /* INET6 */
			tcp_trace(TA_INPUT, ostate, tp, (caddr_t) &tcp_saveti, 0, tlen);
	}
#endif /* TCPDEBUG */

	/*
	 * Return any desired output.
	 */
	if (needoutput || (tp->t_flags & TF_ACKNOW)) {
		(void) tcp_output(tp);
	}
	return;

dropafterack:
	/*
	 * Generate an ACK dropping incoming segment if it occupies
	 * sequence space, where the ACK reflects our state.
	 */
	if (tiflags & TH_RST)
		goto drop;
	m_freem(m);
	tp->t_flags |= TF_ACKNOW;
	(void) tcp_output(tp);
	return;

dropwithreset:
	/*
	 * Generate a RST, dropping incoming segment.
	 * Make ACK acceptable to originator of segment.
	 * Don't bother to respond if destination was broadcast/multicast.
	 */
	if ((tiflags & TH_RST) || m->m_flags & (M_BCAST|M_MCAST))
	  goto drop;
#ifdef INET6
	if (is_ipv6) {
	  /* For following calls to tcp_respond */
	  ti = mtod(m, struct tcpiphdr *);
	  if (IN6_IS_ADDR_MULTICAST(&ipv6->ip6_dst))
	    goto drop;
	} else {
#endif /* INET6 */
	    if (IN_MULTICAST(ti->ti_dst.s_addr))
	      goto drop;
#ifdef INET6
	}
#endif /* INET6 */
	if (tiflags & TH_ACK)
		tcp_respond(tp, (caddr_t) ti, m, (tcp_seq)0, th->th_ack, TH_RST);
	else {
		if (tiflags & TH_SYN)
			tlen++;
		tcp_respond(tp, (caddr_t) ti, m, th->th_seq+tlen, (tcp_seq)0,
		    TH_RST|TH_ACK);
	}
	/* destroy temporarily created socket */
	if (dropsocket)
		(void) soabort(so);
	return;

drop:
	/*
	 * Drop space held by incoming segment and return.
	 */
#ifdef TCPDEBUG
	if (tp && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) {
#ifdef INET6
	  if (tp->pf == PF_INET6)
	    tcp_trace(TA_DROP, ostate, tp, (caddr_t) &tcp_saveti6, 0, tlen);
	  else
#endif /* INET6 */
	    tcp_trace(TA_DROP, ostate, tp, (caddr_t) &tcp_saveti, 0, tlen);
	}
#endif /* TCPDEBUG */

	m_freem(m);
	/* destroy temporarily created socket */
	if (dropsocket)
		(void) soabort(so);
	return;
#ifndef TUBA_INCLUDE
}

void
tcp_dooptions(tp, cp, cnt, th, ts_present, ts_val, ts_ecr)
	struct tcpcb *tp;
	u_char *cp;
	int cnt;
	struct tcphdr *th;
	int *ts_present;
	u_int32_t *ts_val, *ts_ecr;
{
	u_int16_t mss = 0;
	int opt, optlen;

	for (; cnt > 0; cnt -= optlen, cp += optlen) {
		opt = cp[0];
		if (opt == TCPOPT_EOL)
			break;
		if (opt == TCPOPT_NOP)
			optlen = 1;
		else {
			optlen = cp[1];
			if (optlen <= 0)
				break;
		}
		switch (opt) {

		default:
			continue;

		case TCPOPT_MAXSEG:
			if (optlen != TCPOLEN_MAXSEG)
				continue;
			if (!(th->th_flags & TH_SYN))
				continue;
			bcopy((char *) cp + 2, (char *) &mss, sizeof(mss));
			NTOHS(mss);
			break;

		case TCPOPT_WINDOW:
			if (optlen != TCPOLEN_WINDOW)
				continue;
			if (!(th->th_flags & TH_SYN))
				continue;
			tp->t_flags |= TF_RCVD_SCALE;
			tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
			break;

		case TCPOPT_TIMESTAMP:
			if (optlen != TCPOLEN_TIMESTAMP)
				continue;
			*ts_present = 1;
			bcopy((char *)cp + 2, (char *) ts_val, sizeof(*ts_val));
			NTOHL(*ts_val);
			bcopy((char *)cp + 6, (char *) ts_ecr, sizeof(*ts_ecr));
			NTOHL(*ts_ecr);

			/* 
			 * A timestamp received in a SYN makes
			 * it ok to send timestamp requests and replies.
			 */
			if (th->th_flags & TH_SYN) {
				tp->t_flags |= TF_RCVD_TSTMP;
				tp->ts_recent = *ts_val;
				tp->ts_recent_age = tcp_now;
			}
			break;
		
#ifdef TCP_SACK 
		case TCPOPT_SACK_PERMITTED:
			if (tp->sack_disable || optlen!=TCPOLEN_SACK_PERMITTED)
				continue;
			if (th->th_flags & TH_SYN)
				/* MUST only be set on SYN */
				tp->t_flags |= TF_SACK_PERMIT;
			break;
		case TCPOPT_SACK:
			if (tcp_sack_option(tp, th, cp, optlen))
				continue;
			break;
#endif          
		}
	}
	/* Update t_maxopd and t_maxseg after all options are processed */
	if (th->th_flags & TH_SYN)
		(void) tcp_mss(tp, mss);	/* sets t_maxseg */
}

#if defined(TCP_SACK) || defined(TCP_NEWRENO)
u_long 
tcp_seq_subtract(a, b)
	u_long a, b;
{ 
	return ((long)(a - b)); 
}
#endif


#ifdef TCP_SACK 
/*
 * This function is called upon receipt of new valid data (while not in header
 * prediction mode), and it updates the ordered list of sacks. 
 */
void 
tcp_update_sack_list(tp)
	struct tcpcb *tp; 
{    
	/* 
	 * First reported block MUST be the most recent one.  Subsequent
	 * blocks SHOULD be in the order in which they arrived at the
	 * receiver.  These two conditions make the implementation fully
	 * compliant with RFC 2018.
	 */     
	int i, j = 0, count = 0, lastpos = -1;
	struct sackblk sack, firstsack, temp[MAX_SACK_BLKS];
    
	/* First clean up current list of sacks */
	for (i = 0; i < tp->rcv_numsacks; i++) {
		sack = tp->sackblks[i];
		if (sack.start == 0 && sack.end == 0) {
			count++; /* count = number of blocks to be discarded */
			continue;
		}
		if (SEQ_LEQ(sack.end, tp->rcv_nxt)) {
			tp->sackblks[i].start = tp->sackblks[i].end = 0;
			count++;
		} else { 
			temp[j].start = tp->sackblks[i].start;
			temp[j++].end = tp->sackblks[i].end;
		}
	}   
	tp->rcv_numsacks -= count;
	if (tp->rcv_numsacks == 0) { /* no sack blocks currently (fast path) */
		tcp_clean_sackreport(tp);
		if (SEQ_LT(tp->rcv_nxt, tp->rcv_laststart)) {
			/* ==> need first sack block */
			tp->sackblks[0].start = tp->rcv_laststart;
			tp->sackblks[0].end = tp->rcv_lastend;
			tp->rcv_numsacks = 1;
		}
		return;
	}
	/* Otherwise, sack blocks are already present. */
	for (i = 0; i < tp->rcv_numsacks; i++)
		tp->sackblks[i] = temp[i]; /* first copy back sack list */
	if (SEQ_GEQ(tp->rcv_nxt, tp->rcv_lastend)) 
		return;     /* sack list remains unchanged */
	/* 
	 * From here, segment just received should be (part of) the 1st sack.
	 * Go through list, possibly coalescing sack block entries.
	 */
	firstsack.start = tp->rcv_laststart;
	firstsack.end = tp->rcv_lastend;
	for (i = 0; i < tp->rcv_numsacks; i++) {
		sack = tp->sackblks[i];
		if (SEQ_LT(sack.end, firstsack.start) ||
		    SEQ_GT(sack.start, firstsack.end))
			continue; /* no overlap */
		if (sack.start == firstsack.start && sack.end == firstsack.end){
			/* 
			 * identical block; delete it here since we will
			 * move it to the front of the list.
			 */
			tp->sackblks[i].start = tp->sackblks[i].end = 0;
			lastpos = i;    /* last posn with a zero entry */
			continue;
		}
		if (SEQ_LEQ(sack.start, firstsack.start))
			firstsack.start = sack.start; /* merge blocks */
		if (SEQ_GEQ(sack.end, firstsack.end))
			firstsack.end = sack.end;     /* merge blocks */
		tp->sackblks[i].start = tp->sackblks[i].end = 0;
		lastpos = i;    /* last posn with a zero entry */
	}
	if (lastpos != -1) {    /* at least one merge */
		for (i = 0, j = 1; i < tp->rcv_numsacks; i++) {
			sack = tp->sackblks[i];
			if (sack.start == 0 && sack.end == 0)
				continue;
			temp[j++] = sack;
		}
		tp->rcv_numsacks = j; /* including first blk (added later) */
		for (i = 1; i < tp->rcv_numsacks; i++) /* now copy back */
			tp->sackblks[i] = temp[i];
	} else {        /* no merges -- shift sacks by 1 */
		if (tp->rcv_numsacks < MAX_SACK_BLKS)
			tp->rcv_numsacks++;
		for (i = tp->rcv_numsacks-1; i > 0; i--)
			tp->sackblks[i] = tp->sackblks[i-1];
	}
	tp->sackblks[0] = firstsack;
	return;
}  

/*
 * Process the TCP SACK option.  Returns 1 if tcp_dooptions() should continue,
 * and 0 otherwise, if the option was fine.  tp->snd_holes is an ordered list
 * of holes (oldest to newest, in terms of the sequence space).  
 */             
int
tcp_sack_option(tp, th, cp, optlen)
	struct tcpcb *tp;
	struct tcphdr *th;
	u_char *cp;
	int    optlen;
{       
	int tmp_olen;
	u_char *tmp_cp;
	struct sackhole *cur, *p, *temp;
   
	if (tp->sack_disable)
		return 1;
           
	/* Note: TCPOLEN_SACK must be 2*sizeof(tcp_seq) */
	if (optlen <= 2 || (optlen - 2) % TCPOLEN_SACK != 0)
		return 1;
	tmp_cp = cp + 2;
	tmp_olen = optlen - 2;
	if (tp->snd_numholes < 0)
		tp->snd_numholes = 0;
	if (tp->t_maxseg == 0)
		panic("tcp_sack_option"); /* Should never happen */
	while (tmp_olen > 0) {
		struct sackblk sack;
            
		bcopy((char *) tmp_cp, (char *) &(sack.start), sizeof(tcp_seq));
		NTOHL(sack.start); 
		bcopy((char *) tmp_cp + sizeof(tcp_seq),
		    (char *) &(sack.end), sizeof(tcp_seq));
		NTOHL(sack.end);
		tmp_olen -= TCPOLEN_SACK;
		tmp_cp += TCPOLEN_SACK;
		if (SEQ