static int
tcp_reass(register struct tcpcb *tp, register struct tcpiphdr *ti,
struct mbuf *m)
{
register struct tcpiphdr *q;
struct socket *so = tp->t_socket;
int flags;
/*
* Call with ti==NULL after become established to
* force pre-ESTABLISHED data up to user socket.
*/
if (ti == NULL)
goto present;
/*
* Find a segment which begins after this one does.
*/
for (q = tcpfrag_list_first(tp); !tcpfrag_list_end(q, tp);
q = tcpiphdr_next(q))
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 (!tcpfrag_list_end(tcpiphdr_prev(q), tp)) {
register int i;
q = tcpiphdr_prev(q);
/* 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) {
m_free(m);
/*
* Try to present any queued data
* at the left window edge to the user.
* This is needed after the 3-WHS
* completes.
*/
goto present; /* ??? */
}
m_adj(m, i);
ti->ti_len -= i;
ti->ti_seq += i;
}
q = tcpiphdr_next(q);
}
ti->ti_mbuf = m;
/*
* While we overlap succeeding segments trim them or,
* if they are completely covered, dequeue them.
*/
while (!tcpfrag_list_end(q, 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(q->ti_mbuf, i);
break;
}
q = tcpiphdr_next(q);
m = tcpiphdr_prev(q)->ti_mbuf;
remque(tcpiphdr2qlink(tcpiphdr_prev(q)));
m_free(m);
}
/*
* Stick new segment in its place.
*/
insque(tcpiphdr2qlink(ti), tcpiphdr2qlink(tcpiphdr_prev(q)));
present:
/*
* Present data to user, advancing rcv_nxt through
* completed sequence space.
*/
if (!TCPS_HAVEESTABLISHED(tp->t_state))
return (0);
ti = tcpfrag_list_first(tp);
if (tcpfrag_list_end(ti, 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(tcpiphdr2qlink(ti));
m = ti->ti_mbuf;
ti = tcpiphdr_next(ti);
if (so->so_state & SS_FCANTSENDMORE)
m_free(m);
else {
if (so->so_emu) {
if (tcp_emu(so,m)) sbappend(so, m);
} else
sbappend(so, m);
}
} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
return (flags);
}
/*
* TCP input routine, follows pages 65-76 of the
* protocol specification dated September, 1981 very closely.
*/
void
tcp_input(struct mbuf *m, int iphlen, struct socket *inso)
{
struct ip save_ip, *ip;
register struct tcpiphdr *ti;
caddr_t optp = NULL;
int optlen = 0;
int len, tlen, off;
register struct tcpcb *tp = NULL;
register int tiflags;
struct socket *so = NULL;
int todrop, acked, ourfinisacked, needoutput = 0;
int iss = 0;
u_long tiwin;
int ret;
struct ex_list *ex_ptr;
Slirp *slirp;
DEBUG_CALL("tcp_input");
DEBUG_ARGS((dfd, " m = %8lx iphlen = %2d inso = %lx\n",
(long )m, iphlen, (long )inso ));
/*
* If called with m == 0, then we're continuing the connect
*/
if (m == NULL) {
so = inso;
slirp = so->slirp;
/* Re-set a few variables */
tp = sototcpcb(so);
m = so->so_m;
so->so_m = NULL;
ti = so->so_ti;
tiwin = ti->ti_win;
tiflags = ti->ti_flags;
goto cont_conn;
}
slirp = m->slirp;
/*
* Get IP and TCP header together in first mbuf.
* Note: IP leaves IP header in first mbuf.
*/
ti = mtod(m, struct tcpiphdr *);
if (iphlen > sizeof(struct ip )) {
ip_stripoptions(m, (struct mbuf *)0);
iphlen=sizeof(struct ip );
}
/* XXX Check if too short */
/*
* Save a copy of the IP header in case we want restore it
* for sending an ICMP error message in response.
*/
ip=mtod(m, struct ip *);
save_ip = *ip;
save_ip.ip_len+= iphlen;
/*
* Checksum extended TCP header and data.
*/
tlen = ((struct ip *)ti)->ip_len;
tcpiphdr2qlink(ti)->next = tcpiphdr2qlink(ti)->prev = NULL;
memset(&ti->ti_i.ih_mbuf, 0 , sizeof(struct mbuf_ptr));
ti->ti_x1 = 0;
ti->ti_len = htons((uint16_t)tlen);
len = sizeof(struct ip ) + tlen;
if(cksum(m, len)) {
goto drop;
}
/*
* Check that TCP offset makes sense,
* pull out TCP options and adjust length. XXX
*/
off = ti->ti_off << 2;
if (off < sizeof (struct tcphdr) || off > tlen) {
goto drop;
}
tlen -= off;
ti->ti_len = tlen;
if (off > sizeof (struct tcphdr)) {
optlen = off - sizeof (struct tcphdr);
optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
}
tiflags = ti->ti_flags;
/*
* Convert TCP protocol specific fields to host format.
*/
NTOHL(ti->ti_seq);
NTOHL(ti->ti_ack);
NTOHS(ti->ti_win);
NTOHS(ti->ti_urp);
/*
* Drop TCP, IP headers and TCP options.
*/
m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
/*
* Locate pcb for segment.
*/
findso:
so = slirp->tcp_last_so;
if (so->so_fport != ti->ti_dport ||
so->so_lport != ti->ti_sport ||
so->so_laddr.s_addr != ti->ti_src.s_addr ||
so->so_faddr.s_addr != ti->ti_dst.s_addr) {
so = solookup(&slirp->tcb, ti->ti_src, ti->ti_sport,
ti->ti_dst, ti->ti_dport);
if (so)
slirp->tcp_last_so = so;
}
/*
* If the state is CLOSED (i.e., TCB does not exist) then
* all data in the incoming segment is discarded.
* If the TCB exists but is in CLOSED state, it is embryonic,
* but should either do a listen or a connect soon.
*
* state == CLOSED means we've done socreate() but haven't
* attached it to a protocol yet...
*
* XXX If a TCB does not exist, and the TH_SYN flag is
* the only flag set, then create a session, mark it
* as if it was LISTENING, and continue...
*/
if (so == NULL) {
if (slirp->restricted) {
/* Any hostfwds will have an existing socket, so we only get here
* for non-hostfwd connections. These should be dropped, unless it
* happens to be a guestfwd.
*/
for (ex_ptr = slirp->exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
if (ex_ptr->ex_fport == ti->ti_dport &&
ti->ti_dst.s_addr == ex_ptr->ex_addr.s_addr) {
break;
}
}
if (!ex_ptr) {
goto dropwithreset;
}
}
if ((tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) != TH_SYN)
goto dropwithreset;
if ((so = socreate(slirp)) == NULL)
goto dropwithreset;
if (tcp_attach(so) < 0) {
free(so); /* Not sofree (if it failed, it's not insqued) */
goto dropwithreset;
}
sbreserve(&so->so_snd, TCP_SNDSPACE);
sbreserve(&so->so_rcv, TCP_RCVSPACE);
so->so_laddr = ti->ti_src;
so->so_lport = ti->ti_sport;
so->so_faddr = ti->ti_dst;
so->so_fport = ti->ti_dport;
if ((so->so_iptos = tcp_tos(so)) == 0)
so->so_iptos = ((struct ip *)ti)->ip_tos;
tp = sototcpcb(so);
tp->t_state = TCPS_LISTEN;
}
/*
* If this is a still-connecting socket, this probably
* a retransmit of the SYN. Whether it's a retransmit SYN
* or something else, we nuke it.
*/
if (so->so_state & SS_ISFCONNECTING)
goto drop;
tp = sototcpcb(so);
/* XXX Should never fail */
if (tp == NULL)
goto dropwithreset;
if (tp->t_state == TCPS_CLOSED)
goto drop;
tiwin = ti->ti_win;
/*
* Segment received on connection.
* Reset idle time and keep-alive timer.
*/
tp->t_idle = 0;
if (SO_OPTIONS)
tp->t_timer[TCPT_KEEP] = TCPTV_KEEPINTVL;
else
tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_IDLE;
/*
* Process options if not in LISTEN state,
* else do it below (after getting remote address).
*/
if (optp && tp->t_state != TCPS_LISTEN)
tcp_dooptions(tp, (u_char *)optp, optlen, ti);
/*
* Header prediction: check for the two common cases
* of a uni-directional data xfer. If the packet has
* no control flags, is in-sequence, the window didn't
* change and we're not retransmitting, it's a
* candidate. If the length is zero and the ack moved
* forward, we're the sender side of the xfer. Just
* free the data acked & wake any higher level process
* that was blocked waiting for space. If the length
* is non-zero and the ack didn't move, we're the
* receiver side. If we're getting packets in-order
* (the reassembly queue is empty), add the data to
* the socket buffer and note that we need a delayed ack.
*
* XXX Some of these tests are not needed
* eg: the tiwin == tp->snd_wnd prevents many more
* predictions.. with no *real* advantage..
*/
if (tp->t_state == TCPS_ESTABLISHED &&
(tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK &&
ti->ti_seq == tp->rcv_nxt &&
tiwin && tiwin == tp->snd_wnd &&
tp->snd_nxt == tp->snd_max) {
if (ti->ti_len == 0) {
if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
tp->snd_cwnd >= tp->snd_wnd) {
/*
* this is a pure ack for outstanding data.
*/
if (tp->t_rtt &&
SEQ_GT(ti->ti_ack, tp->t_rtseq))
tcp_xmit_timer(tp, tp->t_rtt);
acked = ti->ti_ack - tp->snd_una;
sbdrop(&so->so_snd, acked);
tp->snd_una = ti->ti_ack;
m_free(m);
/*
* If all outstanding data are acked, stop
* retransmit timer, otherwise restart timer
* using current (possibly backed-off) value.
* If process is waiting for space,
* wakeup/selwakeup/signal. If data
* are ready to send, let tcp_output
* decide between more output or persist.
*/
if (tp->snd_una == tp->snd_max)
tp->t_timer[TCPT_REXMT] = 0;
else if (tp->t_timer[TCPT_PERSIST] == 0)
tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
/*
* This is called because sowwakeup might have
* put data into so_snd. Since we don't so sowwakeup,
* we don't need this.. XXX???
*/
if (so->so_snd.sb_cc)
(void) tcp_output(tp);
return;
}
} else if (ti->ti_ack == tp->snd_una &&
tcpfrag_list_empty(tp) &&
ti->ti_len <= sbspace(&so->so_rcv)) {
/*
* this is a pure, in-sequence data packet
* with nothing on the reassembly queue and
* we have enough buffer space to take it.
*/
tp->rcv_nxt += ti->ti_len;
/*
* Add data to socket buffer.
*/
if (so->so_emu) {
if (tcp_emu(so,m)) sbappend(so, m);
} else
sbappend(so, m);
/*
* If this is a short packet, then ACK now - with Nagel
* congestion avoidance sender won't send more until
* he gets an ACK.
*
* It is better to not delay acks at all to maximize
* TCP throughput. See RFC 2581.
*/
tp->t_flags |= TF_ACKNOW;
tcp_output(tp);
return;
}
} /* header prediction */
/*
* 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.
*/
{ int win;
win = sbspace(&so->so_rcv);
if (win < 0)
win = 0;
tp->rcv_wnd = max(win, (int)(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>
* Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
* Fill in remote peer address fields if not previously specified.
* Enter SYN_RECEIVED state, and process any other fields of this
* segment in this state.
*/
case TCPS_LISTEN: {
if (tiflags & TH_RST)
goto drop;
if (tiflags & TH_ACK)
goto dropwithreset;
if ((tiflags & TH_SYN) == 0)
goto drop;
/*
* This has way too many gotos...
* But a bit of spaghetti code never hurt anybody :)
*/
/*
* If this is destined for the control address, then flag to
* tcp_ctl once connected, otherwise connect
*/
if ((so->so_faddr.s_addr & slirp->vnetwork_mask.s_addr) ==
slirp->vnetwork_addr.s_addr) {
if (so->so_faddr.s_addr != slirp->vhost_addr.s_addr &&
so->so_faddr.s_addr != slirp->vnameserver_addr.s_addr) {
/* May be an add exec */
for (ex_ptr = slirp->exec_list; ex_ptr;
ex_ptr = ex_ptr->ex_next) {
if(ex_ptr->ex_fport == so->so_fport &&
so->so_faddr.s_addr == ex_ptr->ex_addr.s_addr) {
so->so_state |= SS_CTL;
break;
}
}
if (so->so_state & SS_CTL) {
goto cont_input;
}
}
/* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */
}
if (so->so_emu & EMU_NOCONNECT) {
so->so_emu &= ~EMU_NOCONNECT;
goto cont_input;
}
if ((tcp_fconnect(so) == -1) &&
#if defined(_WIN32)
socket_error() != WSAEWOULDBLOCK
#else
(errno != EINPROGRESS) && (errno != EWOULDBLOCK)
#endif
) {
u_char code=ICMP_UNREACH_NET;
DEBUG_MISC((dfd, " tcp fconnect errno = %d-%s\n",
errno,strerror(errno)));
if(errno == ECONNREFUSED) {
/* ACK the SYN, send RST to refuse the connection */
tcp_respond(tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
TH_RST|TH_ACK);
} else {
if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
HTONL(ti->ti_seq); /* restore tcp header */
HTONL(ti->ti_ack);
HTONS(ti->ti_win);
HTONS(ti->ti_urp);
m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
*ip=save_ip;
icmp_error(m, ICMP_UNREACH,code, 0,strerror(errno));
}
tcp_close(tp);
m_free(m);
} else {
/*
* Haven't connected yet, save the current mbuf
* and ti, and return
* XXX Some OS's don't tell us whether the connect()
* succeeded or not. So we must time it out.
*/
so->so_m = m;
so->so_ti = ti;
tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
tp->t_state = TCPS_SYN_RECEIVED;
tcp_template(tp);
}
return;
cont_conn:
/* m==NULL
* Check if the connect succeeded
*/
if (so->so_state & SS_NOFDREF) {
tp = tcp_close(tp);
goto dropwithreset;
}
cont_input:
tcp_template(tp);
if (optp)
tcp_dooptions(tp, (u_char *)optp, optlen, ti);
if (iss)
tp->iss = iss;
else
tp->iss = slirp->tcp_iss;
slirp->tcp_iss += TCP_ISSINCR/2;
tp->irs = ti->ti_seq;
tcp_sendseqinit(tp);
tcp_rcvseqinit(tp);
tp->t_flags |= TF_ACKNOW;
tp->t_state = TCPS_SYN_RECEIVED;
tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
goto trimthenstep6;
} /* case TCPS_LISTEN */
/*
* If the state is SYN_SENT:
* if seg contains an ACK, but not for our SYN, drop the input.
* if seg contains a RST, then drop the connection.
* if seg does not contain SYN, then drop it.
* Otherwise this is an acceptable SYN segment
* initialize tp->rcv_nxt and tp->irs
* if seg contains ack then advance tp->snd_una
* if SYN has been acked change to ESTABLISHED else SYN_RCVD state
* arrange for segment to be acked (eventually)
* continue processing rest of data/controls, beginning with URG
*/
case TCPS_SYN_SENT:
if ((tiflags & TH_ACK) &&
(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, 0); /* XXX Check t_softerror! */
}
goto drop;
}
if ((tiflags & TH_SYN) == 0)
goto drop;
if (tiflags & TH_ACK) {
tp->snd_una = ti->ti_ack;
if (SEQ_LT(tp->snd_nxt, tp->snd_una))
tp->snd_nxt = tp->snd_una;
}
tp->t_timer[TCPT_REXMT] = 0;
tp->irs = ti->ti_seq;
tcp_rcvseqinit(tp);
tp->t_flags |= TF_ACKNOW;
if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
soisfconnected(so);
tp->t_state = TCPS_ESTABLISHED;
(void) tcp_reass(tp, (struct tcpiphdr *)0,
(struct mbuf *)0);
/*
* if we didn't have to retransmit the SYN,
* use its rtt as our initial srtt & rtt var.
*/
if (tp->t_rtt)
tcp_xmit_timer(tp, tp->t_rtt);
} else
tp->t_state = TCPS_SYN_RECEIVED;
trimthenstep6:
/*
* Advance ti->ti_seq to correspond to first data byte.
* If data, trim to stay within window,
* dropping FIN if necessary.
*/
ti->ti_seq++;
if (ti->ti_len > tp->rcv_wnd) {
todrop = ti->ti_len - tp->rcv_wnd;
m_adj(m, -todrop);
ti->ti_len = tp->rcv_wnd;
tiflags &= ~TH_FIN;
}
tp->snd_wl1 = ti->ti_seq - 1;
tp->rcv_up = ti->ti_seq;
goto step6;
} /* switch tp->t_state */
/*
* States other than LISTEN or SYN_SENT.
* Check that at least some bytes of segment are within
* receive window. If segment begins before rcv_nxt,
* drop leading data (and SYN); if nothing left, just ack.
*/
todrop = tp->rcv_nxt - ti->ti_seq;
if (todrop > 0) {
if (tiflags & TH_SYN) {
tiflags &= ~TH_SYN;
ti->ti_seq++;
if (ti->ti_urp > 1)
ti->ti_urp--;
else
tiflags &= ~TH_URG;
todrop--;
}
/*
* Following if statement from Stevens, vol. 2, p. 960.
*/
if (todrop > ti->ti_len
|| (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
/*
* Any valid FIN must be to the left of the window.
* At this point the FIN must be a duplicate or out
* of sequence; drop it.
*/
tiflags &= ~TH_FIN;
/*
* Send an ACK to resynchronize and drop any data.
* But keep on processing for RST or ACK.
*/
tp->t_flags |= TF_ACKNOW;
todrop = ti->ti_len;
}
m_adj(m, todrop);
ti->ti_seq += todrop;
ti->ti_len -= todrop;
if (ti->ti_urp > todrop)
ti->ti_urp -= todrop;
else {
tiflags &= ~TH_URG;
ti->ti_urp = 0;
}
}
/*
* If new data are 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 segment ends after window, drop trailing data
* (and PUSH and FIN); if nothing left, just ACK.
*/
todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
if (todrop > 0) {
if (todrop >= ti->ti_len) {
/*
* If a new connection request is received
* while in TIME_WAIT, drop the old connection
* and start over if the sequence numbers
* are above the previous ones.
*/
if (tiflags & TH_SYN &&
tp->t_state == TCPS_TIME_WAIT &&
SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
iss = tp->rcv_nxt + TCP_ISSINCR;
tp = tcp_close(tp);
goto findso;
}
/*
* If window is closed can only take segments at
* window edge, and have to drop data and PUSH from
* incoming segments. Continue processing, but
* remember to ack. Otherwise, drop segment
* and ack.
*/
if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
tp->t_flags |= TF_ACKNOW;
} else {
goto dropafterack;
}
}
m_adj(m, -todrop);
ti->ti_len -= todrop;
tiflags &= ~(TH_PUSH|TH_FIN);
}
/*
* If the RST bit is set examine the state:
* SYN_RECEIVED STATE:
* If passive open, return to LISTEN state.
* If active open, inform user that connection was refused.
* ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
* Inform user that connection was reset, and close tcb.
* CLOSING, LAST_ACK, TIME_WAIT STATES
* Close the tcb.
*/
if (tiflags&TH_RST) switch (tp->t_state) {
case TCPS_SYN_RECEIVED:
case TCPS_ESTABLISHED:
case TCPS_FIN_WAIT_1:
case TCPS_FIN_WAIT_2:
case TCPS_CLOSE_WAIT:
tp->t_state = TCPS_CLOSED;
tcp_close(tp);
goto drop;
case TCPS_CLOSING:
case TCPS_LAST_ACK:
case TCPS_TIME_WAIT:
tcp_close(tp);
goto drop;
}
/*
* If a SYN is in the window, then this is an
* error and we send an RST and drop the connection.
*/
if (tiflags & TH_SYN) {
tp = tcp_drop(tp,0);
goto dropwithreset;
}
/*
* If the ACK bit is off we drop the segment and return.
*/
if ((tiflags & TH_ACK) == 0) goto drop;
/*
* Ack processing.
*/
switch (tp->t_state) {
/*
* In SYN_RECEIVED state if the ack ACKs our SYN then enter
* ESTABLISHED state and continue processing, otherwise
* send an RST. una<=ack<=max
*/
case TCPS_SYN_RECEIVED:
if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
SEQ_GT(ti->ti_ack, tp->snd_max))
goto dropwithreset;
tp->t_state = TCPS_ESTABLISHED;
/*
* The sent SYN is ack'ed with our sequence number +1
* The first data byte already in the buffer will get
* lost if no correction is made. This is only needed for
* SS_CTL since the buffer is empty otherwise.
* tp->snd_una++; or:
*/
tp->snd_una=ti->ti_ack;
if (so->so_state & SS_CTL) {
/* So tcp_ctl reports the right state */
ret = tcp_ctl(so);
if (ret == 1) {
soisfconnected(so);
so->so_state &= ~SS_CTL; /* success XXX */
} else if (ret == 2) {
so->so_state &= SS_PERSISTENT_MASK;
so->so_state |= SS_NOFDREF; /* CTL_CMD */
} else {
needoutput = 1;
tp->t_state = TCPS_FIN_WAIT_1;
}
} else {
soisfconnected(so);
}
(void) tcp_reass(tp, (struct tcpiphdr *)0, (struct mbuf *)0);
tp->snd_wl1 = ti->ti_seq - 1;
/* Avoid ack processing; snd_una==ti_ack => dup ack */
goto synrx_to_est;
/* fall into ... */
/*
* In ESTABLISHED state: drop duplicate ACKs; ACK out of range
* ACKs. If the ack is in the range
* tp->snd_una < ti->ti_ack <= tp->snd_max
* then advance tp->snd_una to ti->ti_ack and drop
* data from the retransmission queue. If this ACK reflects
* more up to date window information we update our window information.
*/
case TCPS_ESTABLISHED:
case TCPS_FIN_WAIT_1:
case TCPS_FIN_WAIT_2:
case TCPS_CLOSE_WAIT:
case TCPS_CLOSING:
case TCPS_LAST_ACK:
case TCPS_TIME_WAIT:
if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
DEBUG_MISC((dfd, " dup ack m = %lx so = %lx\n",
(long )m, (long )so));
/*
* If we have outstanding data (other than
* a window probe), this is a completely
* duplicate ack (ie, window info didn't
* change), the ack is the biggest we've
* seen and we've seen exactly our rexmt
* threshold of them, assume a packet
* has been dropped and retransmit it.
* Kludge snd_nxt & the congestion
* window so we send only this one
* packet.
*
* We know we're losing at the current
* window size so do congestion avoidance
* (set ssthresh to half the current window
* and pull our congestion window back to
* the new ssthresh).
*
* Dup acks mean that packets have left the
* network (they're now cached at the receiver)
* so bump cwnd by the amount in the receiver
* to keep a constant cwnd packets in the
* network.
*/
if (tp->t_timer[TCPT_REXMT] == 0 ||
ti->ti_ack != tp->snd_una)
tp->t_dupacks = 0;
else if (++tp->t_dupacks == TCPREXMTTHRESH) {
tcp_seq onxt = tp->snd_nxt;
u_int win =
min(tp->snd_wnd, tp->snd_cwnd) / 2 /
tp->t_maxseg;
if (win < 2)
win = 2;
tp->snd_ssthresh = win * tp->t_maxseg;
tp->t_timer[TCPT_REXMT] = 0;
tp->t_rtt = 0;
tp->snd_nxt = ti->ti_ack;
tp->snd_cwnd = tp->t_maxseg;
(void) tcp_output(tp);
tp->snd_cwnd = tp->snd_ssthresh +
tp->t_maxseg * tp->t_dupacks;
if (SEQ_GT(onxt, tp->snd_nxt))
tp->snd_nxt = onxt;
goto drop;
} else if (tp->t_dupacks > TCPREXMTTHRESH) {
tp->snd_cwnd += tp->t_maxseg;
(void) tcp_output(tp);
goto drop;
}
} else
tp->t_dupacks = 0;
break;
}
synrx_to_est:
/*
* If the congestion window was inflated to account
* for the other side's cached packets, retract it.
*/
if (tp->t_dupacks > TCPREXMTTHRESH &&
tp->snd_cwnd > tp->snd_ssthresh)
tp->snd_cwnd = tp->snd_ssthresh;
tp->t_dupacks = 0;
if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
goto dropafterack;
}
acked = ti->ti_ack - tp->snd_una;
/*
* If transmit timer is running and timed sequence
* number was acked, update smoothed round trip time.
* Since we now have an rtt measurement, cancel the
* timer backoff (cf., Phil Karn's retransmit alg.).
* Recompute the initial retransmit timer.
*/
if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
tcp_xmit_timer(tp,tp->t_rtt);
/*
* 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 (ti->ti_ack == tp->snd_max) {
tp->t_timer[TCPT_REXMT] = 0;
needoutput = 1;
} else if (tp->t_timer[TCPT_PERSIST] == 0)
tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
/*
* When new data is acked, open the congestion window.
* If the window gives us less than ssthresh packets
* in flight, open exponentially (maxseg per packet).
* Otherwise open linearly: maxseg per window
* (maxseg^2 / cwnd per packet).
*/
{
register u_int cw = tp->snd_cwnd;
register u_int incr = tp->t_maxseg;
if (cw > tp->snd_ssthresh)
incr = incr * incr / cw;
tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
}
if (acked > so->so_snd.sb_cc) {
tp->snd_wnd -= so->so_snd.sb_cc;
sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
ourfinisacked = 1;
} else {
sbdrop(&so->so_snd, acked);
tp->snd_wnd -= acked;
ourfinisacked = 0;
}
tp->snd_una = ti->ti_ack;
if (SEQ_LT(tp->snd_nxt, tp->snd_una))
tp->snd_nxt = tp->snd_una;
switch (tp->t_state) {
/*
* In FIN_WAIT_1 STATE in addition to the processing
* for the ESTABLISHED state if our FIN is now acknowledged
* then enter FIN_WAIT_2.
*/
case TCPS_FIN_WAIT_1:
if (ourfinisacked) {
/*
* 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_FCANTRCVMORE) {
tp->t_timer[TCPT_2MSL] = TCP_MAXIDLE;
}
tp->t_state = TCPS_FIN_WAIT_2;
}
break;
/*
* In CLOSING STATE in addition to the processing for
* the ESTABLISHED state if the ACK acknowledges our FIN
* then enter the TIME-WAIT state, otherwise ignore
* the segment.
*/
case TCPS_CLOSING:
if (ourfinisacked) {
tp->t_state = TCPS_TIME_WAIT;
tcp_canceltimers(tp);
tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
}
break;
/*
* In LAST_ACK, we may still be waiting for data to drain
* and/or to be acked, as well as for the ack of our FIN.
* If our FIN is now acknowledged, delete the TCB,
* enter the closed state and return.
*/
case TCPS_LAST_ACK:
if (ourfinisacked) {
tcp_close(tp);
goto drop;
}
break;
/*
* In TIME_WAIT state the only thing that should arrive
* is a retransmission of the remote FIN. Acknowledge
* it and restart the finack timer.
*/
case TCPS_TIME_WAIT:
tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
goto dropafterack;
}
} /* switch(tp->t_state) */
step6:
/*
* Update window information.
* Don't look at window if no ACK: TAC's send garbage on first SYN.
*/
if ((tiflags & TH_ACK) &&
(SEQ_LT(tp->snd_wl1, ti->ti_seq) ||
(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) ||
(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
tp->snd_wnd = tiwin;
tp->snd_wl1 = ti->ti_seq;
tp->snd_wl2 = ti->ti_ack;
if (tp->snd_wnd > tp->max_sndwnd)
tp->max_sndwnd = tp->snd_wnd;
needoutput = 1;
}
/*
* Process segments with URG.
*/
if ((tiflags & TH_URG) && ti->ti_urp &&
TCPS_HAVERCVDFIN(tp->t_state) == 0) {
/*
* This is a kludge, but if we receive and 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 + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
ti->ti_urp = 0;
tiflags &= ~TH_URG;
goto dodata;
}
/*
* 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(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
tp->rcv_up = ti->ti_seq + ti->ti_urp;
so->so_urgc = so->so_rcv.sb_cc +
(tp->rcv_up - tp->rcv_nxt); /* -1; */
tp->rcv_up = ti->ti_seq + ti->ti_urp;
}
} 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:
/*
* If this is a small packet, then ACK now - with Nagel
* congestion avoidance sender won't send more until
* he gets an ACK.
*/
if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
tp->t_flags |= TF_ACKNOW;
}
/*
* 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 ((ti->ti_len || (tiflags&TH_FIN)) &&
TCPS_HAVERCVDFIN(tp->t_state) == 0) {
TCP_REASS(tp, ti, m, so, tiflags);
} else {
m_free(m);
tiflags &= ~TH_FIN;
}
/*
* If FIN is received ACK the FIN and let the user know
* that the connection is closing.
*/
if (tiflags & TH_FIN) {
if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
/*
* If we receive a FIN we can't send more data,
* set it SS_FDRAIN
* Shutdown the socket if there is no rx data in the
* buffer.
* soread() is called on completion of shutdown() and
* will got to TCPS_LAST_ACK, and use tcp_output()
* to send the FIN.
*/
sofwdrain(so);
tp->t_flags |= TF_ACKNOW;
tp->rcv_nxt++;
}
switch (tp->t_state) {
/*
* In SYN_RECEIVED and ESTABLISHED STATES
* enter the CLOSE_WAIT state.
*/
case TCPS_SYN_RECEIVED:
case TCPS_ESTABLISHED:
if(so->so_emu == EMU_CTL) /* no shutdown on socket */
tp->t_state = TCPS_LAST_ACK;
else
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;
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;
}
}
/*
* 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_free(m);
tp->t_flags |= TF_ACKNOW;
(void) tcp_output(tp);
return;
dropwithreset:
/* reuses m if m!=NULL, m_free() unnecessary */
if (tiflags & TH_ACK)
tcp_respond(tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
else {
if (tiflags & TH_SYN) ti->ti_len++;
tcp_respond(tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
TH_RST|TH_ACK);
}
return;
drop:
/*
* Drop space held by incoming segment and return.
*/
m_free(m);
}
static int
uipc_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
struct mbuf *control, struct proc *p)
{
int error = 0;
struct unpcb *unp = sotounpcb(so);
struct socket *so2;
u_long newhiwat;
if (unp == 0) {
error = EINVAL;
goto release;
}
if (flags & PRUS_OOB) {
error = EOPNOTSUPP;
goto release;
}
if (control && (error = unp_internalize(control, p)))
goto release;
switch (so->so_type) {
case SOCK_DGRAM:
{
struct sockaddr *from;
if (nam) {
if (unp->unp_conn) {
error = EISCONN;
break;
}
error = unp_connect(so, nam, p);
if (error)
break;
} else {
if (unp->unp_conn == 0) {
error = ENOTCONN;
break;
}
}
so2 = unp->unp_conn->unp_socket;
if (unp->unp_addr)
from = (struct sockaddr *)unp->unp_addr;
else
from = &sun_noname;
if (sbappendaddr(&so2->so_rcv, from, m, control)) {
sorwakeup(so2);
m = 0;
control = 0;
} else
error = ENOBUFS;
if (nam)
unp_disconnect(unp);
break;
}
case SOCK_STREAM:
/* Connect if not connected yet. */
/*
* Note: A better implementation would complain
* if not equal to the peer's address.
*/
if ((so->so_state & SS_ISCONNECTED) == 0) {
if (nam) {
error = unp_connect(so, nam, p);
if (error)
break; /* XXX */
} else {
error = ENOTCONN;
break;
}
}
if (so->so_state & SS_CANTSENDMORE) {
error = EPIPE;
break;
}
if (unp->unp_conn == 0)
panic("uipc_send connected but no connection?");
so2 = unp->unp_conn->unp_socket;
/*
* Send to paired receive port, and then reduce
* send buffer hiwater marks to maintain backpressure.
* Wake up readers.
*/
if (control) {
if (sbappendcontrol(&so2->so_rcv, m, control))
control = 0;
} else
sbappend(&so2->so_rcv, m);
so->so_snd.sb_mbmax -=
so2->so_rcv.sb_mbcnt - unp->unp_conn->unp_mbcnt;
unp->unp_conn->unp_mbcnt = so2->so_rcv.sb_mbcnt;
newhiwat = so->so_snd.sb_hiwat -
(so2->so_rcv.sb_cc - unp->unp_conn->unp_cc);
(void)chgsbsize(so->so_cred->cr_uidinfo, &so->so_snd.sb_hiwat,
newhiwat, RLIM_INFINITY);
unp->unp_conn->unp_cc = so2->so_rcv.sb_cc;
sorwakeup(so2);
m = 0;
break;
default:
panic("uipc_send unknown socktype");
}
/*
* SEND_EOF is equivalent to a SEND followed by
* a SHUTDOWN.
*/
if (flags & PRUS_EOF) {
socantsendmore(so);
unp_shutdown(unp);
}
if (control && error != 0)
unp_dispose(control);
release:
if (control)
m_freem(control);
if (m)
m_freem(m);
return error;
}
/*
* Do a send by putting data in output queue and updating urgent
* marker if URG set. Possibly send more data. Unlike the other
* pru_*() routines, the mbuf chains are our responsibility. We
* must either enqueue them or free them. The other pru_* routines
* generally are caller-frees.
*/
static int
tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
struct sockaddr *nam, struct mbuf *control, struct proc *p)
{
int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
#ifdef INET6
int isipv6;
#endif
TCPDEBUG0;
if (inp == NULL) {
/*
* OOPS! we lost a race, the TCP session got reset after
* we checked SS_CANTSENDMORE, eg: while doing uiomove or a
* network interrupt in the non-splnet() section of sosend().
*/
if (m)
m_freem(m);
if (control)
m_freem(control);
error = ECONNRESET; /* XXX EPIPE? */
tp = NULL;
TCPDEBUG1();
goto out;
}
#ifdef INET6
isipv6 = nam && nam->sa_family == AF_INET6;
#endif /* INET6 */
tp = intotcpcb(inp);
TCPDEBUG1();
if (control) {
/* TCP doesn't do control messages (rights, creds, etc) */
if (control->m_len) {
m_freem(control);
if (m)
m_freem(m);
error = EINVAL;
goto out;
}
m_freem(control); /* empty control, just free it */
}
if(!(flags & PRUS_OOB)) {
sbappend(&so->so_snd, m);
if (nam && tp->t_state < TCPS_SYN_SENT) {
/*
* Do implied connect if not yet connected,
* initialize window to default value, and
* initialize maxseg/maxopd using peer's cached
* MSS.
*/
#ifdef INET6
if (isipv6)
error = tcp6_connect(tp, nam, p);
else
#endif /* INET6 */
error = tcp_connect(tp, nam, p);
if (error)
goto out;
tp->snd_wnd = TTCP_CLIENT_SND_WND;
tcp_mss(tp, -1);
}
if (flags & PRUS_EOF) {
/*
* Close the send side of the connection after
* the data is sent.
*/
socantsendmore(so);
tp = tcp_usrclosed(tp);
}
if (tp != NULL) {
if (flags & PRUS_MORETOCOME)
tp->t_flags |= TF_MORETOCOME;
error = tcp_output(tp);
if (flags & PRUS_MORETOCOME)
tp->t_flags &= ~TF_MORETOCOME;
}
} else {
if (sbspace(&so->so_snd) < -512) {
m_freem(m);
error = ENOBUFS;
goto out;
}
/*
* 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.
* Otherwise, snd_up should be one lower.
*/
sbappend(&so->so_snd, m);
if (nam && tp->t_state < TCPS_SYN_SENT) {
/*
* Do implied connect if not yet connected,
* initialize window to default value, and
* initialize maxseg/maxopd using peer's cached
* MSS.
*/
#ifdef INET6
if (isipv6)
error = tcp6_connect(tp, nam, p);
else
#endif /* INET6 */
error = tcp_connect(tp, nam, p);
if (error)
goto out;
tp->snd_wnd = TTCP_CLIENT_SND_WND;
tcp_mss(tp, -1);
}
tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
tp->t_force = 1;
error = tcp_output(tp);
tp->t_force = 0;
}
COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
}
/*
* TCP input routine, follows pages 65-76 of the
* protocol specification dated September, 1981 very closely.
*/
void
tcp_input(usn_mbuf_t *m, int iphlen)
{
struct tcpiphdr *ti;
struct inpcb *inp;
u_char *optp = NULL;
int optlen;
int len, tlen, off;
struct tcpcb *tp = 0;
int tiflags;
struct usn_socket *so = 0;
int todrop, acked, ourfinisacked;
int needoutput = 0;
short ostate;
struct usn_in_addr laddr;
int dropsocket = 0;
int iss = 0;
u_long tiwin, ts_val, ts_ecr;
int ts_present = 0;
(void)needoutput;
g_tcpstat.tcps_rcvtotal++;
// Get IP and TCP header together in first mbuf.
// Note: IP leaves IP header in first mbuf.
ti = mtod(m, struct tcpiphdr *);
if (iphlen > sizeof (usn_ip_t))
ip_stripoptions(m, (usn_mbuf_t *)0);
if (m->mlen < sizeof (struct tcpiphdr)) {
if ((m = m_pullup(m, sizeof (struct tcpiphdr))) == 0) {
g_tcpstat.tcps_rcvshort++;
return;
}
ti = mtod(m, struct tcpiphdr *);
}
#ifdef DUMP_PAYLOAD
dump_chain(m,"tcp");
#endif
/*
* Checksum extended TCP header and data.
*/
tlen = ntohs(((usn_ip_t *)ti)->ip_len);
len = sizeof (usn_ip_t) + tlen;
ti->ti_next = ti->ti_prev = 0;
ti->ti_x1 = 0;
ti->ti_len = (u_short)tlen;
HTONS(ti->ti_len);
ti->ti_sum = in_cksum(m, len);
if (ti->ti_sum) {
g_tcpstat.tcps_rcvbadsum++;
goto drop;
}
/*
* Check that TCP offset makes sense,
* pull out TCP options and adjust length. XXX
*/
off = ti->ti_off << 2;
if (off < sizeof (struct tcphdr) || off > tlen) {
g_tcpstat.tcps_rcvbadoff++;
goto drop;
}
tlen -= off;
ti->ti_len = tlen;
if (off > sizeof (struct tcphdr)) {
if (m->mlen < sizeof(usn_ip_t) + off) {
if ((m = m_pullup(m, sizeof (usn_ip_t) + off)) == 0) {
g_tcpstat.tcps_rcvshort++;
return;
}
ti = mtod(m, struct tcpiphdr *);
}
optlen = off - sizeof (struct tcphdr);
optp = mtod(m, u_char *) + sizeof (struct tcpiphdr);
// Do quick retrieval of timestamp options ("options
// prediction?"). If timestamp is the only option and it's
// formatted as recommended in RFC 1323 appendix A, we
// quickly get the values now and not bother calling
// tcp_dooptions(), etc.
if ((optlen == TCPOLEN_TSTAMP_APPA ||
(optlen > TCPOLEN_TSTAMP_APPA &&
optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
*(u_int *)optp == htonl(TCPOPT_TSTAMP_HDR) &&
(ti->ti_flags & TH_SYN) == 0) {
ts_present = 1;
ts_val = ntohl(*(u_long *)(optp + 4));
ts_ecr = ntohl(*(u_long *)(optp + 8));
optp = NULL; // we've parsed the options
}
}
tiflags = ti->ti_flags;
// Convert TCP protocol specific fields to host format.
NTOHL(ti->ti_seq);
NTOHL(ti->ti_ack);
NTOHS(ti->ti_win);
NTOHS(ti->ti_urp);
// Locate pcb for segment.
findpcb:
inp = g_tcp_last_inpcb;
if (inp->inp_lport != ti->ti_dport ||
inp->inp_fport != ti->ti_sport ||
inp->inp_faddr.s_addr != ti->ti_src.s_addr ||
inp->inp_laddr.s_addr != ti->ti_dst.s_addr) {
inp = in_pcblookup(&g_tcb, ti->ti_src, ti->ti_sport,
ti->ti_dst, ti->ti_dport, INPLOOKUP_WILDCARD);
if (inp)
g_tcp_last_inpcb = inp;
++g_tcpstat.tcps_pcbcachemiss;
}
// If the state is CLOSED (i.e., TCB does not exist) then
// all data in the incoming segment is discarded.
// If the TCB exists but is in CLOSED state, it is embryonic,
// but should either do a listen or a connect soon.
if (inp == 0)
goto dropwithreset;
tp = intotcpcb(inp);
DEBUG("found inp cb, laddr=%x, lport=%d, faddr=%x,"
" fport=%d, tp_state=%d, tp_flags=%d",
inp->inp_laddr.s_addr,
inp->inp_lport,
inp->inp_faddr.s_addr,
inp->inp_fport, tp->t_state, tp->t_flags);
if (tp == 0)
goto dropwithreset;
if (tp->t_state == TCPS_CLOSED)
goto drop;
// Unscale the window into a 32-bit value.
if ((tiflags & TH_SYN) == 0)
tiwin = ti->ti_win << tp->snd_scale;
else
tiwin = ti->ti_win;
so = inp->inp_socket;
DEBUG("socket info, options=%x", so->so_options);
if (so->so_options & (SO_DEBUG|SO_ACCEPTCONN)) {
if (so->so_options & SO_DEBUG) {
ostate = tp->t_state;
g_tcp_saveti = *ti;
}
if (so->so_options & SO_ACCEPTCONN) {
if ((tiflags & (TH_RST|TH_ACK|TH_SYN)) != TH_SYN) {
// Note: dropwithreset makes sure we don't
// send a reset in response to a RST.
if (tiflags & TH_ACK) {
g_tcpstat.tcps_badsyn++;
goto dropwithreset;
}
DEBUG("SYN is expected, tiflags=%d", tiflags);
goto drop;
}
so = sonewconn(so, 0);
if (so == 0) {
DEBUG("failed to create new connection, tiflags=%d", tiflags);
goto drop;
}
// 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;
// BSD >= 4.3
inp->inp_options = ip_srcroute();
tp = intotcpcb(inp);
tp->t_state = TCPS_LISTEN;
// Compute proper scaling value from buffer space
while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
TCP_MAXWIN << tp->request_r_scale < so->so_rcv->sb_hiwat)
tp->request_r_scale++;
}
}
// Segment received on connection.
// Reset idle time and keep-alive timer.
tp->t_idle = 0;
tp->t_timer[TCPT_KEEP] = g_tcp_keepidle;
// Process options if not in LISTEN state,
// else do it below (after getting remote address).
if (optp && tp->t_state != TCPS_LISTEN)
tcp_dooptions(tp, optp, optlen, ti,
&ts_present, &ts_val, &ts_ecr);
// Header prediction: check for the two common cases
// of a uni-directional data xfer. If the packet has
// no control flags, is in-sequence, the window didn't
// change and we're not retransmitting, it's a
// candidate. If the length is zero and the ack moved
// forward, we're the sender side of the xfer. Just
// free the data acked & wake any higher level process
// that was blocked waiting for space. If the length
// is non-zero and the ack didn't move, we're the
// receiver side. If we're getting packets in-order
// (the reassembly queue is empty), add the data to
// the socket buffer and note that we need a delayed ack.
if (tp->t_state == TCPS_ESTABLISHED &&
(tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK &&
(!ts_present || TSTMP_GEQ(ts_val, tp->ts_recent)) &&
ti->ti_seq == tp->rcv_nxt &&
tiwin && tiwin == tp->snd_wnd &&
tp->snd_nxt == tp->snd_max) {
// If last ACK falls within this segment's sequence numbers,
// record the timestamp.
if ( ts_present && TSTMP_GEQ(ts_val, tp->ts_recent) &&
SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) ){
tp->ts_recent_age = g_tcp_now;
tp->ts_recent = ts_val;
}
if (ti->ti_len == 0) {
if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
tp->snd_cwnd >= tp->snd_wnd) {
// this is a pure ack for outstanding data.
++g_tcpstat.tcps_predack;
if (ts_present)
tcp_xmit_timer(tp, g_tcp_now-ts_ecr+1);
else if (tp->t_rtt &&
SEQ_GT(ti->ti_ack, tp->t_rtseq))
tcp_xmit_timer(tp, tp->t_rtt);
acked = ti->ti_ack - tp->snd_una;
g_tcpstat.tcps_rcvackpack++;
g_tcpstat.tcps_rcvackbyte += acked;
TRACE("drop so_snd buffer, drop_bytes=%d, len=%d",
acked, so->so_snd.sb_cc);
sbdrop(so->so_snd, acked);
tp->snd_una = ti->ti_ack;
usn_free_cmbuf(m);
// If all outstanding data are acked, stop
// retransmit timer, otherwise restart timer
// using current (possibly backed-off) value.
// If process is waiting for space,
// wakeup/selwakeup/signal. If data
// are ready to send, let tcp_output
// decide between more output or persist.
if (tp->snd_una == tp->snd_max)
tp->t_timer[TCPT_REXMT] = 0;
else if (tp->t_timer[TCPT_PERSIST] == 0)
tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
if (so->so_options & SO_DEBUG)
tcp_trace(TA_INPUT, ostate, tp, &g_tcp_saveti, 0);
//if (so->so_snd->sb_flags & SB_NOTIFY) {
// usnet_tcpin_wwakeup(so, USN_TCP_IN, usn_tcpev_sbnotify, 0);
// sowwakeup(so);
//}
// send buffer is available for app thread.
usnet_tcpin_wwakeup(so, USN_TCP_IN, USN_TCPEV_WRITE, 0);
if (so->so_snd->sb_cc)
tcp_output(tp);
return;
}
} else if (ti->ti_ack == tp->snd_una &&
tp->seg_next == (struct tcpiphdr *)tp &&
ti->ti_len <= sbspace(so->so_rcv)) {
// this is a pure, in-sequence data packet
// with nothing on the reassembly queue and
// we have enough buffer space to take it.
++g_tcpstat.tcps_preddat;
tp->rcv_nxt += ti->ti_len;
g_tcpstat.tcps_rcvpack++;
g_tcpstat.tcps_rcvbyte += ti->ti_len;
// Drop TCP, IP headers and TCP options then add data
// to socket buffer.
m->head += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
m->mlen -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
TRACE("add data to rcv buf");
sbappend(so->so_rcv, m);
sorwakeup(so);
// new data is available for app threads.
usnet_tcpin_rwakeup(so, USN_TCP_IN, USN_TCPEV_READ, m);
if (so->so_options & SO_DEBUG) {
TRACE("tcp trace, so_options=%d", so->so_options);
tcp_trace(TA_INPUT, ostate, tp, &g_tcp_saveti, 0);
}
tp->t_flags |= TF_DELACK;
return;
}
}
// Drop TCP, IP headers and TCP options.
m->head += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
m->mlen -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
// 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.
{
int win;
win = sbspace(so->so_rcv);
if (win < 0)
win = 0;
tp->rcv_wnd = max(win, (int)(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>
// Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
// Fill in remote peer address fields if not previously specified.
// Enter SYN_RECEIVED state, and process any other fields of this
// segment in this state.
case TCPS_LISTEN: {
usn_mbuf_t *am;
struct usn_sockaddr_in *sin;
if (tiflags & TH_RST)
goto drop;
if (tiflags & TH_ACK)
goto dropwithreset;
if ((tiflags & TH_SYN) == 0)
goto drop;
// RFC1122 4.2.3.10, p. 104: discard bcast/mcast SYN
// in_broadcast() should never return true on a received
// packet with M_BCAST not set.
//if (m->m_flags & (M_BCAST|M_MCAST) ||
// IN_MULTICAST(ntohl(ti->ti_dst.s_addr)))
// goto drop;
am = usn_get_mbuf(0, BUF_MSIZE, 0); // XXX: the size!
if (am == NULL)
goto drop;
am->mlen = sizeof (struct usn_sockaddr_in);
sin = mtod(am, struct usn_sockaddr_in *);
sin->sin_family = AF_INET;
sin->sin_len = sizeof(*sin);
sin->sin_addr = ti->ti_src;
sin->sin_port = ti->ti_sport;
bzero((caddr_t)sin->sin_zero, sizeof(sin->sin_zero));
laddr = inp->inp_laddr;
if (inp->inp_laddr.s_addr == USN_INADDR_ANY)
inp->inp_laddr = ti->ti_dst;
if (in_pcbconnect(inp, am)) {
inp->inp_laddr = laddr;
usn_free_mbuf(am);
goto drop;
}
usn_free_mbuf(am);
tp->t_template = tcp_template(tp);
if (tp->t_template == 0) {
tp = tcp_drop(tp, ENOBUFS);
dropsocket = 0; // socket is already gone
goto drop;
}
if (optp)
tcp_dooptions(tp, optp, optlen, ti,
&ts_present, &ts_val, &ts_ecr);
if (iss)
tp->iss = iss;
else
tp->iss = g_tcp_iss;
g_tcp_iss += TCP_ISSINCR/4;
tp->irs = ti->ti_seq;
tcp_sendseqinit(tp);
tcp_rcvseqinit(tp);
tp->t_flags |= TF_ACKNOW;
TRACE("change tcp state to TCPS_SYN_RECEIVED, state=%d, tp_flags=%d",
tp->t_state, tp->t_flags);
tp->t_state = TCPS_SYN_RECEIVED;
// tcp event
usnet_tcpin_ewakeup(so, USN_TCP_IN, USN_TCPST_SYN_RECEIVED, 0);
tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
dropsocket = 0; // committed to socket
g_tcpstat.tcps_accepts++;
goto trimthenstep6;
}
// If the state is SYN_SENT:
// if seg contains an ACK, but not for our SYN, drop the input.
// if seg contains a RST, then drop the connection.
// if seg does not contain SYN, then drop it.
// Otherwise this is an acceptable SYN segment
// initialize tp->rcv_nxt and tp->irs
// if seg contains ack then advance tp->snd_una
// if SYN has been acked change to ESTABLISHED else SYN_RCVD state
// arrange for segment to be acked (eventually)
// continue processing rest of data/controls, beginning with URG
case TCPS_SYN_SENT:
if ((tiflags & TH_ACK) &&
(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)
tp = tcp_drop(tp, ECONNREFUSED);
goto drop;
}
if ((tiflags & TH_SYN) == 0)
goto drop;
if (tiflags & TH_ACK) {
tp->snd_una = ti->ti_ack;
if (SEQ_LT(tp->snd_nxt, tp->snd_una))
tp->snd_nxt = tp->snd_una;
tp->t_timer[TCPT_REXMT] = 0;
}
tp->irs = ti->ti_seq;
tcp_rcvseqinit(tp);
tp->t_flags |= TF_ACKNOW;
TRACE("ack now, tp flags=%d", tp->t_flags);
// XXX: remove second test.
if (tiflags & TH_ACK /*&& SEQ_GT(tp->snd_una, tp->iss)*/) {
g_tcpstat.tcps_connects++;
soisconnected(so);
TRACE("change tcp state to TCPS_ESTABLISHED,"
" state=%d, tp_flags=%d", tp->t_state, tp->t_flags);
tp->t_state = TCPS_ESTABLISHED;
// Do window scaling on this connection?
if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
(TF_RCVD_SCALE|TF_REQ_SCALE)) {
tp->snd_scale = tp->requested_s_scale;
tp->rcv_scale = tp->request_r_scale;
}
tcp_reass(tp, (struct tcpiphdr *)0, (usn_mbuf_t *)0);
// if we didn't have to retransmit the SYN,
// use its rtt as our initial srtt & rtt var.
if (tp->t_rtt)
tcp_xmit_timer(tp, tp->t_rtt);
} else {
TRACE("change tcp state to TCPS_SYN_RECEIVED, state=%d, tp_flags=%d",
tp->t_state, tp->t_flags);
tp->t_state = TCPS_SYN_RECEIVED;
// tcp event
usnet_tcpin_ewakeup(so, USN_TCP_IN, USN_TCPST_SYN_RECEIVED, 0);
}
trimthenstep6:
// Advance ti->ti_seq to correspond to first data byte.
// If data, trim to stay within window,
// dropping FIN if necessary.
ti->ti_seq++;
if (ti->ti_len > tp->rcv_wnd) {
todrop = ti->ti_len - tp->rcv_wnd;
m_adj(m, -todrop);
ti->ti_len = tp->rcv_wnd;
tiflags &= ~TH_FIN;
g_tcpstat.tcps_rcvpackafterwin++;
g_tcpstat.tcps_rcvbyteafterwin += todrop;
}
tp->snd_wl1 = ti->ti_seq - 1;
tp->rcv_up = ti->ti_seq;
goto step6;
}
// States other than LISTEN or SYN_SENT.
// First check timestamp, if present.
// Then check that at least some bytes of segment are within
// receive window. If segment begins before rcv_nxt,
// drop leading data (and SYN); if nothing left, just ack.
//
// RFC 1323 PAWS: If we have a timestamp reply on this segment
// and it's less than ts_recent, drop it.
if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
TSTMP_LT(ts_val, tp->ts_recent)) {
// Check to see if ts_recent is over 24 days old.
if ((int)(g_tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
// Invalidate ts_recent. If this segment updates
// ts_recent, the age will be reset later and ts_recent
// will get a valid value. If it does not, setting
// ts_recent to zero will at least satisfy the
// requirement that zero be placed in the timestamp
// echo reply when ts_recent isn't valid. The
// age isn't reset until we get a valid ts_recent
// because we don't want out-of-order segments to be
// dropped when ts_recent is old.
tp->ts_recent = 0;
} else {
g_tcpstat.tcps_rcvduppack++;
g_tcpstat.tcps_rcvdupbyte += ti->ti_len;
g_tcpstat.tcps_pawsdrop++;
goto dropafterack;
}
}
todrop = tp->rcv_nxt - ti->ti_seq;
if (todrop > 0) {
if (tiflags & TH_SYN) {
tiflags &= ~TH_SYN;
ti->ti_seq++;
if (ti->ti_urp > 1)
ti->ti_urp--;
else
tiflags &= ~TH_URG;
todrop--;
}
if ( todrop >= ti->ti_len ||
( todrop == ti->ti_len && (tiflags & TH_FIN ) == 0 ) ) {
// Any valid FIN must be to the left of the window.
// At this point the FIN must be a duplicate or
// out of sequence; drop it.
tiflags &= ~TH_FIN;
// Send an ACK to resynchronize and drop any data
// But keep on processing for RST or ACK.
tp->t_flags |= TF_ACKNOW;
TRACE("send ack now to resync, tp_flags=%d", tp->t_flags);
todrop = ti->ti_len;
g_tcpstat.tcps_rcvdupbyte += ti->ti_len;
g_tcpstat.tcps_rcvduppack++;
} else {
g_tcpstat.tcps_rcvpartduppack++;
g_tcpstat.tcps_rcvpartdupbyte += ti->ti_len;
}
m_adj(m, todrop);
ti->ti_seq += todrop;
ti->ti_len -= todrop;
if (ti->ti_urp > todrop)
ti->ti_urp -= todrop;
else {
tiflags &= ~TH_URG;
ti->ti_urp = 0;
}
}
// If new data are received on a connection after the
// user processes are gone, then RST the other end.
if ((so->so_state & USN_NOFDREF) &&
tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
tp = tcp_close(tp);
g_tcpstat.tcps_rcvafterclose++;
goto dropwithreset;
}
// If segment ends after window, drop trailing data
// (and PUSH and FIN); if nothing left, just ACK.
todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
if (todrop > 0) {
g_tcpstat.tcps_rcvpackafterwin++;
if (todrop >= ti->ti_len) {
g_tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
// If a new connection request is received
// while in TIME_WAIT, drop the old connection
// and start over if the sequence numbers
// are above the previous ones.
if (tiflags & TH_SYN &&
tp->t_state == TCPS_TIME_WAIT &&
SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
iss = tp->snd_nxt + TCP_ISSINCR;
tp = tcp_close(tp);
goto findpcb;
}
// If window is closed can only take segments at
// window edge, and have to drop data and PUSH from
// incoming segments. Continue processing, but
// remember to ack. Otherwise, drop segment
// and ack.
if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
tp->t_flags |= TF_ACKNOW;
g_tcpstat.tcps_rcvwinprobe++;
} else
goto dropafterack;
} else
g_tcpstat.tcps_rcvbyteafterwin += todrop;
m_adj(m, -todrop);
ti->ti_len -= todrop;
tiflags &= ~(TH_PUSH|TH_FIN);
}
// check valid timestamp. Replace code above.
if (ts_present && TSTMP_GEQ(ts_val, tp->ts_recent) &&
SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) ) {
tp->ts_recent_age = g_tcp_now;
tp->ts_recent = ts_val;
}
// If the RST bit is set examine the state:
// SYN_RECEIVED STATE:
// If passive open, return to LISTEN state.
// If active open, inform user that connection was refused.
// ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
// Inform user that connection was reset, and close tcb.
// CLOSING, LAST_ACK, TIME_WAIT STATES
// Close the tcb.
if (tiflags&TH_RST) switch (tp->t_state) {
case TCPS_SYN_RECEIVED:
so->so_error = ECONNREFUSED;
goto close;
case TCPS_ESTABLISHED:
case TCPS_FIN_WAIT_1:
case TCPS_FIN_WAIT_2:
case TCPS_CLOSE_WAIT:
so->so_error = ECONNRESET;
close:
DEBUG("change tcp state to TCPS_CLOSED, state=%d", tp->t_state);
tp->t_state = TCPS_CLOSED;
// tcp event
usnet_tcpin_ewakeup(so, USN_TCP_IN, USN_TCPST_CLOSED, 0);
g_tcpstat.tcps_drops++;
tp = tcp_close(tp);
goto drop;
case TCPS_CLOSING:
case TCPS_LAST_ACK:
case TCPS_TIME_WAIT:
tp = tcp_close(tp);
goto drop;
}
// If a SYN is in the window, then this is an
// error and we send an RST and drop the connection.
if (tiflags & TH_SYN) {
tp = tcp_drop(tp, ECONNRESET);
goto dropwithreset;
}
// If the ACK bit is off we drop the segment and return.
if ((tiflags & TH_ACK) == 0)
goto drop;
// Ack processing.
switch (tp->t_state) {
// In SYN_RECEIVED state if the ack ACKs our SYN then enter
// ESTABLISHED state and continue processing, otherwise
// send an RST.
case TCPS_SYN_RECEIVED:
if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
SEQ_GT(ti->ti_ack, tp->snd_max))
goto dropwithreset;
g_tcpstat.tcps_connects++;
DEBUG("change tcp state to TCPS_ESTABLISHED, state=%d", tp->t_state);
tp->t_state = TCPS_ESTABLISHED;
soisconnected(so);
// Do window scaling?
if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
(TF_RCVD_SCALE|TF_REQ_SCALE)) {
tp->snd_scale = tp->requested_s_scale;
tp->rcv_scale = tp->request_r_scale;
}
tcp_reass(tp, (struct tcpiphdr *)0, (usn_mbuf_t *)0);
tp->snd_wl1 = ti->ti_seq - 1;
// fall into ...
// In ESTABLISHED state: drop duplicate ACKs; ACK out of range
// ACKs. If the ack is in the range
// tp->snd_una < ti->ti_ack <= tp->snd_max
// then advance tp->snd_una to ti->ti_ack and drop
// data from the retransmission queue. If this ACK reflects
// more up to date window information we update our window information.
case TCPS_ESTABLISHED:
case TCPS_FIN_WAIT_1:
case TCPS_FIN_WAIT_2:
case TCPS_CLOSE_WAIT:
case TCPS_CLOSING:
case TCPS_LAST_ACK:
case TCPS_TIME_WAIT:
if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
g_tcpstat.tcps_rcvdupack++;
// If we have outstanding data (other than
// a window probe), this is a completely
// duplicate ack (ie, window info didn't
// change), the ack is the biggest we've
// seen and we've seen exactly our rexmt
// threshhold of them, assume a packet
// has been dropped and retransmit it.
// Kludge snd_nxt & the congestion
// window so we send only this one
// packet.
//
// We know we're losing at the current
// window size so do congestion avoidance
// (set ssthresh to half the current window
// and pull our congestion window back to
// the new ssthresh).
//
// Dup acks mean that packets have left the
// network (they're now cached at the receiver)
// so bump cwnd by the amount in the receiver
// to keep a constant cwnd packets in the
// network.
if (tp->t_timer[TCPT_REXMT] == 0 ||
ti->ti_ack != tp->snd_una)
tp->t_dupacks = 0;
else if (++tp->t_dupacks == g_tcprexmtthresh) {
// congestion avoidance
tcp_seq onxt = tp->snd_nxt;
u_int win =
min(tp->snd_wnd, tp->snd_cwnd) / 2 / tp->t_maxseg;
if (win < 2)
win = 2;
tp->snd_ssthresh = win * tp->t_maxseg;
tp->t_timer[TCPT_REXMT] = 0;
tp->t_rtt = 0;
tp->snd_nxt = ti->ti_ack;
tp->snd_cwnd = tp->t_maxseg;
tcp_output(tp);
tp->snd_cwnd = tp->snd_ssthresh +
tp->t_maxseg * tp->t_dupacks;
if (SEQ_GT(onxt, tp->snd_nxt))
tp->snd_nxt = onxt;
goto drop;
} else if (tp->t_dupacks > g_tcprexmtthresh) {
tp->snd_cwnd += tp->t_maxseg;
tcp_output(tp);
goto drop;
}
} else
tp->t_dupacks = 0;
break;
}
// If the congestion window was inflated to account
// for the other side's cached packets, retract it.
if (tp->t_dupacks > g_tcprexmtthresh &&
tp->snd_cwnd > tp->snd_ssthresh)
tp->snd_cwnd = tp->snd_ssthresh;
tp->t_dupacks = 0;
if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
g_tcpstat.tcps_rcvacktoomuch++;
goto dropafterack;
}
acked = ti->ti_ack - tp->snd_una;
g_tcpstat.tcps_rcvackpack++;
g_tcpstat.tcps_rcvackbyte += acked;
// If we have a timestamp reply, update smoothed
// round trip time. If no timestamp is present but
// transmit timer is running and timed sequence
// number was acked, update smoothed round trip time.
// Since we now have an rtt measurement, cancel the
// timer backoff (cf., Phil Karn's retransmit alg.).
// Recompute the initial retransmit timer.
if (ts_present)
tcp_xmit_timer(tp, g_tcp_now-ts_ecr+1);
else if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
tcp_xmit_timer(tp,tp->t_rtt);
// 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 (ti->ti_ack == tp->snd_max) {
tp->t_timer[TCPT_REXMT] = 0;
DEBUG("change needoutput to 1");
needoutput = 1;
tp->t_flags |= TF_NEEDOUTPUT;
} else if (tp->t_timer[TCPT_PERSIST] == 0)
tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
// When new data is acked, open the congestion window.
// If the window gives us less than ssthresh packets
// in flight, open exponentially (maxseg per packet).
// Otherwise open linearly: maxseg per window
// (maxseg * (maxseg / cwnd) per packet).
{
u_int cw = tp->snd_cwnd;
u_int incr = tp->t_maxseg;
if (cw > tp->snd_ssthresh)
incr = incr * incr / cw;
tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
}
if (acked > so->so_snd->sb_cc) {
tp->snd_wnd -= so->so_snd->sb_cc;
DEBUG("drop all so_snd buffer, drop_bytes=%d, acked=%d",
so->so_snd->sb_cc, acked);
sbdrop(so->so_snd, (int)so->so_snd->sb_cc);
ourfinisacked = 1;
} else {
DEBUG("drop so_snd buffer, drop_bytes=%d, len=%d", acked, so->so_snd->sb_cc);
sbdrop(so->so_snd, acked);
tp->snd_wnd -= acked;
ourfinisacked = 0;
}
//if (so->so_snd->sb_flags & SB_NOTIFY) {
sowwakeup(so);
usnet_tcpin_wwakeup(so, USN_TCP_IN, USN_TCPEV_WRITE, 0);
//}
tp->snd_una = ti->ti_ack;
if (SEQ_LT(tp->snd_nxt, tp->snd_una))
tp->snd_nxt = tp->snd_una;
switch (tp->t_state) {
// In FIN_WAIT_1 STATE in addition to the processing
// for the ESTABLISHED state if our FIN is now acknowledged
// then enter FIN_WAIT_2.
case TCPS_FIN_WAIT_1:
if (ourfinisacked) {
// 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 & USN_CANTRCVMORE) {
soisdisconnected(so);
tp->t_timer[TCPT_2MSL] = g_tcp_maxidle;
}
DEBUG("change tcp state to TCPS_FIN_WAIT_2, state=%d", tp->t_state);
tp->t_state = TCPS_FIN_WAIT_2;
usnet_tcpin_ewakeup(so, USN_TCP_IN, USN_TCPST_FIN_WAIT2, 0);
}
break;
// In CLOSING STATE in addition to the processing for
// the ESTABLISHED state if the ACK acknowledges our FIN
// then enter the TIME-WAIT state, otherwise ignore
// the segment.
case TCPS_CLOSING:
if (ourfinisacked) {
DEBUG("change tcp state to TCPS_TIME_WAIT, state=%d", tp->t_state);
tp->t_state = TCPS_TIME_WAIT;
usnet_tcpin_ewakeup(so, USN_TCP_IN, USN_TCPST_TIME_WAIT, 0);
tcp_canceltimers(tp);
tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
soisdisconnected(so);
}
break;
// In LAST_ACK, we may still be waiting for data to drain
// and/or to be acked, as well as for the ack of our FIN.
// If our FIN is now acknowledged, delete the TCB,
// enter the closed state and return.
case TCPS_LAST_ACK:
if (ourfinisacked) {
tp = tcp_close(tp);
goto drop;
}
break;
// In TIME_WAIT state the only thing that should arrive
// is a retransmission of the remote FIN. Acknowledge
// it and restart the finack timer.
case TCPS_TIME_WAIT:
tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
goto dropafterack;
}
}
step6:
// Update window information.
// Don't look at window if no ACK: TAC's send garbage on first SYN.
if ((tiflags & TH_ACK) &&
(SEQ_LT(tp->snd_wl1, ti->ti_seq) ||
(tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) ||
(tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd) )) )) {
// keep track of pure window updates
if (ti->ti_len == 0 &&
tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd)
g_tcpstat.tcps_rcvwinupd++;
tp->snd_wnd = tiwin;
tp->snd_wl1 = ti->ti_seq;
tp->snd_wl2 = ti->ti_ack;
if (tp->snd_wnd > tp->max_sndwnd)
tp->max_sndwnd = tp->snd_wnd;
DEBUG("change needoutput to 1");
tp->t_flags |= TF_NEEDOUTPUT;
needoutput = 1;
}
// Process segments with URG.
if ((tiflags & TH_URG) && ti->ti_urp &&
TCPS_HAVERCVDFIN(tp->t_state) == 0) {
// This is a kludge, but if we receive and 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 + so->so_rcv->sb_cc > g_sb_max) {
ti->ti_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(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
tp->rcv_up = ti->ti_seq + ti->ti_urp;
so->so_oobmark = so->so_rcv->sb_cc +
(tp->rcv_up - tp->rcv_nxt) - 1;
if (so->so_oobmark == 0)
so->so_state |= USN_RCVATMARK;
sohasoutofband(so);
// send async event to app threads.
usnet_tcpin_ewakeup(so, USN_TCP_IN, USN_TCPEV_OUTOFBOUND, 0);
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 (ti->ti_urp <= ti->ti_len
#ifdef SO_OOBINLINE
&& (so->so_options & SO_OOBINLINE) == 0
#endif
)
tcp_pulloutofband(so, ti, m);
} 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
#ifdef DUMP_PAYLOAD
DEBUG("Handle data");
dump_chain(m,"tcp");
#endif
// 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 ((ti->ti_len || (tiflags&TH_FIN)) &&
TCPS_HAVERCVDFIN(tp->t_state) == 0) {
TCP_REASS(tp, ti, m, so, tiflags);
// 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);
} else {
usn_free_cmbuf(m);
tiflags &= ~TH_FIN;
}
// If FIN is received ACK the FIN and let the user know
// that the connection is closing.
if (tiflags & TH_FIN) {
if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
socantrcvmore(so);
tp->t_flags |= TF_ACKNOW;
TRACE("ack FIN now, tp flags=%d", tp->t_flags);
tp->rcv_nxt++;
}
switch (tp->t_state) {
// In SYN_RECEIVED and ESTABLISHED STATES
// enter the CLOSE_WAIT state.
case TCPS_SYN_RECEIVED:
case TCPS_ESTABLISHED:
TRACE("change tcp state to TCPS_CLOSE_WAIT, state=%d", tp->t_state);
tp->t_state = TCPS_CLOSE_WAIT;
soewakeup(so, 0);
usnet_tcpin_ewakeup(so, USN_TCP_IN, USN_TCPST_CLOSE_WAIT, 0);
break;
// If still in FIN_WAIT_1 STATE FIN has not been acked so
// enter the CLOSING state.
case TCPS_FIN_WAIT_1:
TRACE("change tcp state to TCPS_CLOSING, state=%d", tp->t_state);
tp->t_state = TCPS_CLOSING;
usnet_tcpin_ewakeup(so, USN_TCP_IN, USN_TCPST_CLOSING, 0);
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:
TRACE("change tcp state to TCPS_TIME_WAIT, state=%d", tp->t_state);
tp->t_state = TCPS_TIME_WAIT;
tcp_canceltimers(tp);
tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
soisdisconnected(so);
usnet_tcpin_ewakeup(so, USN_TCP_IN, USN_TCPST_TIME_WAIT, 0);
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;
}
}
if (so->so_options & SO_DEBUG) {
TRACE("tcp trace, so_options=%d", so->so_options);
tcp_trace(TA_INPUT, ostate, tp, &g_tcp_saveti, 0);
}
// Return any desired output.
//if (needoutput || (tp->t_flags & TF_ACKNOW)){
if (tp->t_flags & TF_NEEDOUTPUT || (tp->t_flags & TF_ACKNOW)){
TRACE("ack now or need to ouput, tp->t_flags=%d", tp->t_flags);
tcp_output(tp);
}
return;
dropafterack:
TRACE("dropafterack");
// Generate an ACK dropping incoming segment if it occupies
// sequence space, where the ACK reflects our state.
if (tiflags & TH_RST)
goto drop;
usn_free_cmbuf(m);
tp->t_flags |= TF_ACKNOW;
TRACE("ack now, tp flags=%d", tp->t_flags);
tcp_output(tp);
return;
dropwithreset:
TRACE("dropwithreset");
// Generate a RST, dropping incoming segment.
// Make ACK acceptable to originator of segment.
// Don't bother to respond if destination was broadcast/multicast.
#define USN_MULTICAST(i) (((u_int)(i) & 0xf0000000) == 0xe0000000)
if ((tiflags & TH_RST) || m->flags & (BUF_BCAST|BUF_MCAST) ||
USN_MULTICAST(ntohl(ti->ti_dst.s_addr)))
goto drop;
if (tiflags & TH_ACK)
tcp_respond(tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
else {
if (tiflags & TH_SYN)
ti->ti_len++;
tcp_respond(tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
TH_RST|TH_ACK);
}
// destroy temporarily created socket
if (dropsocket)
soabort(so);
return;
drop:
TRACE("drop");
// Drop space held by incoming segment and return.
if (tp && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)) {
TRACE("tcp trace: drop a socket");
tcp_trace(TA_DROP, ostate, tp, &g_tcp_saveti, 0);
}
usn_free_cmbuf(m);
// destroy temporarily created socket
if (dropsocket)
soabort(so);
return;
}
int
tcp_reass(struct tcpcb *tp,struct tcpiphdr *ti, usn_mbuf_t *m)
{
struct tcpiphdr *q;
struct usn_socket *so = tp->t_inpcb->inp_socket;
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) {
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) {
g_tcpstat.tcps_rcvduppack++;
g_tcpstat.tcps_rcvdupbyte += ti->ti_len;
usn_free_mbuf(m);
return (0);
}
m_adj(m, i);
ti->ti_len -= i;
ti->ti_seq += i;
}
q = (struct tcpiphdr *)(q->ti_next);
}
g_tcpstat.tcps_rcvoopack++;
g_tcpstat.tcps_rcvoobyte += ti->ti_len;
REASS_MBUF(ti) = m; // XXX: wrong assumtion dst and src port for mbuf pointer
// While we overlap succeeding segments trim them or,
// if they are completely covered, dequeue them.
while (q != (struct tcpiphdr *)tp) {
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(REASS_MBUF(q), i);
break;
}
q = (struct tcpiphdr *)q->ti_next;
m = REASS_MBUF((struct tcpiphdr *)q->ti_prev);
// FIXME
//remque(q->ti_prev);
usn_free_mbuf(m);
}
// FIXME: 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;
// FIXME
//remque(ti);
m = REASS_MBUF(ti);
ti = (struct tcpiphdr *)ti->ti_next;
if (so->so_state & USN_CANTRCVMORE) {
usn_free_mbuf(m);
} else {
sbappend(so->so_rcv, m);
}
} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
sorwakeup(so);
usnet_tcpin_rwakeup(so, USN_TCP_IN, USN_TCPEV_READ, m);
return (flags);
}
int tcp_reass(struct tcpcb *tp, struct tcpiphdr *ti, struct mbuf *m)
{
register struct tcpiphdr *q;
struct socket *so = tp->t_socket;
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 = (struct tcpiphdr *)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) {
tcpstat.tcps_rcvduppack++;
tcpstat.tcps_rcvdupbyte += ti->ti_len;
m_freem(m);
/*
* Try to present any queued data
* at the left window edge to the user.
* This is needed after the 3-WHS
* completes.
*/
goto present; /* ??? */
}
m_adj(m, i);
ti->ti_len -= i;
ti->ti_seq += i;
}
q = (struct tcpiphdr *)(q->ti_next);
}
tcpstat.tcps_rcvoopack++;
tcpstat.tcps_rcvoobyte += ti->ti_len;
REASS_MBUF(ti) = (mbufp_32) m; /* XXX */
/*
* 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((struct mbuf *) REASS_MBUF(q), i);
break;
}
q = (struct tcpiphdr *)q->ti_next;
m = (struct mbuf *) REASS_MBUF((struct tcpiphdr *)q->ti_prev);
slirp_remque((void *)(q->ti_prev));
m_freem(m);
}
/*
* Stick new segment in its place.
*/
slirp_insque(ti, (void *)(q->ti_prev));
present:
/*
* Present data to user, advancing rcv_nxt through
* completed sequence space.
*/
if (!TCPS_HAVEESTABLISHED(tp->t_state))
return (0);
ti = (struct tcpiphdr *) 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;
slirp_remque(ti);
m = (struct mbuf *) REASS_MBUF(ti); /* XXX */
ti = (struct tcpiphdr *)ti->ti_next;
/* if (so->so_state & SS_FCANTRCVMORE) */
if (so->so_state & SS_FCANTSENDMORE)
m_freem(m);
else {
if (so->so_emu) {
if (tcp_emu(so,m)) sbappend(so, m);
} else
sbappend(so, m);
}
} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
/* sorwakeup(so); */
return (flags);
}
int
tcp_usrreq(struct socket * so,
struct mbuf * m,
struct mbuf * nam)
{
struct inpcb * inp;
struct tcpcb * tp;
int error = 0;
int req;
#ifdef DO_TCPTRACE
int ostate;
#endif
req = so->so_req; /* get request from socket struct */
inp = sotoinpcb(so);
/*
* When a TCP is attached to a socket, then there will be
* a (struct inpcb) pointed at by the socket, and this
* structure will point at a subsidary (struct tcpcb).
*/
if (inp == 0 && req != PRU_ATTACH)
{
return (EINVAL);
}
if (inp)
tp = intotcpcb(inp);
else /* inp and tp not set, make sure this is OK: */
{
if (req == PRU_ATTACH)
tp = NULL; /* stifle compiler warnings about using unassigned tp*/
else
{
dtrap(); /* programming error? */
return EINVAL;
}
}
switch (req)
{
/*
* TCP attaches to socket via PRU_ATTACH, reserving space,
* and an internet control block.
*/
case PRU_ATTACH:
if (inp)
{
error = EISCONN;
break;
}
error = tcp_attach(so);
if (error)
break;
if ((so->so_options & SO_LINGER) && so->so_linger == 0)
so->so_linger = TCP_LINGERTIME;
#ifdef DO_TCPTRACE
SETTP(tp, sototcpcb(so));
#endif
break;
/*
* PRU_DETACH detaches the TCP protocol from the socket.
* If the protocol state is non-embryonic, then can't
* do this directly: have to initiate a PRU_DISCONNECT,
* which may finish later; embryonic TCB's can just
* be discarded here.
*/
case PRU_DETACH:
if (tp->t_state > TCPS_LISTEN)
SETTP(tp, tcp_disconnect(tp));
else
SETTP(tp, tcp_close(tp));
break;
/*
* Give the socket an address.
*/
case PRU_BIND:
/* bind is quite different for IPv4 and v6, so we use two
* seperate pcbbind routines. so_domain was checked for
* validity way up in t_bind()
*/
#ifdef IP_V4
if(inp->inp_socket->so_domain == AF_INET)
{
error = in_pcbbind(inp, nam);
break;
}
#endif /* IP_V4 */
#ifdef IP_V6
if(inp->inp_socket->so_domain == AF_INET6)
{
error = ip6_pcbbind(inp, nam);
break;
}
#endif /* IP_V6 */
dtrap(); /* not v4 or v6? */
error = EINVAL;
break;
/*
* Prepare to accept connections.
*/
case PRU_LISTEN:
if (inp->inp_lport == 0)
error = in_pcbbind(inp, (struct mbuf *)0);
if (error == 0)
tp->t_state = TCPS_LISTEN;
break;
/*
* Initiate connection to peer.
* Create a template for use in transmissions on this connection.
* Enter SYN_SENT state, and mark socket as connecting.
* Start keep-alive timer, and seed output sequence space.
* Send initial segment on connection.
*/
case PRU_CONNECT:
if (inp->inp_lport == 0)
{
#ifdef IP_V4
#ifndef IP_V6 /* v4 only */
error = in_pcbbind(inp, (struct mbuf *)0);
#else /* dual mode */
if(so->so_domain == AF_INET)
error = in_pcbbind(inp, (struct mbuf *)0);
else
error = ip6_pcbbind(inp, (struct mbuf *)0);
#endif /* end dual mode code */
#else /* no v4, v6 only */
error = ip6_pcbbind(inp, (struct mbuf *)0);
#endif /* end v6 only */
if (error)
break;
}
#ifdef IP_V4
#ifndef IP_V6 /* v4 only */
error = in_pcbconnect(inp, nam);
#else /* dual mode */
if(so->so_domain == AF_INET)
error = in_pcbconnect(inp, nam);
else
error = ip6_pcbconnect(inp, nam);
#endif /* end dual mode code */
#else /* no v4, v6 only */
error = ip6_pcbconnect(inp, nam);
#endif /* end v6 only */
if (error)
break;
tp->t_template = tcp_template(tp);
if (tp->t_template == 0)
{
#ifdef IP_V4
#ifndef IP_V6 /* v4 only */
in_pcbdisconnect(inp);
#else /* dual mode */
if(so->so_domain == AF_INET)
in_pcbdisconnect(inp);
else
ip6_pcbdisconnect(inp);
#endif /* end dual mode code */
#else /* no v4, v6 only */
ip6_pcbdisconnect(inp);
#endif /* end v6 only */
error = ENOBUFS;
break;
}
soisconnecting(so);
tcpstat.tcps_connattempt++;
tp->t_state = TCPS_SYN_SENT;
tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
tp->iss = tcp_iss;
tcp_iss += (tcp_seq)(TCP_ISSINCR/2);
tcp_sendseqinit(tp);
error = tcp_output(tp);
if (!error)
TCP_MIB_INC(tcpActiveOpens); /* keep MIB stats */
break;
/*
* Create a TCP connection between two sockets.
*/
case PRU_CONNECT2:
error = EOPNOTSUPP;
break;
/*
* Initiate disconnect from peer.
* If connection never passed embryonic stage, just drop;
* else if don't need to let data drain, then can just drop anyways,
* else have to begin TCP shutdown process: mark socket disconnecting,
* drain unread data, state switch to reflect user close, and
* send segment (e.g. FIN) to peer. Socket will be really disconnected
* when peer sends FIN and acks ours.
*
* SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
*/
case PRU_DISCONNECT:
SETTP(tp, tcp_disconnect(tp));
break;
/*
* Accept a connection. Essentially all the work is
* done at higher levels; just return the address
* of the peer, storing through addr.
*/
case PRU_ACCEPT:
{
struct sockaddr_in * sin = mtod(nam, struct sockaddr_in *);
#ifdef IP_V6
struct sockaddr_in6 * sin6 = mtod(nam, struct sockaddr_in6 *);
#endif
#ifdef IP_V6
if (so->so_domain == AF_INET6)
{
nam->m_len = sizeof (struct sockaddr_in6);
sin6->sin6_port = inp->inp_fport;
sin6->sin6_family = AF_INET6;
IP6CPY(&sin6->sin6_addr, &inp->ip6_faddr);
}
#endif
#ifdef IP_V4
if (so->so_domain == AF_INET)
{
nam->m_len = sizeof (struct sockaddr_in);
sin->sin_family = AF_INET;
sin->sin_port = inp->inp_fport;
sin->sin_addr = inp->inp_faddr;
}
#endif
if ( !(so->so_domain == AF_INET) &&
!(so->so_domain == AF_INET6)
)
{
dprintf("*** PRU_ACCEPT bad domain = %d\n", so->so_domain);
dtrap();
}
TCP_MIB_INC(tcpPassiveOpens); /* keep MIB stats */
break;
}
/*
* Mark the connection as being incapable of further output.
*/
case PRU_SHUTDOWN:
socantsendmore(so);
tp = tcp_usrclosed(tp);
if (tp)
error = tcp_output(tp);
break;
/*
* After a receive, possibly send window update to peer.
*/
case PRU_RCVD:
(void) tcp_output(tp);
break;
/*
* Do a send by putting data in output queue and updating urgent
* marker if URG set. Possibly send more data.
*/
case PRU_SEND:
if (so->so_pcb == NULL)
{ /* Return EPIPE error if socket is not connected */
error = EPIPE;
break;
}
sbappend(&so->so_snd, m);
error = tcp_output(tp);
if (error == ENOBUFS)
sbdropend(&so->so_snd,m); /* Remove data from socket buffer */
break;
/*
* Abort the TCP.
*/
case PRU_ABORT:
SETTP(tp, tcp_drop(tp, ECONNABORTED));
break;
case PRU_SENSE:
/* ((struct stat *) m)->st_blksize = so->so_snd.sb_hiwat; */
dtrap(); /* does this ever happen? */
return (0);
case PRU_RCVOOB:
if ((so->so_oobmark == 0 &&
(so->so_state & SS_RCVATMARK) == 0) ||
#ifdef SO_OOBINLINE
so->so_options & SO_OOBINLINE ||
#endif
tp->t_oobflags & TCPOOB_HADDATA)
{
error = EINVAL;
break;
}
if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0)
{
error = EWOULDBLOCK;
break;
}
m->m_len = 1;
*mtod(m, char *) = tp->t_iobc;
if ((MBUF2LONG(nam) & MSG_PEEK) == 0)
tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
break;
case PRU_SENDOOB:
if (so->so_pcb == NULL)
{ /* Return EPIPE error if socket is not connected */
error = EPIPE;
break;
}
if (sbspace(&so->so_snd) == 0)
{
m_freem(m);
error = ENOBUFS;
break;
}
/*
* 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.
* Otherwise, snd_up should be one lower.
*/
sbappend(&so->so_snd, m);
tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
tp->t_force = 1;
error = tcp_output(tp);
if (error == ENOBUFS)
sbdropend(&so->so_snd,m); /* Remove data from socket buffer */
tp->t_force = 0;
break;
case PRU_SOCKADDR:
/* sockaddr and peeraddr have to switch based on IP type */
#ifdef IP_V4
#ifndef IP_V6 /* v4 only */
in_setsockaddr(inp, nam);
#else /* dual mode */
if(so->so_domain == AF_INET6)
ip6_setsockaddr(inp, nam);
else
in_setsockaddr(inp, nam);
#endif /* dual mode */
#else /* IP_V6 */
ip6_setsockaddr(inp, nam);
#endif
break;
case PRU_PEERADDR:
#ifdef IP_V4
#ifndef IP_V6 /* v4 only */
in_setpeeraddr(inp, nam);
#else /* dual mode */
if(so->so_domain == AF_INET6)
ip6_setpeeraddr(inp, nam);
else
in_setpeeraddr(inp, nam);
#endif /* dual mode */
#else /* IP_V6 */
ip6_setpeeraddr(inp, nam);
#endif
break;
case PRU_SLOWTIMO:
SETTP(tp, tcp_timers(tp, (int)MBUF2LONG(nam)));
#ifdef DO_TCPTRACE
req |= (long)nam << 8; /* for debug's sake */
#endif
break;
default:
panic("tcp_usrreq");
}
#ifdef DO_TCPTRACE
if (tp && (so->so_options & SO_DEBUG))
tcp_trace("usrreq: state: %d, tcpcb: %x, req: %d",
ostate, tp, req);
#endif
return (error);
}
/*
* XXX This is ugly
* We create and bind a socket, then fork off to another
* process, which connects to this socket, after which we
* exec the wanted program. If something (strange) happens,
* the accept() call could block us forever.
*
* do_pty = 0 Fork/exec inetd style
* do_pty = 1 Fork/exec using slirp.telnetd
* do_ptr = 2 Fork/exec using pty
*/
int fork_exec(struct socket *so, char *ex, int do_pty)
{
int s;
struct sockaddr_in addr;
socklen_t addrlen = sizeof(addr);
int opt;
int master;
char *argv[256];
#if 0
char buff[256];
#endif
/* don't want to clobber the original */
char *bptr;
char *curarg;
int c, i, ret;
DEBUG_CALL("fork_exec");
DEBUG_ARG("so = %lx", (long)so);
DEBUG_ARG("ex = %lx", (long)ex);
DEBUG_ARG("do_pty = %lx", (long)do_pty);
if (do_pty == 2) {
if (slirp_openpty(&master, &s) == -1) {
lprint("Error: openpty failed: %s\n", strerror(errno));
return 0;
}
} else {
memset(&addr, 0, sizeof(struct sockaddr_in));
addr.sin_family = AF_INET;
addr.sin_port = 0;
addr.sin_addr.s_addr = INADDR_ANY;
if ((s = socket(AF_INET, SOCK_STREAM, 0)) < 0 ||
bind(s, (struct sockaddr *)&addr, addrlen) < 0 ||
listen(s, 1) < 0) {
lprint("Error: inet socket: %s\n", strerror(errno));
closesocket(s);
return 0;
}
}
switch(fork()) {
case -1:
lprint("Error: fork failed: %s\n", strerror(errno));
close(s);
if (do_pty == 2)
close(master);
return 0;
case 0:
/* Set the DISPLAY */
if (do_pty == 2) {
(void) close(master);
#ifdef TIOCSCTTY /* XXXXX */
(void) setsid();
ioctl(s, TIOCSCTTY, (char *)NULL);
#endif
} else {
getsockname(s, (struct sockaddr *)&addr, &addrlen);
close(s);
/*
* Connect to the socket
* XXX If any of these fail, we're in trouble!
*/
s = socket(AF_INET, SOCK_STREAM, 0);
addr.sin_addr = loopback_addr;
do {
ret = connect(s, (struct sockaddr *)&addr, addrlen);
} while (ret < 0 && errno == EINTR);
}
#if 0
if (x_port >= 0) {
#ifdef HAVE_SETENV
sprintf(buff, "%s:%d.%d", inet_ntoa(our_addr), x_port, x_screen);
setenv("DISPLAY", buff, 1);
#else
sprintf(buff, "DISPLAY=%s:%d.%d", inet_ntoa(our_addr), x_port, x_screen);
putenv(buff);
#endif
}
#endif
dup2(s, 0);
dup2(s, 1);
dup2(s, 2);
for (s = 3; s <= 255; s++)
close(s);
i = 0;
bptr = strdup(ex); /* No need to free() this */
if (do_pty == 1) {
/* Setup "slirp.telnetd -x" */
argv[i++] = "slirp.telnetd";
argv[i++] = "-x";
argv[i++] = bptr;
} else
do {
/* Change the string into argv[] */
curarg = bptr;
while (*bptr != ' ' && *bptr != (char)0)
bptr++;
c = *bptr;
*bptr++ = (char)0;
argv[i++] = strdup(curarg);
} while (c);
argv[i] = 0;
execvp(argv[0], argv);
/* Ooops, failed, let's tell the user why */
{
char buff[256];
sprintf(buff, "Error: execvp of %s failed: %s\n",
argv[0], strerror(errno));
write(2, buff, strlen(buff)+1);
}
close(0); close(1); close(2); /* XXX */
exit(1);
default:
if (do_pty == 2) {
close(s);
so->s = master;
} else {
/*
* XXX this could block us...
* XXX Should set a timer here, and if accept() doesn't
* return after X seconds, declare it a failure
* The only reason this will block forever is if socket()
* of connect() fail in the child process
*/
do {
so->s = accept(s, (struct sockaddr *)&addr, &addrlen);
} while (so->s < 0 && errno == EINTR);
closesocket(s);
opt = 1;
setsockopt(so->s,SOL_SOCKET,SO_REUSEADDR,(char *)&opt,sizeof(int));
opt = 1;
setsockopt(so->s,SOL_SOCKET,SO_OOBINLINE,(char *)&opt,sizeof(int));
}
fd_nonblock(so->s);
/* Append the telnet options now */
if (so->so_m != 0 && do_pty == 1) {
sbappend(so, so->so_m);
so->so_m = 0;
}
return 1;
}
}
/*
* Implement receive operations on a socket.
*
* We depend on the way that records are added to the signalsockbuf
* by sbappend*. In particular, each record (mbufs linked through m_next)
* must begin with an address if the protocol so specifies,
* followed by an optional mbuf or mbufs containing ancillary data,
* and then zero or more mbufs of data.
*
* Although the signalsockbuf is locked, new data may still be appended.
* A token inside the ssb_lock deals with MP issues and still allows
* the network to access the socket if we block in a uio.
*
* The caller may receive the data as a single mbuf chain by supplying
* an mbuf **mp0 for use in returning the chain. The uio is then used
* only for the count in uio_resid.
*/
int
soreceive(struct socket *so, struct sockaddr **psa, struct uio *uio,
struct sockbuf *sio, struct mbuf **controlp, int *flagsp)
{
struct mbuf *m, *n;
struct mbuf *free_chain = NULL;
int flags, len, error, offset;
struct protosw *pr = so->so_proto;
int moff, type = 0;
size_t resid, orig_resid;
if (uio)
resid = uio->uio_resid;
else
resid = (size_t)(sio->sb_climit - sio->sb_cc);
orig_resid = resid;
if (psa)
*psa = NULL;
if (controlp)
*controlp = NULL;
if (flagsp)
flags = *flagsp &~ MSG_EOR;
else
flags = 0;
if (flags & MSG_OOB) {
m = m_get(MB_WAIT, MT_DATA);
if (m == NULL)
return (ENOBUFS);
error = so_pru_rcvoob(so, m, flags & MSG_PEEK);
if (error)
goto bad;
if (sio) {
do {
sbappend(sio, m);
KKASSERT(resid >= (size_t)m->m_len);
resid -= (size_t)m->m_len;
} while (resid > 0 && m);
} else {
do {
uio->uio_resid = resid;
error = uiomove(mtod(m, caddr_t),
(int)szmin(resid, m->m_len),
uio);
resid = uio->uio_resid;
m = m_free(m);
} while (uio->uio_resid && error == 0 && m);
}
bad:
if (m)
m_freem(m);
return (error);
}
if ((so->so_state & SS_ISCONFIRMING) && resid)
so_pru_rcvd(so, 0);
/*
* The token interlocks against the protocol thread while
* ssb_lock is a blocking lock against other userland entities.
*/
lwkt_gettoken(&so->so_rcv.ssb_token);
restart:
error = ssb_lock(&so->so_rcv, SBLOCKWAIT(flags));
if (error)
goto done;
m = so->so_rcv.ssb_mb;
/*
* If we have less data than requested, block awaiting more
* (subject to any timeout) if:
* 1. the current count is less than the low water mark, or
* 2. MSG_WAITALL is set, and it is possible to do the entire
* receive operation at once if we block (resid <= hiwat).
* 3. MSG_DONTWAIT is not set
* If MSG_WAITALL is set but resid is larger than the receive buffer,
* we have to do the receive in sections, and thus risk returning
* a short count if a timeout or signal occurs after we start.
*/
if (m == NULL || (((flags & MSG_DONTWAIT) == 0 &&
(size_t)so->so_rcv.ssb_cc < resid) &&
(so->so_rcv.ssb_cc < so->so_rcv.ssb_lowat ||
((flags & MSG_WAITALL) && resid <= (size_t)so->so_rcv.ssb_hiwat)) &&
m->m_nextpkt == 0 && (pr->pr_flags & PR_ATOMIC) == 0)) {
KASSERT(m != NULL || !so->so_rcv.ssb_cc, ("receive 1"));
if (so->so_error) {
if (m)
goto dontblock;
error = so->so_error;
if ((flags & MSG_PEEK) == 0)
so->so_error = 0;
goto release;
}
if (so->so_state & SS_CANTRCVMORE) {
if (m)
goto dontblock;
else
goto release;
}
for (; m; m = m->m_next) {
if (m->m_type == MT_OOBDATA || (m->m_flags & M_EOR)) {
m = so->so_rcv.ssb_mb;
goto dontblock;
}
}
if ((so->so_state & (SS_ISCONNECTED|SS_ISCONNECTING)) == 0 &&
(pr->pr_flags & PR_CONNREQUIRED)) {
error = ENOTCONN;
goto release;
}
if (resid == 0)
goto release;
if (flags & (MSG_FNONBLOCKING|MSG_DONTWAIT)) {
error = EWOULDBLOCK;
goto release;
}
ssb_unlock(&so->so_rcv);
error = ssb_wait(&so->so_rcv);
if (error)
goto done;
goto restart;
}
dontblock:
if (uio && uio->uio_td && uio->uio_td->td_proc)
uio->uio_td->td_lwp->lwp_ru.ru_msgrcv++;
/*
* note: m should be == sb_mb here. Cache the next record while
* cleaning up. Note that calling m_free*() will break out critical
* section.
*/
KKASSERT(m == so->so_rcv.ssb_mb);
/*
* Skip any address mbufs prepending the record.
*/
if (pr->pr_flags & PR_ADDR) {
KASSERT(m->m_type == MT_SONAME, ("receive 1a"));
orig_resid = 0;
if (psa)
*psa = dup_sockaddr(mtod(m, struct sockaddr *));
if (flags & MSG_PEEK)
m = m->m_next;
else
m = sbunlinkmbuf(&so->so_rcv.sb, m, &free_chain);
}
/*
* XXX This is ugly
* We create and bind a socket, then fork off to another
* process, which connects to this socket, after which we
* exec the wanted program. If something (strange) happens,
* the accept() call could block us forever.
*
* do_pty = 0 Fork/exec inetd style
* do_pty = 1 Fork/exec using slirp.telnetd
* do_ptr = 2 Fork/exec using pty
*/
int fork_exec(struct socket *so, const char *ex, int do_pty)
{
int s;
struct sockaddr_in addr;
socklen_t addrlen = sizeof(addr);
int opt;
int master = -1;
const char *argv[256];
/* don't want to clobber the original */
char *bptr;
const char *curarg;
int c, i, ret;
pid_t pid;
DEBUG_CALL("fork_exec so = %lx ex = %lx do_pty = %lx", (long)so, (long)ex, (long)do_pty);
if (do_pty == 2) {
return 0;
} else {
addr.sin_family = AF_INET;
addr.sin_port = 0;
addr.sin_addr.s_addr = INADDR_ANY;
if ((s = clo_socket(AF_INET, SOCK_STREAM, 0)) < 0 ||
bind(s, (struct sockaddr *)&addr, addrlen) < 0 ||
listen(s, 1) < 0) {
vmiPrintf("Error: inet socket: %s\n", strerror(errno));
closesocket(s);
return 0;
}
}
pid = fork();
switch(pid) {
case -1:
vmiPrintf("Error: fork failed: %s\n", strerror(errno));
close(s);
if (do_pty == 2)
close(master);
return 0;
case 0:
setsid();
/* Set the DISPLAY */
if (do_pty == 2) {
(void) close(master);
#ifdef TIOCSCTTY /* XXXXX */
ioctl(s, TIOCSCTTY, (char *)NULL);
#endif
} else {
getsockname(s, (struct sockaddr *)&addr, &addrlen);
close(s);
/*
* Connect to the socket
* XXX If any of these fail, we're in trouble!
*/
s = clo_socket(AF_INET, SOCK_STREAM, 0);
addr.sin_addr = loopback_addr;
do {
ret = connect(s, (struct sockaddr *)&addr, addrlen);
} while (ret < 0 && errno == EINTR);
}
dup2(s, 0);
dup2(s, 1);
dup2(s, 2);
for (s = getdtablesize() - 1; s >= 3; s--)
close(s);
i = 0;
bptr = strdup(ex);
if (do_pty == 1) {
/* Setup "slirp.telnetd -x" */
argv[i++] = "slirp.telnetd";
argv[i++] = "-x";
argv[i++] = bptr;
} else
do {
/* Change the string into argv[] */
curarg = bptr;
while (*bptr != ' ' && *bptr != (char)0)
bptr++;
c = *bptr;
*bptr++ = (char)0;
argv[i++] = strdup(curarg);
} while (c);
argv[i] = NULL;
execvp(argv[0], (char **)argv);
/* Ooops, failed, let's tell the user why */
fprintf(stderr, "Error: execvp of %s failed: %s\n",
argv[0], strerror(errno));
close(0); close(1); close(2); /* XXX */
exit(1);
default:
// MWH qemu_add_child_watch(pid);
if (do_pty == 2) {
close(s);
so->s = master;
} else {
/*
* XXX this could block us...
* XXX Should set a timer here, and if accept() doesn't
* return after X seconds, declare it a failure
* The only reason this will block forever is if socket()
* of connect() fail in the child process
*/
do {
so->s = accept(s, (struct sockaddr *)&addr, &addrlen);
} while (so->s < 0 && errno == EINTR);
closesocket(s);
opt = 1;
setsockopt(so->s,SOL_SOCKET,SO_REUSEADDR,(char *)&opt,sizeof(int));
opt = 1;
setsockopt(so->s,SOL_SOCKET,SO_OOBINLINE,(char *)&opt,sizeof(int));
}
fd_nonblock(so->s);
/* Append the telnet options now */
if (so->so_m != NULL && do_pty == 1) {
sbappend(so, so->so_m);
so->so_m = NULL;
}
return 1;
}
}
