static void
tcp_cubic_post_fr(struct tcpcb *tp, struct tcphdr *th)
{
uint32_t flight_size = 0;
if (SEQ_LEQ(th->th_ack, tp->snd_max))
flight_size = tp->snd_max - th->th_ack;
/*
* Complete ack. The current window was inflated for fast recovery.
* It has to be deflated post recovery.
*
* Window inflation should have left us with approx snd_ssthresh
* outstanding data. If the flight size is zero or one segment,
* make congestion window to be at least as big as 2 segments to
* avoid delayed acknowledgements. This is according to RFC 6582.
*/
if (flight_size < tp->snd_ssthresh)
tp->snd_cwnd = max(flight_size, tp->t_maxseg)
+ tp->t_maxseg;
else
tp->snd_cwnd = tp->snd_ssthresh;
tp->t_ccstate->cub_tcp_win = 0;
tp->t_ccstate->cub_target_win = 0;
tp->t_ccstate->cub_tcp_bytes_acked = 0;
}
/**
* \brief Update stream with SACK records from a TCP packet.
*
* \param stream The stream to update.
* \param p packet to get the SACK records from
*
* \retval -1 error
* \retval 0 ok
*/
int StreamTcpSackUpdatePacket(TcpStream *stream, Packet *p) {
int records = TCP_GET_SACK_CNT(p);
int record = 0;
TCPOptSackRecord *sack_rec = (TCPOptSackRecord *)(TCP_GET_SACK_PTR(p));
for (record = 0; record < records; record++) {
SCLogDebug("%p last_ack %u, left edge %u, right edge %u", sack_rec,
stream->last_ack, ntohl(sack_rec->le), ntohl(sack_rec->re));
if (SEQ_LEQ(ntohl(sack_rec->re), stream->last_ack)) {
SCLogDebug("record before last_ack");
goto next;
}
/** \todo need a metric to a check for a right edge limit */
/*
if (SEQ_GT(ntohl(sack_rec->re), stream->next_seq)) {
SCLogDebug("record beyond next_seq %u", stream->next_seq);
goto next;
}
*/
if (SEQ_GEQ(ntohl(sack_rec->le), ntohl(sack_rec->re))) {
SCLogDebug("invalid record: le >= re");
goto next;
}
if (StreamTcpSackInsertRange(stream, ntohl(sack_rec->le),
ntohl(sack_rec->re)) == -1)
{
SCReturnInt(-1);
}
next:
sack_rec++;
}
#ifdef DEBUG
StreamTcpSackPrintList(stream);
#endif
SCReturnInt(0);
}
/*
* Ertt_packet_measurements uses a small amount of state kept on each packet
* sent to match incoming acknowledgements. This enables more accurate and
* secure round trip time measurements. The resulting measurement is used for
* congestion control algorithms which require a more accurate time.
* Ertt_packet_measurements is called via the helper hook in tcp_input.c
*/
static int
ertt_packet_measurement_hook(int hhook_type, int hhook_id, void *udata,
void *ctx_data, void *hdata, struct osd *hosd)
{
struct ertt *e_t;
struct tcpcb *tp;
struct tcphdr *th;
struct tcpopt *to;
struct tcp_hhook_data *thdp;
struct txseginfo *txsi;
int acked, measurenext_len, multiack, new_sacked_bytes, rtt_bytes_adjust;
uint32_t measurenext, rts;
tcp_seq ack;
KASSERT(ctx_data != NULL, ("%s: ctx_data is NULL!", __func__));
KASSERT(hdata != NULL, ("%s: hdata is NULL!", __func__));
e_t = (struct ertt *)hdata;
thdp = ctx_data;
tp = thdp->tp;
th = thdp->th;
to = thdp->to;
new_sacked_bytes = (tp->sackhint.last_sack_ack != 0);
measurenext = measurenext_len = multiack = rts = rtt_bytes_adjust = 0;
acked = th->th_ack - tp->snd_una;
INP_WLOCK_ASSERT(tp->t_inpcb);
/* Packet has provided new acknowledgements. */
if (acked > 0 || new_sacked_bytes) {
if (acked == 0 && new_sacked_bytes) {
/* Use last sacked data. */
ack = tp->sackhint.last_sack_ack;
} else
ack = th->th_ack;
txsi = TAILQ_FIRST(&e_t->txsegi_q);
while (txsi != NULL) {
rts = 0;
/* Acknowledgement is acking more than this txsi. */
if (SEQ_GT(ack, txsi->seq + txsi->len)) {
if (txsi->flags & TXSI_RTT_MEASURE_START ||
measurenext) {
marked_packet_rtt(txsi, e_t, tp,
&measurenext, &measurenext_len,
&rtt_bytes_adjust, MULTI_ACK);
}
TAILQ_REMOVE(&e_t->txsegi_q, txsi, txsegi_lnk);
uma_zfree(txseginfo_zone, txsi);
txsi = TAILQ_FIRST(&e_t->txsegi_q);
continue;
}
/*
* Guess if delayed acks are being used by the receiver.
*
* XXXDH: A simple heuristic that could be improved
*/
if (!new_sacked_bytes) {
if (acked > tp->t_maxseg) {
e_t->dlyack_rx +=
(e_t->dlyack_rx < DLYACK_SMOOTH) ?
1 : 0;
multiack = 1;
} else if (acked > txsi->len) {
multiack = 1;
e_t->dlyack_rx +=
(e_t->dlyack_rx < DLYACK_SMOOTH) ?
1 : 0;
} else if (acked == tp->t_maxseg ||
acked == txsi->len) {
e_t->dlyack_rx -=
(e_t->dlyack_rx > 0) ? 1 : 0;
}
/* Otherwise leave dlyack_rx the way it was. */
}
/*
* Time stamps are only to help match the txsi with the
* received acknowledgements.
*/
if (e_t->timestamp_errors < MAX_TS_ERR &&
(to->to_flags & TOF_TS) != 0 && to->to_tsecr) {
/*
* Note: All packets sent with the offload will
* have the same time stamp. If we are sending
* on a fast interface and the t_maxseg is much
* smaller than one tick, this will be fine. The
* time stamp would be the same whether we were
* using tso or not. However, if the interface
* is slow, this will cause problems with the
* calculations. If the interface is slow, there
* is not reason to be using tso, and it should
* be turned off.
*/
/*
* If there are too many time stamp errors, time
* stamps won't be trusted
*/
rts = to->to_tsecr;
/* Before this packet. */
if (!e_t->dlyack_rx && TSTMP_LT(rts, txsi->tx_ts))
/* When delayed acking is used, the
* reflected time stamp is of the first
* packet and thus may be before
* txsi->tx_ts.
*/
break;
if (TSTMP_GT(rts, txsi->tx_ts)) {
/*
* If reflected time stamp is later than
* tx_tsi, then this txsi is old.
*/
if (txsi->flags & TXSI_RTT_MEASURE_START
|| measurenext) {
marked_packet_rtt(txsi, e_t, tp,
&measurenext, &measurenext_len,
&rtt_bytes_adjust, OLD_TXSI);
}
TAILQ_REMOVE(&e_t->txsegi_q, txsi,
txsegi_lnk);
uma_zfree(txseginfo_zone, txsi);
txsi = TAILQ_FIRST(&e_t->txsegi_q);
continue;
}
if (rts == txsi->tx_ts &&
TSTMP_LT(to->to_tsval, txsi->rx_ts)) {
/*
* Segment received before sent!
* Something is wrong with the received
* timestamps so increment errors. If
* this keeps up we will ignore
* timestamps.
*/
e_t->timestamp_errors++;
}
}
/*
* Acknowledging a sequence number before this txsi.
* If it is an old txsi that may have had the same seq
* numbers, it should have been removed if time stamps
* are being used.
*/
if (SEQ_LEQ(ack, txsi->seq))
break; /* Before first packet in txsi. */
/*
* Only ack > txsi->seq and ack <= txsi->seq+txsi->len
* past this point.
*
* If delayed acks are being used, an acknowledgement
* for a single segment will have been delayed by the
* receiver and will yield an inaccurate measurement. In
* this case, we only make the measurement if more than
* one segment is being acknowledged or sack is
* currently being used.
*/
if (!e_t->dlyack_rx || multiack || new_sacked_bytes) {
/* Make an accurate new measurement. */
e_t->rtt = tcp_ts_getticks() - txsi->tx_ts + 1;
if (e_t->rtt < e_t->minrtt || e_t->minrtt == 0)
e_t->minrtt = e_t->rtt;
if (e_t->rtt > e_t->maxrtt || e_t->maxrtt == 0)
e_t->maxrtt = e_t->rtt;
}
if (txsi->flags & TXSI_RTT_MEASURE_START || measurenext)
marked_packet_rtt(txsi, e_t, tp,
&measurenext, &measurenext_len,
&rtt_bytes_adjust, CORRECT_ACK);
if (txsi->flags & TXSI_TSO) {
if (txsi->len > acked) {
txsi->len -= acked;
/*
* This presumes ack for first bytes in
* txsi, this may not be true but it
* shouldn't cause problems for the
* timing.
*
* We remeasure RTT even though we only
* have a single txsi. The rationale
* behind this is that it is better to
* have a slightly inaccurate
* measurement than no additional
* measurement for the rest of the bulk
* transfer. Since TSO is only used on
* high speed interface cards, so the
* packets should be transmitted at line
* rate back to back with little
* difference in transmission times (in
* ticks).
*/
txsi->seq += acked;
/*
* Reset txsi measure flag so we don't
* use it for another RTT measurement.
*/
txsi->flags &= ~TXSI_RTT_MEASURE_START;
/*
* There is still more data to be acked
* from tso bulk transmission, so we
* won't remove it from the TAILQ yet.
*/
break;
}
txsi->len = 0;
}
TAILQ_REMOVE(&e_t->txsegi_q, txsi, txsegi_lnk);
uma_zfree(txseginfo_zone, txsi);
break;
}
if (measurenext) {
/*
* We need to do a RTT measurement. It won't be the best
* if we do it here.
*/
marked_packet_rtt(txsi, e_t, tp,
&measurenext, &measurenext_len,
&rtt_bytes_adjust, FORCED_MEASUREMENT);
}
}
return (0);
}
/*
* 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);
}
/*
* 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(struct tcpcb *tp, tcp_seq rcv_start, tcp_seq rcv_end)
{
/*
* 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.
*/
struct sackblk head_blk, saved_blks[MAX_SACK_BLKS];
int num_head, num_saved, i;
/* SACK block for the received segment. */
head_blk.start = rcv_start;
head_blk.end = rcv_end;
/*
* Merge updated SACK blocks into head_blk, and
* save unchanged SACK blocks into saved_blks[].
* num_saved will have the number of the saved SACK blocks.
*/
num_saved = 0;
for (i = 0; i < tp->rcv_numsacks; i++) {
tcp_seq start = tp->sackblks[i].start;
tcp_seq end = tp->sackblks[i].end;
if (SEQ_GEQ(start, end) || SEQ_LEQ(start, tp->rcv_nxt)) {
/*
* Discard this SACK block.
*/
} else if (SEQ_LEQ(head_blk.start, end) &&
SEQ_GEQ(head_blk.end, start)) {
/*
* Merge this SACK block into head_blk.
* This SACK block itself will be discarded.
*/
if (SEQ_GT(head_blk.start, start))
head_blk.start = start;
if (SEQ_LT(head_blk.end, end))
head_blk.end = end;
} else {
/*
* Save this SACK block.
*/
saved_blks[num_saved].start = start;
saved_blks[num_saved].end = end;
num_saved++;
}
}
/*
* Update SACK list in tp->sackblks[].
*/
num_head = 0;
if (SEQ_GT(head_blk.start, tp->rcv_nxt)) {
/*
* The received data segment is an out-of-order segment.
* Put head_blk at the top of SACK list.
*/
tp->sackblks[0] = head_blk;
num_head = 1;
/*
* If the number of saved SACK blocks exceeds its limit,
* discard the last SACK block.
*/
if (num_saved >= MAX_SACK_BLKS)
num_saved--;
}
if (num_saved > 0) {
/*
* Copy the saved SACK blocks back.
*/
bcopy(saved_blks, &tp->sackblks[num_head],
sizeof(struct sackblk) * num_saved);
}
/* Save the number of SACK blocks. */
tp->rcv_numsacks = num_head + num_saved;
/* If we are requesting SACK recovery, reset the stretch-ack state
* so that connection will generate more acks after recovery and
* sender's cwnd will open.
*/
if ((tp->t_flags & TF_STRETCHACK) != 0 && tp->rcv_numsacks > 0)
tcp_reset_stretch_ack(tp);
#if TRAFFIC_MGT
if (tp->acc_iaj > 0 && tp->rcv_numsacks > 0)
reset_acc_iaj(tp);
#endif /* TRAFFIC_MGT */
}
/*
* Process cumulative ACK and the TCP SACK option to update the scoreboard.
* tp->snd_holes is an ordered list of holes (oldest to newest, in terms of
* the sequence space).
*/
void
tcp_sack_doack(struct tcpcb *tp, struct tcpopt *to, tcp_seq th_ack)
{
struct sackhole *cur, *temp;
struct sackblk sack, sack_blocks[TCP_MAX_SACK + 1], *sblkp;
int i, j, num_sack_blks;
//ScenSim-Port// INP_WLOCK_ASSERT(tp->t_inpcb);
num_sack_blks = 0;
/*
* If SND.UNA will be advanced by SEG.ACK, and if SACK holes exist,
* treat [SND.UNA, SEG.ACK) as if it is a SACK block.
*/
if (SEQ_LT(tp->snd_una, th_ack) && !TAILQ_EMPTY(&tp->snd_holes)) {
sack_blocks[num_sack_blks].start = tp->snd_una;
sack_blocks[num_sack_blks++].end = th_ack;
}
/*
* Append received valid SACK blocks to sack_blocks[], but only if we
* received new blocks from the other side.
*/
if (to->to_flags & TOF_SACK) {
for (i = 0; i < to->to_nsacks; i++) {
bcopy((to->to_sacks + i * TCPOLEN_SACK),
&sack, sizeof(sack));
sack.start = ntohl(sack.start);
sack.end = ntohl(sack.end);
if (SEQ_GT(sack.end, sack.start) &&
SEQ_GT(sack.start, tp->snd_una) &&
SEQ_GT(sack.start, th_ack) &&
SEQ_LT(sack.start, tp->snd_max) &&
SEQ_GT(sack.end, tp->snd_una) &&
SEQ_LEQ(sack.end, tp->snd_max))
sack_blocks[num_sack_blks++] = sack;
}
}
/*
* Return if SND.UNA is not advanced and no valid SACK block is
* received.
*/
if (num_sack_blks == 0)
return;
/*
* Sort the SACK blocks so we can update the scoreboard with just one
* pass. The overhead of sorting upto 4+1 elements is less than
* making upto 4+1 passes over the scoreboard.
*/
for (i = 0; i < num_sack_blks; i++) {
for (j = i + 1; j < num_sack_blks; j++) {
if (SEQ_GT(sack_blocks[i].end, sack_blocks[j].end)) {
sack = sack_blocks[i];
sack_blocks[i] = sack_blocks[j];
sack_blocks[j] = sack;
}
}
}
if (TAILQ_EMPTY(&tp->snd_holes))
/*
* Empty scoreboard. Need to initialize snd_fack (it may be
* uninitialized or have a bogus value). Scoreboard holes
* (from the sack blocks received) are created later below
* (in the logic that adds holes to the tail of the
* scoreboard).
*/
tp->snd_fack = SEQ_MAX(tp->snd_una, th_ack);
/*
* In the while-loop below, incoming SACK blocks (sack_blocks[]) and
* SACK holes (snd_holes) are traversed from their tails with just
* one pass in order to reduce the number of compares especially when
* the bandwidth-delay product is large.
*
* Note: Typically, in the first RTT of SACK recovery, the highest
* three or four SACK blocks with the same ack number are received.
* In the second RTT, if retransmitted data segments are not lost,
* the highest three or four SACK blocks with ack number advancing
* are received.
*/
sblkp = &sack_blocks[num_sack_blks - 1]; /* Last SACK block */
tp->sackhint.last_sack_ack = sblkp->end;
if (SEQ_LT(tp->snd_fack, sblkp->start)) {
/*
* The highest SACK block is beyond fack. Append new SACK
* hole at the tail. If the second or later highest SACK
* blocks are also beyond the current fack, they will be
* inserted by way of hole splitting in the while-loop below.
*/
temp = tcp_sackhole_insert(tp, tp->snd_fack,sblkp->start,NULL);
if (temp != NULL) {
tp->snd_fack = sblkp->end;
/* Go to the previous sack block. */
sblkp--;
} else {
/*
* We failed to add a new hole based on the current
* sack block. Skip over all the sack blocks that
* fall completely to the right of snd_fack and
* proceed to trim the scoreboard based on the
* remaining sack blocks. This also trims the
* scoreboard for th_ack (which is sack_blocks[0]).
*/
while (sblkp >= sack_blocks &&
SEQ_LT(tp->snd_fack, sblkp->start))
sblkp--;
if (sblkp >= sack_blocks &&
SEQ_LT(tp->snd_fack, sblkp->end))
tp->snd_fack = sblkp->end;
}
} else if (SEQ_LT(tp->snd_fack, sblkp->end))
/* fack is advanced. */
tp->snd_fack = sblkp->end;
/* We must have at least one SACK hole in scoreboard. */
//ScenSim-Port// KASSERT(!TAILQ_EMPTY(&tp->snd_holes),
//ScenSim-Port// ("SACK scoreboard must not be empty"));
cur = TAILQ_LAST(&tp->snd_holes, sackhole_head); /* Last SACK hole. */
/*
* Since the incoming sack blocks are sorted, we can process them
* making one sweep of the scoreboard.
*/
while (sblkp >= sack_blocks && cur != NULL) {
if (SEQ_GEQ(sblkp->start, cur->end)) {
/*
* SACKs data beyond the current hole. Go to the
* previous sack block.
*/
sblkp--;
continue;
}
if (SEQ_LEQ(sblkp->end, cur->start)) {
/*
* SACKs data before the current hole. Go to the
* previous hole.
*/
cur = TAILQ_PREV(cur, sackhole_head, scblink);
continue;
}
tp->sackhint.sack_bytes_rexmit -= (cur->rxmit - cur->start);
//ScenSim-Port// KASSERT(tp->sackhint.sack_bytes_rexmit >= 0,
//ScenSim-Port// ("sackhint bytes rtx >= 0"));
if (SEQ_LEQ(sblkp->start, cur->start)) {
/* Data acks at least the beginning of hole. */
if (SEQ_GEQ(sblkp->end, cur->end)) {
/* Acks entire hole, so delete hole. */
temp = cur;
cur = TAILQ_PREV(cur, sackhole_head, scblink);
tcp_sackhole_remove(tp, temp);
/*
* The sack block may ack all or part of the
* next hole too, so continue onto the next
* hole.
*/
continue;
} else {
/* Move start of hole forward. */
cur->start = sblkp->end;
cur->rxmit = SEQ_MAX(cur->rxmit, cur->start);
}
} else {
/* Data acks at least the end of hole. */
if (SEQ_GEQ(sblkp->end, cur->end)) {
/* Move end of hole backward. */
cur->end = sblkp->start;
cur->rxmit = SEQ_MIN(cur->rxmit, cur->end);
} else {
/*
* ACKs some data in middle of a hole; need
* to split current hole
*/
temp = tcp_sackhole_insert(tp, sblkp->end,
cur->end, cur);
if (temp != NULL) {
if (SEQ_GT(cur->rxmit, temp->rxmit)) {
temp->rxmit = cur->rxmit;
tp->sackhint.sack_bytes_rexmit
+= (temp->rxmit
- temp->start);
}
cur->end = sblkp->start;
cur->rxmit = SEQ_MIN(cur->rxmit,
cur->end);
}
}
}
tp->sackhint.sack_bytes_rexmit += (cur->rxmit - cur->start);
/*
* Testing sblkp->start against cur->start tells us whether
* we're done with the sack block or the sack hole.
* Accordingly, we advance one or the other.
*/
if (SEQ_LEQ(sblkp->start, cur->start))
cur = TAILQ_PREV(cur, sackhole_head, scblink);
else
sblkp--;
}
}
/*
* 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(struct tcpcb *tp, tcp_seq rcv_start, tcp_seq rcv_end)
{
/*
* 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.
*/
struct sackblk head_blk, saved_blks[MAX_SACK_BLKS];
int num_head, num_saved, i;
//ScenSim-Port// INP_WLOCK_ASSERT(tp->t_inpcb);
/* Check arguments. */
//ScenSim-Port// KASSERT(SEQ_LT(rcv_start, rcv_end), ("rcv_start < rcv_end"));
/* SACK block for the received segment. */
head_blk.start = rcv_start;
head_blk.end = rcv_end;
/*
* Merge updated SACK blocks into head_blk, and save unchanged SACK
* blocks into saved_blks[]. num_saved will have the number of the
* saved SACK blocks.
*/
num_saved = 0;
for (i = 0; i < tp->rcv_numsacks; i++) {
tcp_seq start = tp->sackblks[i].start;
tcp_seq end = tp->sackblks[i].end;
if (SEQ_GEQ(start, end) || SEQ_LEQ(start, tp->rcv_nxt)) {
/*
* Discard this SACK block.
*/
} else if (SEQ_LEQ(head_blk.start, end) &&
SEQ_GEQ(head_blk.end, start)) {
/*
* Merge this SACK block into head_blk. This SACK
* block itself will be discarded.
*/
if (SEQ_GT(head_blk.start, start))
head_blk.start = start;
if (SEQ_LT(head_blk.end, end))
head_blk.end = end;
} else {
/*
* Save this SACK block.
*/
saved_blks[num_saved].start = start;
saved_blks[num_saved].end = end;
num_saved++;
}
}
/*
* Update SACK list in tp->sackblks[].
*/
num_head = 0;
if (SEQ_GT(head_blk.start, tp->rcv_nxt)) {
/*
* The received data segment is an out-of-order segment. Put
* head_blk at the top of SACK list.
*/
tp->sackblks[0] = head_blk;
num_head = 1;
/*
* If the number of saved SACK blocks exceeds its limit,
* discard the last SACK block.
*/
if (num_saved >= MAX_SACK_BLKS)
num_saved--;
}
if (num_saved > 0) {
/*
* Copy the saved SACK blocks back.
*/
bcopy(saved_blks, &tp->sackblks[num_head],
sizeof(struct sackblk) * num_saved);
}
/* Save the number of SACK blocks. */
tp->rcv_numsacks = num_head + num_saved;
}
/*
* 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;
}
/*
* To insert a new blk to the array of SACK blk in receiver.
*
* Parameters:
* sack_blk_t *head: pointer to the array of SACK blks.
* tcp_seq begin: starting seq num of the new blk.
* tcp_seq end: ending seq num of the new blk.
* int32_t *num: (referenced) total num of SACK blks on the list.
*/
void
tcp_sack_insert(sack_blk_t *head, tcp_seq begin, tcp_seq end, int32_t *num)
{
int32_t i, j, old_num, new_num;
sack_blk_t tmp[MAX_SACK_BLK - 1];
/* The array is empty, just add the new one. */
if (*num == 0) {
head[0].begin = begin;
head[0].end = end;
*num = 1;
return;
}
/*
* Check for overlap. There are five cases.
*
* 1. there is no overlap with any other SACK blks.
* 2. new SACK blk is completely contained in another blk.
* 3. tail part of new SACK blk overlaps with another blk.
* 4. head part of new SACK blk overlaps with another blk.
* 5. new SACK blk completely contains another blk.
*
* Use tmp to hold old SACK blks. After the loop, copy them back
* to head.
*/
old_num = *num;
if (old_num > MAX_SACK_BLK - 1) {
old_num = MAX_SACK_BLK - 1;
}
new_num = old_num;
j = 0;
for (i = 0; i < old_num; i++) {
if (SEQ_LT(end, head[i].begin) || SEQ_GT(begin, head[i].end)) {
/* Case 1: continue to check. */
tmp[j].begin = head[i].begin;
tmp[j].end = head[i].end;
j++;
continue;
} else if (SEQ_GEQ(begin, head[i].begin) &&
SEQ_LEQ(end, head[i].end)) {
/* Case 2: re-insert the old blk to the head. */
begin = head[i].begin;
end = head[i].end;
} else if (SEQ_LEQ(end, head[i].end) &&
SEQ_GEQ(end, head[i].begin)) {
/*
* Case 3: Extend the new blk, remove the old one
* and continue to check.
*/
end = head[i].end;
} else if (SEQ_GEQ(begin, head[i].begin) &&
SEQ_LEQ(begin, head[i].end)) {
/* Case 4 */
begin = head[i].begin;
}
/*
* Common code for all cases except the first one, which
* copies the original SACK blk into the tmp storage. Other
* cases remove the original SACK blk by not copying into
* tmp storage.
*/
new_num--;
}
head[0].begin = begin;
head[0].end = end;
for (i = 0; i < new_num; i++) {
head[i+1].begin = tmp[i].begin;
head[i+1].end = tmp[i].end;
}
*num = new_num + 1;
}
/**
* \brief insert a SACK range
*
* \param le left edge in host order
* \param re right edge in host order
*
* \retval 0 all is good
* \retval -1 error
*/
static int StreamTcpSackInsertRange(TcpStream *stream, uint32_t le, uint32_t re) {
SCLogDebug("le %u, re %u", le, re);
#ifdef DEBUG
StreamTcpSackPrintList(stream);
#endif
/* if to the left of last_ack then ignore */
if (SEQ_LT(re, stream->last_ack)) {
SCLogDebug("too far left. discarding");
goto end;
}
/* if to the right of the tcp window then ignore */
if (SEQ_GT(le, (stream->last_ack + stream->window))) {
SCLogDebug("too far right. discarding");
goto end;
}
if (stream->sack_head != NULL) {
StreamTcpSackRecord *rec;
for (rec = stream->sack_head; rec != NULL; rec = rec->next) {
SCLogDebug("rec %p, le %u, re %u", rec, rec->le, rec->re);
if (SEQ_LT(le, rec->le)) {
SCLogDebug("SEQ_LT(le, rec->le)");
if (SEQ_LT(re, rec->le)) {
SCLogDebug("SEQ_LT(re, rec->le)");
// entirely before, prepend
StreamTcpSackRecord *stsr = StreamTcpSackRecordAlloc();
if (unlikely(stsr == NULL)) {
SCReturnInt(-1);
}
stsr->le = le;
stsr->re = re;
stsr->next = stream->sack_head;
stream->sack_head = stsr;
goto end;
} else if (SEQ_EQ(re, rec->le)) {
SCLogDebug("SEQ_EQ(re, rec->le)");
// starts before, ends on rec->le, expand
rec->le = le;
} else if (SEQ_GT(re, rec->le)) {
SCLogDebug("SEQ_GT(re, rec->le)");
// starts before, ends beyond rec->le
if (SEQ_LEQ(re, rec->re)) {
SCLogDebug("SEQ_LEQ(re, rec->re)");
// ends before rec->re, expand
rec->le = le;
} else { // implied if (re > rec->re)
SCLogDebug("implied if (re > rec->re), le set to %u", rec->re);
le = rec->re;
continue;
}
}
} else if (SEQ_EQ(le, rec->le)) {
SCLogDebug("SEQ_EQ(le, rec->le)");
if (SEQ_LEQ(re, rec->re)) {
SCLogDebug("SEQ_LEQ(re, rec->re)");
// new record fully overlapped
SCReturnInt(0);
} else { // implied re > rec->re
SCLogDebug("implied re > rec->re");
if (rec->next != NULL) {
if (SEQ_LEQ(re, rec->next->le)) {
rec->re = re;
goto end;
} else {
rec->re = rec->next->le;
le = rec->next->le;
SCLogDebug("le is now %u", le);
continue;
}
} else {
rec->re = re;
goto end;
}
}
} else { // implied (le > rec->le)
SCLogDebug("implied (le > rec->le)");
if (SEQ_LT(le, rec->re)) {
SCLogDebug("SEQ_LT(le, rec->re))");
// new record fully overlapped
if (SEQ_GT(re, rec->re)) {
SCLogDebug("SEQ_GT(re, rec->re)");
if (rec->next != NULL) {
if (SEQ_LEQ(re, rec->next->le)) {
rec->re = re;
goto end;
} else {
rec->re = rec->next->le;
le = rec->next->le;
continue;
}
} else {
rec->re = re;
goto end;
}
}
SCLogDebug("new range fully overlapped");
SCReturnInt(0);
} else if (SEQ_EQ(le, rec->re)) {
SCLogDebug("here");
// new record fully overlapped
//int r = StreamTcpSackInsertRange(stream, rec->re+1, re);
//SCReturnInt(r);
le = rec->re;
continue;
} else { /* implied le > rec->re */
SCLogDebug("implied le > rec->re");
if (rec->next == NULL) {
SCLogDebug("rec->next == NULL");
StreamTcpSackRecord *stsr = StreamTcpSackRecordAlloc();
if (unlikely(stsr == NULL)) {
SCReturnInt(-1);
}
stsr->le = le;
stsr->re = re;
stsr->next = NULL;
stream->sack_tail->next = stsr;
stream->sack_tail = stsr;
goto end;
} else {
SCLogDebug("implied rec->next != NULL");
if (SEQ_LT(le, rec->next->le) && SEQ_LT(re, rec->next->le)) {
SCLogDebug("SEQ_LT(le, rec->next->le) && SEQ_LT(re, rec->next->le)");
StreamTcpSackRecord *stsr = StreamTcpSackRecordAlloc();
if (unlikely(stsr == NULL)) {
SCReturnInt(-1);
}
stsr->le = le;
stsr->re = re;
stsr->next = rec->next;
rec->next = stsr;
} else if (SEQ_LT(le, rec->next->le) && SEQ_GEQ(re, rec->next->le)) {
SCLogDebug("SEQ_LT(le, rec->next->le) && SEQ_GEQ(re, rec->next->le)");
StreamTcpSackRecord *stsr = StreamTcpSackRecordAlloc();
if (unlikely(stsr == NULL)) {
SCReturnInt(-1);
}
stsr->le = le;
stsr->re = rec->next->le;
stsr->next = rec->next;
rec->next = stsr;
le = rec->next->le;
}
}
}
}
}
} else {
SCLogDebug("implied empty list");
StreamTcpSackRecord *stsr = StreamTcpSackRecordAlloc();
if (unlikely(stsr == NULL)) {
SCReturnInt(-1);
}
stsr->le = le;
stsr->re = re;
stsr->next = NULL;
stream->sack_head = stsr;
stream->sack_tail = stsr;
}
StreamTcpSackPruneList(stream);
end:
SCReturnInt(0);
}
int tcp_input(struct ifnet * __if, struct iphdr * iph,
struct tcphdr * th, int len)
{
struct tcp_listen_pcb * mux;
struct tcp_pcb * tp;
#if (ENABLE_NET_TCP_CHECKSUM)
unsigned int sum;
#endif
int ti_len;
int acked = 0;
int ourfinisacked = 0;
int needoutput = 0;
unsigned int optlen;
int tiflags;
int todrop;
uint32_t snd_una;
uint32_t snd_nxt;
uint32_t snd_max;
uint32_t ti_seq;
uint32_t ti_ack;
int rcv_wnd;
int tiwin;
int hdrlen;
uint8_t * data;
int ret;
#if (ENABLE_TCPDUMP)
tcp_dump(iph, th, TCPDUMP_RX);
#endif
/* get TCP options, if any */
optlen = ((th->th_off << 2) - sizeof(struct tcphdr));
hdrlen = sizeof(struct tcphdr) + optlen;
data = (uint8_t *)&th->th_opt[optlen];
ti_len = len - hdrlen;
#if (ENABLE_NET_TCP_CHECKSUM)
/* initialize checksum */
sum = htons(len) + (IPPROTO_TCP << 8);
sum = in_chksum(sum, &iph->saddr, 8);
sum = in_chksum(sum, th, hdrlen);
if (ti_len) {
sum = in_chksum(sum, data, ti_len);
}
if (sum != 0x0000ffff) {
DCC_LOG3(LOG_WARNING, "checksum error: 0x%04x hdrlen=%d, len=%d",
sum, hdrlen, len);
TCP_PROTO_STAT_ADD(rx_err, 1);
goto drop;
}
#endif
tiflags = th->th_flags;
/* convert TCP protocol specific fields to host format */
tiwin = ntohs(th->th_win);
ti_seq = ntohl(th->th_seq);
ti_ack = ntohl(th->th_ack);
TCP_PROTO_STAT_ADD(rx_ok, 1);
/* Serch in active list first */
if ((tp = tcp_active_lookup(iph->saddr, th->th_sport,
iph->daddr, th->th_dport)) == NULL) {
/* lookup into listening pcb list */
if ((mux = tcp_listen_lookup(iph->saddr, th->th_sport,
iph->daddr, th->th_dport)) == NULL) {
DCC_LOG(LOG_WARNING, "invalid peer ???");
goto dropwithreset;
}
if ((tiflags & TH_ACK)) {
DCC_LOG(LOG_WARNING, "listen ACK ?");
goto dropwithreset;
}
if (ti_len != 0) {
DCC_LOG(LOG_WARNING, "ti_len != 0");
goto dropwithreset;
}
/* Completion of Passive Open
Ref.: TCP/IP Illustrated Volume 2, pg. 942 */
if (!(tiflags & TH_SYN)) {
DCC_LOG(LOG_WARNING, "listen !SYN ?");
goto drop;
}
/* In the LISTEN state, we check for incoming SYN segments,
creates a new PCB, and responds with a SYN|ACK. */
if ((tiflags & TH_RST)) {
DCC_LOG(LOG_WARNING, "listen RST?");
goto drop;
}
if ((tp = tcp_passive_open(mux, iph, th, optlen)) == NULL) {
DCC_LOG(LOG_WARNING, "tcp_passive_open()");
goto dropwithreset;
}
/* schedule output */
tcp_output_sched(tp);
/* packet handled */
return 0;
}
DCC_LOG1(LOG_MSG, "<%05x> active", (int)tp);
snd_una = tp->snd_seq;
snd_nxt = tp->snd_seq + tp->snd_off;
snd_max = tp->snd_seq + tp->snd_max;
/* Remove acknowledged bytes from the send buffer */
/* Wakeup processes waiting on send buffer */
/* Segment received on a connection.
Reset the idle detection timer
Ref.: TCP/IP Illustrated Volume 2, pg. 932 */
tp->t_conn_tmr = tcp_idle_det_tmo;
if (tp->t_flags & TF_IDLE) {
/* exits from the idle state */
tp->t_flags &= ~TF_IDLE;
DCC_LOG1(LOG_INFO, "<%05x> IDLE exit", (int)tp);
}
#if 0
/* Process options, we don't need to check if the socket is
in the LISTEN state, because only active (non LISTENING) sockets
will actually fall into this code.
XXX: options after connection stablished ???
*/
if (optlen)
tcp_parse_options(tp, th, th->th_opt, optlen);
#endif
/* Ref.: TCP/IP Illustrated Volume 2, pg. 934 */
#if (TCP_ENABLE_HEADER_PREDICTION)
if ((tp->t_state == TCPS_ESTABLISHED) &&
(tiflags & (TH_SYN | TH_FIN | TH_RST | TH_URG | TH_ACK)) == TH_ACK &&
(ti_seq == tp->rcv_nxt) &&
(tiwin) &&
(tiwin == tp->snd_wnd) &&
(snd_nxt == snd_max)) {
if (ti_len == 0) {
if (SEQ_GT(ti_ack, snd_una) &&
SEQ_LEQ(ti_ack, snd_max)) {
acked = ti_ack - snd_una;
DCC_LOG(LOG_INFO, "header prediction, ACK ...");
mbuf_queue_trim(&tp->snd_q, acked);
snd_una = ti_ack;
tp->snd_seq = snd_una;
tp->snd_off = snd_nxt - tp->snd_seq;
tp->snd_max = snd_max - tp->snd_seq;
if (snd_una == snd_max) {
tp->t_rxmt_tmr = 0;
tp->t_rxmt_cnt = 0;
DCC_LOG(LOG_INFO, "acked all data, rxmt tmr stopped");
} else {
if (tp->t_rxmt_tmr == 0) {
DCC_LOG(LOG_INFO,
"not all data acked restart rxmt tmr");
tp->t_rxmt_tmr = tcp_rxmtintvl[tp->t_rxmt_cnt / 2];
}
}
thinkos_cond_broadcast(tp->t_cond);
if (tp->snd_q.len) {
/* schedule output */
tcp_output_sched(tp);
}
return 0;
}
} else {
if ((ti_ack == snd_una) &&
ti_len <= (tcp_maxrcv - tp->rcv_q.len)) {
int len;
DCC_LOG1(LOG_INFO, "header prediction, data (%d)", ti_len);
/* append data */
len = mbuf_queue_add(&tp->rcv_q, data, ti_len);
tp->rcv_nxt += len;
thinkos_cond_broadcast(tp->t_cond);
if (len != ti_len) {
DCC_LOG1(LOG_WARNING, "<%05x> no more mbufs", (int)tp);
tp->t_flags |= TF_ACKNOW;
/* schedule output */
tcp_output_sched(tp);
} else {
tp->t_flags |= TF_DELACK;
}
return 0;
}
}
}
#endif /* TCP_ENABLE_HEADER_PREDICTION */
/* Slow path input processing
Ref.: TCP/IP Illustrated Volume 2, pg. 941 */
/* TODO: Drop TCP, IP headers and TCP options.
Well, only if these structures were dynamic allocated... */
if (ti_len == 0) {
DCC_LOG(LOG_INFO, "slow path ACK");
} else {
DCC_LOG1(LOG_INFO, "slow path (%d)", ti_len);
}
/* Calculate the 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 advertise window.
Ref.: TCP/IP Illustrated Volume 2, pg. 941 */
{
int win;
/* space left in the input queue */
win = tcp_maxrcv - tp->rcv_q.len;
if (win <= 0) {
win = 0;
DCC_LOG(LOG_INFO, "receive buffer full!");
}
// rcv_wnd = MAX(win, tp->rcv_adv_wnd);
rcv_wnd = win;
DCC_LOG3(LOG_INFO, "adv_wnd=%d rcv_wnd=%d win=%d",
tp->rcv_adv_wnd, rcv_wnd, win);
}
if (tp->t_state == TCPS_SYN_SENT) {
/* response to an active open.
Ref.: TCP/IP Illustrated Volume 2, pg. 947 */
/* Common proccessing for receipt of SYN.
Ref.: TCP/IP Illustrated Volume 2, pg. 950 */
if ((tiflags & TH_RST)) {
goto close;
}
if (!(tiflags & TH_SYN)) {
DCC_LOG(LOG_WARNING, "SYN_SENT SYN ?");
/* TODO: reset */
goto close_and_reset;
}
if (!(tiflags & TH_ACK)) {
DCC_LOG(LOG_WARNING, "SYN_SENT ACK ?");
/* TODO: reset */
goto close_and_reset;
}
if (ti_len != 0) {
DCC_LOG(LOG_WARNING, "ti_len != 0");
/* TODO: reset */
goto close_and_reset;
}
/* update the send sequence */
tp->snd_seq++;
if (tp->snd_seq != ti_ack) {
DCC_LOG3(LOG_WARNING, "<%05x> tp->snd_seq(%d) != ti_ack(%d)",
(int)tp, tp->snd_seq, ti_ack);
/* TODO: reset */
goto close_and_reset;
}
tp->snd_off--;
tp->snd_max--;
// tp->snd_off = 0;
// tp->snd_max = 0;
if (optlen)
tcp_parse_options(tp, th, th->th_opt, optlen);
/* Advance tp->ti_seq to correspond to first data byte. */
ti_seq++;
if (ti_len > rcv_wnd) {
DCC_LOG3(LOG_WARNING, "<%05x> ti_len(%d) > rcv_wnd(%d)",
(int)tp, ti_len, rcv_wnd);
/* TODO: if data, trim to stay within window. */
ti_len = rcv_wnd;
}
/* update the sequence number */
tp->rcv_nxt = ti_seq;
/* update the window size */
tp->snd_wnd = ntohs(th->th_win);
tp->t_state = TCPS_ESTABLISHED;
DCC_LOG1(LOG_INFO, "<%05x> [ESTABLISHED]", (int)tp);
/* TODO: initialization of receive urgent pointer
tcp->rcv_up = ti_seq; */
/* XXX: */
tp->t_flags |= TF_ACKNOW;
thinkos_cond_broadcast(tp->t_cond);
goto step6;
close_and_reset:
tp->t_state = TCPS_CLOSED;
pcb_move((struct pcb *)tp, &__tcp__.active, &__tcp__.closed);
DCC_LOG1(LOG_INFO, "<%05x> [CLOSED]", (int)tp);
/* XXX: discard the data */
mbuf_queue_free(&tp->snd_q);
mbuf_queue_free(&tp->rcv_q);
/* notify the upper layer */
thinkos_cond_broadcast(tp->t_cond);
goto dropwithreset;
}
/* 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 ti_ack. */
/* Trim Segment so Data is Within Window
Ref.: TCP/IP Illustrated Volume 2, pg. 954 */
todrop = tp->rcv_nxt - ti_seq;
if (todrop > 0) {
if (tiflags & TH_SYN) {
DCC_LOG(LOG_INFO, "SYN");
tiflags &= ~TH_SYN;
ti_seq++;
todrop--;
}
if ((todrop > ti_len) ||
((todrop == ti_len) && ((tiflags & TH_FIN) == 0))) {
tiflags &= ~TH_FIN;
tp->t_flags |= TF_ACKNOW;
todrop = ti_len;
}
DCC_LOG4(LOG_WARNING, "<%05x> drop: len=%d drop=%d rem=%d!",
(int)tp, ti_len, todrop, ti_len - todrop);
/* adjust the data pointer */
data += todrop;
ti_seq += todrop;
ti_len -= todrop;
/* TODO: adjust the urgent pointer */
}
/* FIXME: only reset the connection if there are no more
application to handle the incomming data, half-close */
if ((tp->t_state > TCPS_FIN_WAIT_1) && (ti_len)) {
DCC_LOG1(LOG_INFO, "<%05x> segment received after FIN", (int)tp);
/* TODO: stat */
goto dropwithreset;
}
/* If segment ends after window, drop trailing data
and (PUSH and FIN); if nothing left, just ACK.
Ref.: TCP/IP Illustrated Volume 2, pg. 958 */
todrop = (ti_seq + ti_len) - (tp->rcv_nxt + rcv_wnd);
DCC_LOG4(LOG_INFO, "ti_seq=%u ti_len=%d rcv_nxt=%u rcv_wnd=%d",
ti_seq, ti_len, tp->rcv_nxt, rcv_wnd);
/* */
if (todrop > 0) {
// TCP_LOG(tp, "tcp_input: trailing data drop");
if (todrop >= ti_len) {
/*
* If a new connection request is received
* while in TIME_WAIT, drop the old connection ...
* Ref.: TCP/IP Illustrated Volume 2, pg. 958
if ((tiflags & TH_SYN) && (tp->t_state == TCPS_TIMEWAIT) &&
(SEQ_GT(ti_seq, tp->rcv_nxt))) {
__tcp__.iss += tcp_issincr;
tcp_rst(tp);
goto findpcb;
} */
if ((rcv_wnd == 0) && (ti_seq == tp->rcv_nxt)) {
tp->t_flags |= TF_ACKNOW;
} else
goto dropafterack;
}
DCC_LOG2(LOG_WARNING, "<%05x> data drop: %d!", (int)tp, todrop);
ti_len -= todrop;
tiflags &= ~(TH_PSH | TH_FIN);
}
/* If the RST bit is set eximine the state: ...
Ref.: TCP/IP Illustrated Volume 2, pg. 964 */
if ((tiflags & TH_RST)) {
DCC_LOG1(LOG_WARNING, "<%05x> RST received", (int)tp);
switch(tp->t_state) {
case TCPS_SYN_RCVD:
// tp->errno = ECONNREFUSED;
goto close;
case TCPS_ESTABLISHED:
case TCPS_CLOSE_WAIT:
// tp->errno = ECONNRESET;
close:
/* discard the data */
mbuf_queue_free(&tp->snd_q);
mbuf_queue_free(&tp->rcv_q);
tp->t_state = TCPS_CLOSED;
pcb_move((struct pcb *)tp, &__tcp__.active, &__tcp__.closed);
DCC_LOG1(LOG_INFO, "<%05x> [CLOSED]", (int)tp);
/* notify the upper layer */
thinkos_cond_broadcast(tp->t_cond);
/* PCBs in the close state should be cleared by the application */
goto drop;
case TCPS_FIN_WAIT_1:
case TCPS_FIN_WAIT_2:
case TCPS_CLOSING:
case TCPS_LAST_ACK:
case TCPS_TIME_WAIT:
/* Our side was already closed */
tcp_pcb_free(tp);
goto drop;
}
}
/* If a SYN is in the window, then this is an
error and we send an RST and drop the connection.
Ref.: TCP/IP Illustrated Volume 2, pg. 965 */
if ((tiflags & TH_SYN)) {
DCC_LOG1(LOG_WARNING, "<%05x> the SYN bit is set inside the window",
(int)tp);
goto dropwithreset;
}
/* If the ACK bit is off we drop the segment and return. */
if ((!(tiflags & TH_ACK))) {
DCC_LOG1(LOG_WARNING, "<%05x> the ACK bit is off", (int)tp);
goto drop;
}
/*
* ACK processing.
* Ref.: TCP/IP Illustrated Volume 2, pg. 969
*
*/
DCC_LOG4(LOG_INFO, "ack=%u una=%u nxt=%u max=%u",
ti_ack, snd_una, snd_nxt, snd_max);
switch(tp->t_state) {
case TCPS_SYN_RCVD:
if (SEQ_GT(snd_una, ti_ack) ||
SEQ_GT(ti_ack, snd_max)) {
DCC_LOG1(LOG_WARNING,
"<%05x> ti_ack < snd_una || snd_max < ti_ack",
(int)tp);
goto dropwithreset;
}
tp->t_state = TCPS_ESTABLISHED;
tp->snd_off--;
tp->snd_max--;
DCC_LOG1(LOG_INFO, "<%05x> SYN ackd [ESTABLISHED]", (int)tp);
/* notify the upper layer*/
// thinkos_cond_signal(tp->t_cond);
/* TODO: tcp reassembly
tcp_reass(tp); */
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:
/* TODO: tcp reassembly
tcp_reass(tp); */
if (SEQ_LEQ(ti_ack, snd_una)) {
/* TODO: check for completly duplicated ACKs.
Ref.: TCP/IP Illustrated Volume 2, pg. 971 */
if ((ti_len == 0) && (tiwin == tp->snd_wnd)) {
if ((tp->t_rxmt_tmr == 0) || ti_ack != snd_una) {
// dupacks = 0;
} else {
DCC_LOG2(LOG_INFO, "duplicated ACK. ti_ack=%u snd_una=%u",
ti_ack, snd_una);
}
} else {
// dupacks = 0;
}
break;
}
/* Check out of range ACK */
/* Ref.: TCP/IP Illustrated Volume 2, pg. 974 */
if (SEQ_GT(ti_ack, snd_max)) {
/* TODO:
tcpstat.tcps_rcvacktoomuch++;
*/
DCC_LOG3(LOG_WARNING, "(%04x) out of range ACK. "
"th_ack=%u > snd_max=%u !",
(int)tp, ti_ack, snd_max);
goto dropafterack;
}
acked = ti_ack - snd_una;
/* TODO:
tcpstat.tcps_rcvackpack++;
tcpstat.tcps_rcvackbyte += acked;
*/
DCC_LOG1(LOG_INFO, "acked=%d", acked);
/* If all outstanding data is acked, stop retransmit timer else
restarts it ....
Ref.: TCP/IP Illustrated Volume 2, pg. 976 */
if (ti_ack == snd_max) {
tp->t_rxmt_tmr = 0;
tp->t_rxmt_cnt = 0;
needoutput = 1;
DCC_LOG(LOG_INFO, "acked all data, rxmt tmr stopped");
} else {
/* TODO: peristent timer */
// if (tp->t_persist_tmr == 0) {
DCC_LOG(LOG_INFO, "not all data acked restart rxmt tmr");
tp->t_rxmt_tmr = tcp_rxmtintvl[tp->t_rxmt_cnt / 2];
// }
}
/* TODO:
tcpstat.tcps_rcvackpack++;
tcpstat.tcps_rcvackbyte += acked;
*/
/* TODO: remove acknowledged data from send buffer
Ref.: TCP/IP Illustrated Volume 2, pg. 978 */
/* FIXME: send buffer bytes count */
if (acked > tp->snd_q.len) {
mbuf_queue_trim(&tp->snd_q, tp->snd_q.len);
ourfinisacked = 1;
} else {
/* TODO: estimate the send window */
mbuf_queue_trim(&tp->snd_q, acked);
ourfinisacked = 0;
}
/* awaken a thread waiting on the send buffer ... */
thinkos_cond_broadcast(tp->t_cond);
snd_una = ti_ack;
if (SEQ_LT(snd_nxt, snd_una)) {
snd_nxt = snd_una;
}
tp->snd_seq = snd_una;
tp->snd_off = snd_nxt - tp->snd_seq;
tp->snd_max = snd_max - tp->snd_seq;
DCC_LOG4(LOG_INFO, "<%05x> snd_seq=%u snd_max=%u snd_q.len=%d",
(int)tp, tp->snd_seq, snd_max, tp->snd_q.len);
switch(tp->t_state) {
case TCPS_FIN_WAIT_1:
if (ourfinisacked) {
/* FIXME: If we can't receive any more data..
Ref.: TCP/IP Illustrated Volume 2, pg. 979 */
tp->t_conn_tmr = 4 * tcp_msl;
tp->t_state = TCPS_FIN_WAIT_2;
DCC_LOG1(LOG_INFO, "<%05x> [FIN_WAIT_2]", (int)tp);
}
break;
case TCPS_CLOSING:
if (ourfinisacked) {
mbuf_queue_free(&tp->snd_q);
mbuf_queue_free(&tp->rcv_q);
tp->t_state = TCPS_TIME_WAIT;
DCC_LOG1(LOG_INFO, "<%05x> [TIME_WAIT]", (int)tp);
tp->t_rxmt_tmr = 0;
tp->t_conn_tmr = 2 * tcp_msl;
DCC_LOG1(LOG_INFO, "stop rxmt tmr, start 2MSL tmr: %d",
tp->t_conn_tmr);
}
break;
case TCPS_LAST_ACK:
if (ourfinisacked) {
tcp_pcb_free(tp);
goto drop;
}
break;
case TCPS_TIME_WAIT:
/* restart the finack timer
Ref.: TCP/IP Illustrated Volume 2, pg. 981 */
tp->t_conn_tmr = 2 * tcp_msl;
goto dropafterack;
}
break;
}
DCC_LOG4(LOG_INFO, "<%05x> recvd=%d acked=%d rcv_q.len=%d", (int)tp,
ti_len, acked, tp->rcv_q.len);
step6:
/* Update window information
Ref.: TCP/IP Illustrated Volume 2, pg. 982 */
DCC_LOG(LOG_MSG, "setp6");
// if ((tiflags & TH_ACK) && (tiwin > tp->snd_wnd)) {
if ((tiflags & TH_ACK) && (tiwin != tp->snd_wnd)) {
/* Keep track of pure window updates */
/* TODO: TCP Statistics */
/* TODO: Update window information */
DCC_LOG1(LOG_INFO, "window update, win=%d", tiwin);
tp->snd_wnd = tiwin;
needoutput = 1;
}
/* TODO: Urgent mode processing */
/* Process the segment text,
merging it into the TCP sequencing queue,
dodata:
...
Ref.: TCP/IP Illustrated Volume 2, pg. 988 */
if ((ti_len || (tiflags & TH_FIN)) &&
TCPS_HAVERCVDFIN(tp->t_state) == 0) {
if ((ti_seq == tp->rcv_nxt) && (tp->t_state == TCPS_ESTABLISHED)) {
/* append data */
int n;
tp->t_flags |= TF_DELACK;
n = mbuf_queue_add(&tp->rcv_q, data, ti_len);
if (n != ti_len) {
DCC_LOG2(LOG_WARNING, "no more mbufs, %d != %d", n, ti_len);
}
ti_len = n;
tp->rcv_nxt += ti_len;
/* TODO: statistics */
tiflags &= TH_FIN;
// if (tp->rcv_q.len == ti_len) {
// DCC_LOG3(LOG_INFO, "<%05x> rcvd %d, signaling %d ...",
// (int)tp, ti_len, tp->t_cond);
/*
* notify the upper layer of the data arrival...
*/
thinkos_cond_signal(tp->t_cond);
// } else {
// DCC_LOG2(LOG_INFO, "<%05x> rcvd %d", (int)tp, ti_len);
// }
} else {
/* TODO: half-close */
/* TODO: reassembly */
// m = mlink_free(m);
if (tp->t_state == TCPS_ESTABLISHED) {
// DCC_LOG(LOG_WARNING, "out of order, drop!");
DCC_LOG(LOG_WARNING, "out of order, drop");
TCP_PROTO_STAT_ADD(rx_drop, 1);
}
tp->t_flags |= TF_ACKNOW;
}
} else {
DCC_LOG(LOG_INFO, "!!!!!!!!!");
tiflags &= ~TH_FIN;
}
/* FIN Processing */
if (tiflags & TH_FIN) {
if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
tp->t_flags |= TF_ACKNOW;
tp->rcv_nxt++;
}
switch(tp->t_state) {
case TCPS_SYN_RCVD:
case TCPS_ESTABLISHED:
tp->t_state = TCPS_CLOSE_WAIT;
DCC_LOG1(LOG_INFO, "<%05x> [CLOSE_WAIT]", (int)tp);
/* notify the application that our peer
has closed its side. Sockets: marks
the socket as write-only */
if (tp->rcv_q.len == 0) {
thinkos_cond_broadcast(tp->t_cond);
}
break;
case TCPS_FIN_WAIT_1:
tp->t_state = TCPS_CLOSING;
DCC_LOG1(LOG_INFO, "<%05x> [CLOSING]", (int)tp);
break;
case TCPS_FIN_WAIT_2:
mbuf_queue_free(&tp->rcv_q);
mbuf_queue_free(&tp->snd_q);
tp->t_state = TCPS_TIME_WAIT;
DCC_LOG1(LOG_INFO, "<%05x> [TIME_WAIT]", (int)tp);
tp->t_rxmt_tmr = 0;
tp->t_conn_tmr = 2 * tcp_msl;
DCC_LOG1(LOG_INFO, "stop rxmt tmr, start 2MSL tmr: %d",
tp->t_conn_tmr);
break;
case TCPS_TIME_WAIT:
/* restart the counter */
tp->t_conn_tmr = 2 * tcp_msl;
break;
}
}
/* Final Processing */
if (needoutput || (tp->t_flags & TF_ACKNOW)) {
if (needoutput) {
DCC_LOG(LOG_INFO, "needoutput, call tcp_out.");
}
if (tp->t_flags & TF_ACKNOW) {
DCC_LOG(LOG_INFO, "ACKNOW set, call tcp_out.");
}
/* schedule output */
tcp_output_sched(tp);
}
return 0;
dropafterack:
DCC_LOG1(LOG_INFO, "<%05x> drop and ACK", (int)tp);
if (tiflags & TH_RST)
goto drop;
tp->t_flags |= TF_ACKNOW;
/* schedule output */
tcp_output_sched(tp);
return 0;
dropwithreset:
DCC_LOG1(LOG_TRACE, "<%05x> drop and RST", (int)tp);
ret = 0;
/* TODO: check for a broadcast/multicast */
if (!(tiflags & TH_RST)) {
if (tiflags & TH_ACK) {
ret = tcp_respond(iph, th, 0, ti_ack, TH_RST);
} else if (tiflags & TH_SYN) {
ti_len++;
ret = tcp_respond(iph, th, ti_seq + ti_len, 0, TH_ACK | TH_RST);
}
}
TCP_PROTO_STAT_ADD(rx_drop, 1);
return ret;
drop:
DCC_LOG(LOG_TRACE, "drop");
TCP_PROTO_STAT_ADD(rx_drop, 1);
return 0;
}