/*
* After a timeout, the SACK list may be rebuilt. This SACK information
* should be used to avoid retransmitting SACKed data. This function
* traverses the SACK list to see if snd_nxt should be moved forward.
*/
void
tcp_sack_adjust(struct tcpcb *tp)
{
struct sackhole *p, *cur = TAILQ_FIRST(&tp->snd_holes);
//ScenSim-Port// INP_WLOCK_ASSERT(tp->t_inpcb);
if (cur == NULL)
return; /* No holes */
if (SEQ_GEQ(tp->snd_nxt, tp->snd_fack))
return; /* We're already beyond any SACKed blocks */
/*-
* Two cases for which we want to advance snd_nxt:
* i) snd_nxt lies between end of one hole and beginning of another
* ii) snd_nxt lies between end of last hole and snd_fack
*/
while ((p = TAILQ_NEXT(cur, scblink)) != NULL) {
if (SEQ_LT(tp->snd_nxt, cur->end))
return;
if (SEQ_GEQ(tp->snd_nxt, p->start))
cur = p;
else {
tp->snd_nxt = p->start;
return;
}
}
if (SEQ_LT(tp->snd_nxt, cur->end))
return;
tp->snd_nxt = tp->snd_fack;
}
static void
dctcp_ack_received(struct cc_var *ccv, uint16_t type)
{
struct dctcp *dctcp_data;
int bytes_acked = 0;
dctcp_data = ccv->cc_data;
if (CCV(ccv, t_flags) & TF_ECN_PERMIT) {
/*
* DCTCP doesn't treat receipt of ECN marked packet as a
* congestion event. Thus, DCTCP always executes the ACK
* processing out of congestion recovery.
*/
if (IN_CONGRECOVERY(CCV(ccv, t_flags))) {
EXIT_CONGRECOVERY(CCV(ccv, t_flags));
newreno_cc_algo.ack_received(ccv, type);
ENTER_CONGRECOVERY(CCV(ccv, t_flags));
} else
newreno_cc_algo.ack_received(ccv, type);
if (type == CC_DUPACK)
bytes_acked = CCV(ccv, t_maxseg);
if (type == CC_ACK)
bytes_acked = ccv->bytes_this_ack;
/* Update total bytes. */
dctcp_data->bytes_total += bytes_acked;
/* Update total marked bytes. */
if (dctcp_data->ece_curr) {
if (!dctcp_data->ece_prev
&& bytes_acked > CCV(ccv, t_maxseg)) {
dctcp_data->bytes_ecn +=
(bytes_acked - CCV(ccv, t_maxseg));
} else
dctcp_data->bytes_ecn += bytes_acked;
dctcp_data->ece_prev = 1;
} else {
if (dctcp_data->ece_prev
&& bytes_acked > CCV(ccv, t_maxseg))
dctcp_data->bytes_ecn += CCV(ccv, t_maxseg);
dctcp_data->ece_prev = 0;
}
dctcp_data->ece_curr = 0;
/*
* Update the fraction of marked bytes at the end of
* current window size.
*/
if ((IN_FASTRECOVERY(CCV(ccv, t_flags)) &&
SEQ_GEQ(ccv->curack, CCV(ccv, snd_recover))) ||
(!IN_FASTRECOVERY(CCV(ccv, t_flags)) &&
SEQ_GT(ccv->curack, dctcp_data->save_sndnxt)))
dctcp_update_alpha(ccv);
} else
newreno_cc_algo.ack_received(ccv, type);
}
/*
* To remove a SACK block.
*
* Parameters:
* sack_blk_t *head: pointer to the array of SACK blks.
* tcp_seq end: to remove all sack blk with seq num less than end.
* int32_t *num: (referenced) total num of SACK blks in the array.
*/
void
tcp_sack_remove(sack_blk_t *head, tcp_seq end, int32_t *num)
{
sack_blk_t tmp[MAX_SACK_BLK];
int32_t i, j, old_num, new_num;
if (*num == 0)
return;
old_num = *num;
new_num = old_num;
j = 0;
/* Walk thru the whole list and copy the new list to tmp[]. */
for (i = 0; i < old_num; i++) {
if (SEQ_GT(end, head[i].begin)) {
/*
* Check to see if the old SACK blk needs to be
* removed or updated. If the old blk is just
* partially covered, update begin and continue.
* If the old blk is completely covered, remove it
* and continue to check.
*/
if (SEQ_GEQ(end, head[i].end)) {
new_num--;
continue;
} else {
tmp[j].begin = end;
tmp[j].end = head[i].end;
}
} else {
tmp[j].begin = head[i].begin;
tmp[j].end = head[i].end;
}
j++;
}
/* Copy tmp[] back to the original list. */
for (i = 0; i < new_num; i++) {
head[i].begin = tmp[i].begin;
head[i].end = tmp[i].end;
}
*num = new_num;
}
/**
* \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);
}
/*
* USTAT PDU / SOS_READY Processor
*
* Arguments:
* sop pointer to sscop connection block
* m pointer to PDU buffer (without trailer)
* trlr pointer to PDU trailer
*
* Returns:
* none
*
*/
void
sscop_ustat_ready(struct sscop *sop, KBuffer *m, caddr_t trlr)
{
struct ustat_pdu *up = (struct ustat_pdu *)trlr;
struct pdu_hdr *php;
sscop_seq seq1, seq2;
up->ustat_nmr = ntohl(up->ustat_nmr);
up->ustat_nr = ntohl(up->ustat_nr);
/*
* Validate peer's current receive data sequence number
*/
if (SEQ_GT(sop->so_ack, up->ustat_nr, sop->so_ack) ||
SEQ_GEQ(up->ustat_nr, sop->so_send, sop->so_ack)) {
/*
* Bad data sequence number
*/
goto goterr;
}
/*
* Free acknowledged PDUs
*/
for (seq1 = sop->so_ack, SEQ_SET(seq2, up->ustat_nr);
SEQ_LT(seq1, seq2, sop->so_ack);
SEQ_INCR(seq1, 1)) {
sscop_pack_free(sop, seq1);
}
/*
* Update transmit state variables
*/
sop->so_ack = seq2;
SEQ_SET(sop->so_sendmax, up->ustat_nmr);
/*
* Get USTAT list elements
*/
SEQ_SET(seq1, ntohl(up->ustat_le1));
SEQ_SET(seq2, ntohl(up->ustat_le2));
/*
* Validate elements
*/
if (SEQ_GT(sop->so_ack, seq1, sop->so_ack) ||
SEQ_GEQ(seq1, seq2, sop->so_ack) ||
SEQ_GEQ(seq2, sop->so_send, sop->so_ack)) {
/*
* Bad element sequence number
*/
goto goterr;
}
/*
* Process each missing sequence number in this gap
*/
while (SEQ_LT(seq1, seq2, sop->so_ack)) {
/*
* Find corresponding SD PDU on pending ack queue
*/
php = sscop_pack_locate(sop, seq1);
if (php == NULL) {
goto goterr;
}
/*
* Retransmit this SD PDU if it's not
* already scheduled for retranmission.
*/
if ((php->ph_rexmit_lk == NULL) &&
(sop->so_rexmit_tl != php)) {
/*
* Put PDU on retransmit queue and schedule
* transmit servicing
*/
sscop_rexmit_insert(sop, php);
sop->so_flags |= SOF_XMITSRVC;
}
/*
* Bump to next sequence number
*/
SEQ_INCR(seq1, 1);
}
/*
* Report retransmitted PDUs
*/
sscop_maa_error(sop, 'V');
/*
* Free PDU buffer chain
*/
KB_FREEALL(m);
/*
* See if transmit queues need servicing
*/
if (sop->so_flags & SOF_XMITSRVC)
sscop_service_xmit(sop);
return;
goterr:
/*
* Protocol/parameter error encountered
*/
sscop_maa_error(sop, 'T');
/*
* Free PDU buffer chain
*/
KB_FREEALL(m);
if (sop->so_vers == SSCOP_VERS_QSAAL)
/*
* Reestablish a new connection
*/
qsaal1_reestablish(sop);
else
/*
* Initiate error recovery
*/
q2110_error_recovery(sop);
return;
}
/*
* STAT PDU / SOS_READY Processor
*
* Arguments:
* sop pointer to sscop connection block
* m pointer to PDU buffer (without trailer)
* trlr pointer to PDU trailer
*
* Returns:
* none
*
*/
void
sscop_stat_ready(struct sscop *sop, KBuffer *m, caddr_t trlr)
{
struct stat_pdu *sp = (struct stat_pdu *)trlr;
struct pdu_hdr *php;
KBuffer *m0 = m;
sscop_seq seq1, seq2, opa;
int cnt = 0;
sp->stat_nps = ntohl(sp->stat_nps);
sp->stat_nmr = ntohl(sp->stat_nmr);
sp->stat_nr = ntohl(sp->stat_nr);
/*
* Validate peer's received poll sequence number
*/
if (SEQ_GT(sop->so_pollack, sp->stat_nps, sop->so_pollack) ||
SEQ_GT(sp->stat_nps, sop->so_pollsend, sop->so_pollack)) {
/*
* Bad poll sequence number
*/
sscop_maa_error(sop, 'R');
goto goterr;
}
/*
* Validate peer's current receive data sequence number
*/
if (SEQ_GT(sop->so_ack, sp->stat_nr, sop->so_ack) ||
SEQ_GT(sp->stat_nr, sop->so_send, sop->so_ack)) {
/*
* Bad data sequence number
*/
sscop_maa_error(sop, 'S');
goto goterr;
}
/*
* Free acknowledged PDUs
*/
for (seq1 = sop->so_ack, SEQ_SET(seq2, sp->stat_nr);
SEQ_LT(seq1, seq2, sop->so_ack);
SEQ_INCR(seq1, 1)) {
sscop_pack_free(sop, seq1);
}
/*
* Update transmit state variables
*/
opa = sop->so_pollack;
sop->so_ack = seq2;
SEQ_SET(sop->so_pollack, sp->stat_nps);
SEQ_SET(sop->so_sendmax, sp->stat_nmr);
/*
* Get first element in STAT list
*/
while (m && (KB_LEN(m) == 0))
m = KB_NEXT(m);
if (m == NULL)
goto done;
m = sscop_stat_getelem(m, &seq1);
/*
* Make sure there's a second element too
*/
if (m == NULL)
goto done;
/*
* Validate first element (start of missing pdus)
*/
if (SEQ_GT(sop->so_ack, seq1, sop->so_ack) ||
SEQ_GEQ(seq1, sop->so_send, sop->so_ack)) {
/*
* Bad element sequence number
*/
sscop_maa_error(sop, 'S');
goto goterr;
}
/*
* Loop thru all STAT elements in list
*/
while (m) {
/*
* Get next even element (start of received pdus)
*/
m = sscop_stat_getelem(m, &seq2);
/*
* Validate seqence number
*/
if (SEQ_GEQ(seq1, seq2, sop->so_ack) ||
SEQ_GT(seq2, sop->so_send, sop->so_ack)) {
/*
* Bad element sequence number
*/
sscop_maa_error(sop, 'S');
goto goterr;
}
/*
* Process each missing sequence number in this gap
*/
while (SEQ_LT(seq1, seq2, sop->so_ack)) {
/*
* Find corresponding SD PDU on pending ack queue
*/
php = sscop_pack_locate(sop, seq1);
if (php == NULL) {
sscop_maa_error(sop, 'S');
goto goterr;
}
/*
* Retransmit this SD PDU only if it was last sent
* during an earlier poll sequence and it's not
* already scheduled for retranmission.
*/
if (SEQ_LT(php->ph_nps, sp->stat_nps, opa) &&
(php->ph_rexmit_lk == NULL) &&
(sop->so_rexmit_tl != php)) {
/*
* Put PDU on retransmit queue and schedule
* transmit servicing
*/
sscop_rexmit_insert(sop, php);
sop->so_flags |= SOF_XMITSRVC;
cnt++;
}
/*
* Bump to next sequence number
*/
SEQ_INCR(seq1, 1);
}
/*
* Now process series of acknowledged PDUs
*
* Get next odd element (start of missing pdus),
* but make sure there is one and that it's valid
*/
if (m == NULL)
goto done;
m = sscop_stat_getelem(m, &seq2);
if (SEQ_GEQ(seq1, seq2, sop->so_ack) ||
SEQ_GT(seq2, sop->so_send, sop->so_ack)) {
/*
* Bad element sequence number
*/
sscop_maa_error(sop, 'S');
goto goterr;
}
/*
* Process each acked sequence number
*/
while (SEQ_LT(seq1, seq2, sop->so_ack)) {
/*
* Can we clear transmit buffers ??
*/
if ((sop->so_flags & SOF_NOCLRBUF) == 0) {
/*
* Yes, free acked buffers
*/
sscop_pack_free(sop, seq1);
}
/*
* Bump to next sequence number
*/
SEQ_INCR(seq1, 1);
}
}
done:
/*
* Free PDU buffer chain
*/
KB_FREEALL(m0);
/*
* Report retransmitted PDUs
*/
if (cnt)
sscop_maa_error(sop, 'V');
/*
* Record transmit window closed transitions
*/
if (SEQ_LT(sop->so_send, sop->so_sendmax, sop->so_ack)) {
if (sop->so_flags & SOF_NOCREDIT) {
sop->so_flags &= ~SOF_NOCREDIT;
sscop_maa_error(sop, 'X');
}
} else {
if ((sop->so_flags & SOF_NOCREDIT) == 0) {
sop->so_flags |= SOF_NOCREDIT;
sscop_maa_error(sop, 'W');
}
}
if (sop->so_vers == SSCOP_VERS_QSAAL)
/*
* Restart lost poll/stat timer
*/
sop->so_timer[SSCOP_T_NORESP] = sop->so_parm.sp_timeresp;
else {
/*
* Determine new polling phase
*/
if ((sop->so_timer[SSCOP_T_POLL] != 0) &&
((sop->so_flags & SOF_KEEPALIVE) == 0)) {
/*
* Remain in active phase - reset NO-RESPONSE timer
*/
sop->so_timer[SSCOP_T_NORESP] =
sop->so_parm.sp_timeresp;
} else if (sop->so_timer[SSCOP_T_IDLE] == 0) {
/*
* Go from transient to idle phase
*/
sop->so_timer[SSCOP_T_POLL] = 0;
sop->so_flags &= ~SOF_KEEPALIVE;
sop->so_timer[SSCOP_T_NORESP] = 0;
sop->so_timer[SSCOP_T_IDLE] = sop->so_parm.sp_timeidle;
}
}
/*
* See if transmit queues need servicing
*/
if (sop->so_flags & SOF_XMITSRVC)
sscop_service_xmit(sop);
return;
goterr:
/*
* Protocol/parameter error encountered
*/
/*
* Free PDU buffer chain
*/
KB_FREEALL(m0);
if (sop->so_vers == SSCOP_VERS_QSAAL)
/*
* Reestablish a new connection
*/
qsaal1_reestablish(sop);
else
/*
* Initiate error recovery
*/
q2110_error_recovery(sop);
return;
}
/*
* When a new ack with SACK is received, check if it indicates packet
* reordering. If there is packet reordering, the socket is marked and
* the late time offset by which the packet was reordered with
* respect to its closest neighboring packets is computed.
*/
static void
tcp_sack_detect_reordering(struct tcpcb *tp, struct sackhole *s,
tcp_seq sacked_seq, tcp_seq snd_fack)
{
int32_t rext = 0, reordered = 0;
/*
* If the SACK hole is past snd_fack, this is from new SACK
* information, so we can ignore it.
*/
if (SEQ_GT(s->end, snd_fack))
return;
/*
* If there has been a retransmit timeout, then the timestamp on
* the SACK segment will be newer. This might lead to a
* false-positive. Avoid re-ordering detection in this case.
*/
if (tp->t_rxtshift > 0)
return;
/*
* Detect reordering from SACK information by checking
* if recently sacked data was never retransmitted from this hole.
*/
if (SEQ_LT(s->rxmit, sacked_seq)) {
reordered = 1;
tcpstat.tcps_avoid_rxmt++;
}
if (reordered) {
if (tcp_detect_reordering == 1 &&
!(tp->t_flagsext & TF_PKTS_REORDERED)) {
tp->t_flagsext |= TF_PKTS_REORDERED;
tcpstat.tcps_detect_reordering++;
}
tcpstat.tcps_reordered_pkts++;
tp->t_reordered_pkts++;
/*
* If reordering is seen on a connection wth ECN enabled,
* increment the heuristic
*/
if (TCP_ECN_ENABLED(tp)) {
INP_INC_IFNET_STAT(tp->t_inpcb, ecn_fallback_reorder);
tcpstat.tcps_ecn_fallback_reorder++;
tcp_heuristic_ecn_aggressive(tp);
}
VERIFY(SEQ_GEQ(snd_fack, s->rxmit));
if (s->rxmit_start > 0) {
rext = timer_diff(tcp_now, 0, s->rxmit_start, 0);
if (rext < 0)
return;
/*
* We take the maximum reorder window to schedule
* DELAYFR timer as that will take care of jitter
* on the network path.
*
* Computing average and standard deviation seems
* to cause unnecessary retransmissions when there
* is high jitter.
*
* We set a maximum of SRTT/2 and a minimum of
* 10 ms on the reorder window.
*/
tp->t_reorderwin = max(tp->t_reorderwin, rext);
tp->t_reorderwin = min(tp->t_reorderwin,
(tp->t_srtt >> (TCP_RTT_SHIFT - 1)));
tp->t_reorderwin = max(tp->t_reorderwin, 10);
}
}
/*
* 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 */
}
/*
* Tcp output routine: figure out what should be sent and send it.
*/
int
tcp_output(struct tcpcb *tp)
{
struct socket *so = tp->t_inpcb->inp_socket;
long len, recwin, sendwin;
int off, flags, error;
#ifdef TCP_SIGNATURE
int sigoff = 0;
#endif
struct mbuf *m;
struct ip *ip = NULL;
struct ipovly *ipov = NULL;
struct tcphdr *th;
u_char opt[TCP_MAXOLEN];
unsigned ipoptlen, optlen, hdrlen;
int idle, sendalot;
int i, sack_rxmit;
int sack_bytes_rxmt;
struct sackhole *p;
#if 0
int maxburst = TCP_MAXBURST;
#endif
struct rmxp_tao tao;
#ifdef INET6
struct ip6_hdr *ip6 = NULL;
int isipv6;
bzero(&tao, sizeof(tao));
isipv6 = (tp->t_inpcb->inp_vflag & INP_IPV6) != 0;
#endif
#ifdef TCP_ECN
int needect;
#endif
INP_LOCK_ASSERT(tp->t_inpcb);
/*
* Determine length of data that should be transmitted,
* and flags that will be used.
* If there is some data or critical controls (SYN, RST)
* to send, then transmit; otherwise, investigate further.
*/
idle = (tp->t_flags & TF_LASTIDLE) || (tp->snd_max == tp->snd_una);
if (idle && (ticks - tp->t_rcvtime) >= tp->t_rxtcur) {
/*
* We have been idle for "a while" and no acks are
* expected to clock out any data we send --
* slow start to get ack "clock" running again.
*
* Set the slow-start flight size depending on whether
* this is a local network or not.
*/
int ss = ss_fltsz;
#ifdef INET6
if (isipv6) {
if (in6_localaddr(&tp->t_inpcb->in6p_faddr))
ss = ss_fltsz_local;
} else
#endif
if (in_localaddr(tp->t_inpcb->inp_faddr))
ss = ss_fltsz_local;
tp->snd_cwnd = tp->t_maxseg * ss;
}
tp->t_flags &= ~TF_LASTIDLE;
if (idle) {
if (tp->t_flags & TF_MORETOCOME) {
tp->t_flags |= TF_LASTIDLE;
idle = 0;
}
}
again:
/*
* If we've recently taken a timeout, snd_max will be greater than
* snd_nxt. There may be SACK information that allows us to avoid
* resending already delivered data. Adjust snd_nxt accordingly.
*/
if (tp->sack_enable && SEQ_LT(tp->snd_nxt, tp->snd_max))
tcp_sack_adjust(tp);
sendalot = 0;
off = tp->snd_nxt - tp->snd_una;
sendwin = min(tp->snd_wnd, tp->snd_cwnd);
sendwin = min(sendwin, tp->snd_bwnd);
flags = tcp_outflags[tp->t_state];
/*
* Send any SACK-generated retransmissions. If we're explicitly trying
* to send out new data (when sendalot is 1), bypass this function.
* If we retransmit in fast recovery mode, decrement snd_cwnd, since
* we're replacing a (future) new transmission with a retransmission
* now, and we previously incremented snd_cwnd in tcp_input().
*/
/*
* Still in sack recovery , reset rxmit flag to zero.
*/
sack_rxmit = 0;
sack_bytes_rxmt = 0;
len = 0;
p = NULL;
if (tp->sack_enable && IN_FASTRECOVERY(tp) &&
(p = tcp_sack_output(tp, &sack_bytes_rxmt))) {
long cwin;
cwin = min(tp->snd_wnd, tp->snd_cwnd) - sack_bytes_rxmt;
if (cwin < 0)
cwin = 0;
/* Do not retransmit SACK segments beyond snd_recover */
if (SEQ_GT(p->end, tp->snd_recover)) {
/*
* (At least) part of sack hole extends beyond
* snd_recover. Check to see if we can rexmit data
* for this hole.
*/
if (SEQ_GEQ(p->rxmit, tp->snd_recover)) {
/*
* Can't rexmit any more data for this hole.
* That data will be rexmitted in the next
* sack recovery episode, when snd_recover
* moves past p->rxmit.
*/
p = NULL;
goto after_sack_rexmit;
} else
/* Can rexmit part of the current hole */
len = ((long)ulmin(cwin,
tp->snd_recover - p->rxmit));
} else
len = ((long)ulmin(cwin, p->end - p->rxmit));
off = p->rxmit - tp->snd_una;
KASSERT(off >= 0,("%s: sack block to the left of una : %d",
__func__, off));
if (len > 0) {
sack_rxmit = 1;
sendalot = 1;
tcpstat.tcps_sack_rexmits++;
tcpstat.tcps_sack_rexmit_bytes +=
min(len, tp->t_maxseg);
}
}
after_sack_rexmit:
/*
* Get standard flags, and add SYN or FIN if requested by 'hidden'
* state flags.
*/
if (tp->t_flags & TF_NEEDFIN)
flags |= TH_FIN;
if (tp->t_flags & TF_NEEDSYN)
flags |= TH_SYN;
SOCKBUF_LOCK(&so->so_snd);
/*
* If in persist timeout with window of 0, send 1 byte.
* Otherwise, if window is small but nonzero
* and timer expired, we will send what we can
* and go to transmit state.
*/
if (tp->t_force) {
if (sendwin == 0) {
/*
* If we still have some data to send, then
* clear the FIN bit. Usually this would
* happen below when it realizes that we
* aren't sending all the data. However,
* if we have exactly 1 byte of unsent data,
* then it won't clear the FIN bit below,
* and if we are in persist state, we wind
* up sending the packet without recording
* that we sent the FIN bit.
*
* We can't just blindly clear the FIN bit,
* because if we don't have any more data
* to send then the probe will be the FIN
* itself.
*/
if (off < so->so_snd.sb_cc)
flags &= ~TH_FIN;
sendwin = 1;
} else {
callout_stop(tp->tt_persist);
tp->t_rxtshift = 0;
}
}
/*
* If snd_nxt == snd_max and we have transmitted a FIN, the
* offset will be > 0 even if so_snd.sb_cc is 0, resulting in
* a negative length. This can also occur when TCP opens up
* its congestion window while receiving additional duplicate
* acks after fast-retransmit because TCP will reset snd_nxt
* to snd_max after the fast-retransmit.
*
* In the normal retransmit-FIN-only case, however, snd_nxt will
* be set to snd_una, the offset will be 0, and the length may
* wind up 0.
*
* If sack_rxmit is true we are retransmitting from the scoreboard
* in which case len is already set.
*/
if (sack_rxmit == 0) {
if (sack_bytes_rxmt == 0)
len = ((long)ulmin(so->so_snd.sb_cc, sendwin) - off);
else {
long cwin;
/*
* We are inside of a SACK recovery episode and are
* sending new data, having retransmitted all the
* data possible in the scoreboard.
*/
len = ((long)ulmin(so->so_snd.sb_cc, tp->snd_wnd)
- off);
/*
* Don't remove this (len > 0) check !
* We explicitly check for len > 0 here (although it
* isn't really necessary), to work around a gcc
* optimization issue - to force gcc to compute
* len above. Without this check, the computation
* of len is bungled by the optimizer.
*/
if (len > 0) {
cwin = tp->snd_cwnd -
(tp->snd_nxt - tp->sack_newdata) -
sack_bytes_rxmt;
if (cwin < 0)
cwin = 0;
len = lmin(len, cwin);
}
}
}
/*
* Lop off SYN bit if it has already been sent. However, if this
* is SYN-SENT state and if segment contains data and if we don't
* know that foreign host supports TAO, suppress sending segment.
*/
if ((flags & TH_SYN) && SEQ_GT(tp->snd_nxt, tp->snd_una)) {
flags &= ~TH_SYN;
off--, len++;
if (tcp_do_rfc1644)
tcp_hc_gettao(&tp->t_inpcb->inp_inc, &tao);
if (len > 0 && tp->t_state == TCPS_SYN_SENT &&
tao.tao_ccsent == 0)
goto just_return;
}
/*
* Be careful not to send data and/or FIN on SYN segments
* in cases when no CC option will be sent.
* This measure is needed to prevent interoperability problems
* with not fully conformant TCP implementations.
*/
if ((flags & TH_SYN) &&
((tp->t_flags & TF_NOOPT) || !(tp->t_flags & TF_REQ_CC) ||
((flags & TH_ACK) && !(tp->t_flags & TF_RCVD_CC)))) {
len = 0;
flags &= ~TH_FIN;
}
if (len < 0) {
/*
* If FIN has been sent but not acked,
* but we haven't been called to retransmit,
* len will be < 0. Otherwise, window shrank
* after we sent into it. If window shrank to 0,
* cancel pending retransmit, pull snd_nxt back
* to (closed) window, and set the persist timer
* if it isn't already going. If the window didn't
* close completely, just wait for an ACK.
*/
len = 0;
if (sendwin == 0) {
callout_stop(tp->tt_rexmt);
tp->t_rxtshift = 0;
tp->snd_nxt = tp->snd_una;
if (!callout_active(tp->tt_persist))
tcp_setpersist(tp);
}
}
/*
* len will be >= 0 after this point. Truncate to the maximum
* segment length and ensure that FIN is removed if the length
* no longer contains the last data byte.
*/
if (len > tp->t_maxseg) {
len = tp->t_maxseg;
sendalot = 1;
}
if (sack_rxmit) {
if (SEQ_LT(p->rxmit + len, tp->snd_una + so->so_snd.sb_cc))
flags &= ~TH_FIN;
} else {
if (SEQ_LT(tp->snd_nxt + len, tp->snd_una + so->so_snd.sb_cc))
flags &= ~TH_FIN;
}
recwin = sbspace(&so->so_rcv);
/*
* Sender silly window avoidance. We transmit under the following
* conditions when len is non-zero:
*
* - We have a full segment
* - This is the last buffer in a write()/send() and we are
* either idle or running NODELAY
* - we've timed out (e.g. persist timer)
* - we have more then 1/2 the maximum send window's worth of
* data (receiver may be limited the window size)
* - we need to retransmit
*/
if (len) {
if (len == tp->t_maxseg)
goto send;
/*
* NOTE! on localhost connections an 'ack' from the remote
* end may occur synchronously with the output and cause
* us to flush a buffer queued with moretocome. XXX
*
* note: the len + off check is almost certainly unnecessary.
*/
if (!(tp->t_flags & TF_MORETOCOME) && /* normal case */
(idle || (tp->t_flags & TF_NODELAY)) &&
len + off >= so->so_snd.sb_cc &&
(tp->t_flags & TF_NOPUSH) == 0) {
goto send;
}
if (tp->t_force) /* typ. timeout case */
goto send;
if (len >= tp->max_sndwnd / 2 && tp->max_sndwnd > 0)
goto send;
if (SEQ_LT(tp->snd_nxt, tp->snd_max)) /* retransmit case */
goto send;
if (sack_rxmit)
goto send;
}
/*
* Compare available window to amount of window
* known to peer (as advertised window less
* next expected input). If the difference is at least two
* max size segments, or at least 50% of the maximum possible
* window, then want to send a window update to peer.
* Skip this if the connection is in T/TCP half-open state.
*/
if (recwin > 0 && !(tp->t_flags & TF_NEEDSYN)) {
/*
* "adv" is the amount we can increase the window,
* taking into account that we are limited by
* TCP_MAXWIN << tp->rcv_scale.
*/
long adv = min(recwin, (long)TCP_MAXWIN << tp->rcv_scale) -
(tp->rcv_adv - tp->rcv_nxt);
if (adv >= (long) (2 * tp->t_maxseg))
goto send;
if (2 * adv >= (long) so->so_rcv.sb_hiwat)
goto send;
}
/*
* Send if we owe the peer an ACK, RST, SYN, or urgent data. ACKNOW
* is also a catch-all for the retransmit timer timeout case.
*/
if (tp->t_flags & TF_ACKNOW)
goto send;
if ((flags & TH_RST) ||
((flags & TH_SYN) && (tp->t_flags & TF_NEEDSYN) == 0))
goto send;
if (SEQ_GT(tp->snd_up, tp->snd_una))
goto send;
/*
* If our state indicates that FIN should be sent
* and we have not yet done so, then we need to send.
*/
if (flags & TH_FIN &&
((tp->t_flags & TF_SENTFIN) == 0 || tp->snd_nxt == tp->snd_una))
goto send;
/*
* In SACK, it is possible for tcp_output to fail to send a segment
* after the retransmission timer has been turned off. Make sure
* that the retransmission timer is set.
*/
if (tp->sack_enable && SEQ_GT(tp->snd_max, tp->snd_una) &&
!callout_active(tp->tt_rexmt) &&
!callout_active(tp->tt_persist)) {
callout_reset(tp->tt_rexmt, tp->t_rxtcur,
tcp_timer_rexmt, tp);
goto just_return;
}
/*
* TCP window updates are not reliable, rather a polling protocol
* using ``persist'' packets is used to insure receipt of window
* updates. The three ``states'' for the output side are:
* idle not doing retransmits or persists
* persisting to move a small or zero window
* (re)transmitting and thereby not persisting
*
* callout_active(tp->tt_persist)
* is true when we are in persist state.
* tp->t_force
* is set when we are called to send a persist packet.
* callout_active(tp->tt_rexmt)
* is set when we are retransmitting
* The output side is idle when both timers are zero.
*
* If send window is too small, there is data to transmit, and no
* retransmit or persist is pending, then go to persist state.
* If nothing happens soon, send when timer expires:
* if window is nonzero, transmit what we can,
* otherwise force out a byte.
*/
if (so->so_snd.sb_cc && !callout_active(tp->tt_rexmt) &&
!callout_active(tp->tt_persist)) {
tp->t_rxtshift = 0;
tcp_setpersist(tp);
}
/*
* No reason to send a segment, just return.
*/
just_return:
SOCKBUF_UNLOCK(&so->so_snd);
return (0);
send:
SOCKBUF_LOCK_ASSERT(&so->so_snd);
/*
* Before ESTABLISHED, force sending of initial options
* unless TCP set not to do any options.
* NOTE: we assume that the IP/TCP header plus TCP options
* always fit in a single mbuf, leaving room for a maximum
* link header, i.e.
* max_linkhdr + sizeof (struct tcpiphdr) + optlen <= MCLBYTES
*/
optlen = 0;
#ifdef INET6
if (isipv6)
hdrlen = sizeof (struct ip6_hdr) + sizeof (struct tcphdr);
else
#endif
hdrlen = sizeof (struct tcpiphdr);
if (flags & TH_SYN) {
tp->snd_nxt = tp->iss;
if ((tp->t_flags & TF_NOOPT) == 0) {
u_short mss;
opt[0] = TCPOPT_MAXSEG;
opt[1] = TCPOLEN_MAXSEG;
mss = htons((u_short) tcp_mssopt(&tp->t_inpcb->inp_inc));
(void)memcpy(opt + 2, &mss, sizeof(mss));
optlen = TCPOLEN_MAXSEG;
/*
* If this is the first SYN of connection (not a SYN
* ACK), include SACK_PERMIT_HDR option. If this is a
* SYN ACK, include SACK_PERMIT_HDR option if peer has
* already done so. This is only for active connect,
* since the syncache takes care of the passive connect.
*/
if (tp->sack_enable && ((flags & TH_ACK) == 0 ||
(tp->t_flags & TF_SACK_PERMIT))) {
*((u_int32_t *) (opt + optlen)) =
htonl(TCPOPT_SACK_PERMIT_HDR);
optlen += 4;
}
if ((tp->t_flags & TF_REQ_SCALE) &&
((flags & TH_ACK) == 0 ||
(tp->t_flags & TF_RCVD_SCALE))) {
*((u_int32_t *)(opt + optlen)) = htonl(
TCPOPT_NOP << 24 |
TCPOPT_WINDOW << 16 |
TCPOLEN_WINDOW << 8 |
tp->request_r_scale);
optlen += 4;
}
}
}
/*
* Send a timestamp and echo-reply if this is a SYN and our side
* wants to use timestamps (TF_REQ_TSTMP is set) or both our side
* and our peer have sent timestamps in our SYN's.
*/
if ((tp->t_flags & (TF_REQ_TSTMP|TF_NOOPT)) == TF_REQ_TSTMP &&
(flags & TH_RST) == 0 &&
((flags & TH_ACK) == 0 ||
(tp->t_flags & TF_RCVD_TSTMP))) {
u_int32_t *lp = (u_int32_t *)(opt + optlen);
/* Form timestamp option as shown in appendix A of RFC 1323. */
*lp++ = htonl(TCPOPT_TSTAMP_HDR);
*lp++ = htonl(ticks);
*lp = htonl(tp->ts_recent);
optlen += TCPOLEN_TSTAMP_APPA;
}
/*
* Send SACKs if necessary. This should be the last option processed.
* Only as many SACKs are sent as are permitted by the maximum options
* size. No more than three SACKs are sent.
*/
if (tp->sack_enable && tp->t_state == TCPS_ESTABLISHED &&
(tp->t_flags & (TF_SACK_PERMIT|TF_NOOPT)) == TF_SACK_PERMIT &&
tp->rcv_numsacks) {
u_int32_t *lp = (u_int32_t *)(opt + optlen);
u_int32_t *olp = lp++;
int count = 0; /* actual number of SACKs inserted */
int maxsack = (MAX_TCPOPTLEN - (optlen + 4))/TCPOLEN_SACK;
tcpstat.tcps_sack_send_blocks++;
maxsack = min(maxsack, TCP_MAX_SACK);
for (i = 0; (i < tp->rcv_numsacks && count < maxsack); i++) {
struct sackblk sack = tp->sackblks[i];
if (sack.start == 0 && sack.end == 0)
continue;
*lp++ = htonl(sack.start);
*lp++ = htonl(sack.end);
count++;
}
*olp = htonl(TCPOPT_SACK_HDR|(TCPOLEN_SACK*count+2));
optlen += TCPOLEN_SACK*count + 4; /* including leading NOPs */
}
/*
* Send `CC-family' options if our side wants to use them (TF_REQ_CC),
* options are allowed (!TF_NOOPT) and it's not a RST.
*/
if ((tp->t_flags & (TF_REQ_CC|TF_NOOPT)) == TF_REQ_CC &&
(flags & TH_RST) == 0) {
switch (flags & (TH_SYN|TH_ACK)) {
/*
* This is a normal ACK, send CC if we received CC before
* from our peer.
*/
case TH_ACK:
if (!(tp->t_flags & TF_RCVD_CC))
break;
/*FALLTHROUGH*/
/*
* We can only get here in T/TCP's SYN_SENT* state, when
* we're a sending a non-SYN segment without waiting for
* the ACK of our SYN. A check above assures that we only
* do this if our peer understands T/TCP.
*/
case 0:
opt[optlen++] = TCPOPT_NOP;
opt[optlen++] = TCPOPT_NOP;
opt[optlen++] = TCPOPT_CC;
opt[optlen++] = TCPOLEN_CC;
*(u_int32_t *)&opt[optlen] = htonl(tp->cc_send);
optlen += 4;
break;
/*
* This is our initial SYN, check whether we have to use
* CC or CC.new.
*/
case TH_SYN:
opt[optlen++] = TCPOPT_NOP;
opt[optlen++] = TCPOPT_NOP;
opt[optlen++] = tp->t_flags & TF_SENDCCNEW ?
TCPOPT_CCNEW : TCPOPT_CC;
opt[optlen++] = TCPOLEN_CC;
*(u_int32_t *)&opt[optlen] = htonl(tp->cc_send);
optlen += 4;
break;
/*
* This is a SYN,ACK; send CC and CC.echo if we received
* CC from our peer.
*/
case (TH_SYN|TH_ACK):
if (tp->t_flags & TF_RCVD_CC) {
opt[optlen++] = TCPOPT_NOP;
opt[optlen++] = TCPOPT_NOP;
opt[optlen++] = TCPOPT_CC;
opt[optlen++] = TCPOLEN_CC;
*(u_int32_t *)&opt[optlen] =
htonl(tp->cc_send);
optlen += 4;
opt[optlen++] = TCPOPT_NOP;
opt[optlen++] = TCPOPT_NOP;
opt[optlen++] = TCPOPT_CCECHO;
opt[optlen++] = TCPOLEN_CC;
*(u_int32_t *)&opt[optlen] =
htonl(tp->cc_recv);
optlen += 4;
}
break;
}
}
#ifdef TCP_SIGNATURE
#ifdef INET6
if (!isipv6)
#endif
if (tp->t_flags & TF_SIGNATURE) {
int i;
u_char *bp;
/* Initialize TCP-MD5 option (RFC2385) */
bp = (u_char *)opt + optlen;
*bp++ = TCPOPT_SIGNATURE;
*bp++ = TCPOLEN_SIGNATURE;
sigoff = optlen + 2;
for (i = 0; i < TCP_SIGLEN; i++)
*bp++ = 0;
optlen += TCPOLEN_SIGNATURE;
/* Terminate options list and maintain 32-bit alignment. */
*bp++ = TCPOPT_NOP;
*bp++ = TCPOPT_EOL;
optlen += 2;
}
#endif /* TCP_SIGNATURE */
hdrlen += optlen;
#ifdef INET6
if (isipv6)
ipoptlen = ip6_optlen(tp->t_inpcb);
else
#endif
if (tp->t_inpcb->inp_options)
ipoptlen = tp->t_inpcb->inp_options->m_len -
offsetof(struct ipoption, ipopt_list);
else
static int tcpup_state_receive(struct tcpup_info *upp, struct tcpiphdr *tcp, size_t dlen)
{
int xflags = 0;
int snd_una = htonl(tcp->th_ack);
if (tcp->th_flags & TH_RST) {
upp->t_state = TCPS_CLOSED;
return 0;
}
if ((tcp->th_flags & TH_ACK) && SEQ_GT(snd_una, upp->snd_una)) {
/* update snd una from peer */
upp->snd_una = snd_una;
}
switch (upp->t_state) {
case TCPS_SYN_SENT:
xflags = TH_SYN| TH_ACK;
if ((tcp->th_flags & xflags) == TH_SYN) {
assert((tcp->th_flags & TH_FIN) != TH_FIN);
tcp_state_preload(upp, TCPS_SYN_RECEIVED, htonl(tcp->th_seq) + 1);
return 0;
}
if ((tcp->th_flags & xflags) == xflags
&& SEQ_GEQ(htonl(tcp->th_ack), upp->snd_nxt)) {
assert((tcp->th_flags & TH_FIN) != TH_FIN);
tcp_state_preload(upp, TCPS_ESTABLISHED, htonl(tcp->th_seq));
return 0;
}
break;
case TCPS_SYN_RECEIVED:
if ((tcp->th_flags & TH_ACK) == TH_ACK
&& SEQ_GEQ(htonl(tcp->th_ack), upp->snd_nxt)) {
assert((tcp->th_flags & TH_FIN) != TH_FIN);
tcp_state_preload(upp, TCPS_ESTABLISHED, htonl(tcp->th_seq));
return 0;
}
break;
case TCPS_ESTABLISHED:
if ((tcp->th_flags & TH_FIN) == TH_FIN) {
tcp_state_preload(upp, TCPS_CLOSE_WAIT, htonl(tcp->th_seq) + 1);
return 0;
}
break;
case TCPS_FIN_WAIT_1:
xflags = TH_FIN| TH_ACK;
if ((tcp->th_flags & xflags) == xflags
&& SEQ_GEQ(htonl(tcp->th_ack), upp->snd_nxt)) {
tcp_state_preload(upp, TCPS_TIME_WAIT, htonl(tcp->th_seq) + dlen + 1);
return 0;
}
if ((tcp->th_flags & TH_FIN) == TH_FIN) {
tcp_state_preload(upp, TCPS_CLOSING, htonl(tcp->th_seq) + dlen + 1);
return 0;
}
if ((tcp->th_flags & TH_ACK) == TH_ACK
&& SEQ_GEQ(htonl(tcp->th_ack), upp->snd_nxt)) {
tcp_state_preload(upp, TCPS_FIN_WAIT_2, htonl(tcp->th_seq) + dlen);
return 0;
}
break;
case TCPS_FIN_WAIT_2:
if ((tcp->th_flags & TH_FIN) == TH_FIN) {
tcp_state_preload(upp, TCPS_TIME_WAIT, htonl(tcp->th_seq) + dlen + 1);
return 0;
}
break;
case TCPS_CLOSING:
if ((tcp->th_flags & TH_ACK) == TH_ACK
&& SEQ_GEQ(htonl(tcp->th_ack), upp->snd_nxt)) {
tcp_state_preload(upp, TCPS_TIME_WAIT, htonl(tcp->th_seq) + dlen);
return 0;
}
break;
case TCPS_LAST_ACK:
if ((tcp->th_flags & TH_ACK) == TH_ACK
&& SEQ_GEQ(htonl(tcp->th_ack), upp->snd_nxt)) {
tcp_state_preload(upp, TCPS_CLOSED, htonl(tcp->th_seq) + dlen);
return 0;
}
break;
case TCPS_TIME_WAIT:
fprintf(stderr, "before TIME_WAIT -> TIME_WAIT\n");
break;
}
return 0;
}
/*
* 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;
}
/*
* SD PDU / SOS_READY Processor
*
* Arguments:
* sop pointer to sscop connection block
* m pointer to PDU buffer (without trailer)
* trlr pointer to PDU trailer
*
* Returns:
* none
*
*/
static void
sscop_sd_ready(struct sscop *sop, KBuffer *m, caddr_t trlr)
{
struct sd_pdu *sp = (struct sd_pdu *)trlr;
struct pdu_hdr *php;
KBuffer *n;
sscop_seq ns;
int err, space;
/*
* Get PDU sequence number
*/
SEQ_SET(ns, ntohl(sp->sd_ns));
/*
* Ensure that the sequence number fits within the window
*/
if (SEQ_GEQ(ns, sop->so_rcvmax, sop->so_rcvnext)) {
/*
* It doesn't, drop received data
*/
KB_FREEALL(m);
/*
* If next highest PDU hasn't reached window end yet,
* then send a USTAT to inform transmitter of this gap
*/
if (SEQ_LT(sop->so_rcvhigh, sop->so_rcvmax, sop->so_rcvnext)) {
sscop_send_ustat(sop, sop->so_rcvmax);
sop->so_rcvhigh = sop->so_rcvmax;
}
return;
}
/*
* If this is the next in-sequence PDU, hand it to user
*/
if (ns == sop->so_rcvnext) {
STACK_CALL(SSCOP_DATA_IND, sop->so_upper, sop->so_toku,
sop->so_connvc, (int)m, ns, err);
if (err) {
KB_FREEALL(m);
return;
}
/*
* Bump next expected sequence number
*/
SEQ_INCR(sop->so_rcvnext, 1);
/*
* Slide receive window down
*/
SEQ_INCR(sop->so_rcvmax, 1);
/*
* Is this the highest sequence PDU we've received??
*/
if (ns == sop->so_rcvhigh) {
/*
* Yes, bump the limit and exit
*/
sop->so_rcvhigh = sop->so_rcvnext;
return;
}
/*
* This is a retransmitted PDU, so see if we have
* more in-sequence PDUs already queued up
*/
while ((php = sop->so_recv_hd) &&
(php->ph_ns == sop->so_rcvnext)) {
/*
* Yup we do, so remove next PDU from queue and
* pass it up to the user as well
*/
sop->so_recv_hd = php->ph_recv_lk;
if (sop->so_recv_hd == NULL)
sop->so_recv_tl = NULL;
STACK_CALL(SSCOP_DATA_IND, sop->so_upper, sop->so_toku,
sop->so_connvc, (int)php->ph_buf, php->ph_ns,
err);
if (err) {
/*
* Should never happen, but...
*/
KB_FREEALL(php->ph_buf);
sscop_abort(sop, "stack memory\n");
return;
}
/*
* Bump next expected sequence number
*/
SEQ_INCR(sop->so_rcvnext, 1);
/*
* Slide receive window down
*/
SEQ_INCR(sop->so_rcvmax, 1);
}
/*
* Finished with data delivery...
*/
return;
}
/*
* We're gonna have to queue this PDU, so find space
* for the PDU header
*/
KB_HEADROOM(m, space);
/*
* If there's not enough room in the received buffer,
* allocate & link a new buffer for the header
*/
if (space < sizeof(struct pdu_hdr)) {
KB_ALLOC(n, sizeof(struct pdu_hdr), KB_F_NOWAIT, KB_T_HEADER);
if (n == NULL) {
KB_FREEALL(m);
return;
}
KB_HEADSET(n, sizeof(struct pdu_hdr));
KB_LEN(n) = 0;
KB_LINKHEAD(n, m);
m = n;
}
/*
* Build PDU header
*
* We can at least assume/require that the start of
* the user data is aligned. Also note that we don't
* include this header in the buffer len/offset fields.
*/
KB_DATASTART(m, php, struct pdu_hdr *);
php--;
php->ph_ns = ns;
php->ph_buf = m;
/*
* Insert PDU into the receive queue
*/
if (sscop_recv_insert(sop, php)) {
/*
* Oops, a duplicate sequence number PDU is already on
* the queue, somethings wrong here.
*/
sscop_maa_error(sop, 'Q');
/*
* Free buffers
*/
KB_FREEALL(m);
/*
* Go into recovery mode
*/
q2110_error_recovery(sop);
return;
}
/*
* Are we at the high-water mark??
*/
if (ns == sop->so_rcvhigh) {
/*
* Yes, just bump the mark
*/
SEQ_INCR(sop->so_rcvhigh, 1);
return;
}
/*
* Are we beyond the high-water mark??
*/
if (SEQ_GT(ns, sop->so_rcvhigh, sop->so_rcvnext)) {
/*
* Yes, then there's a missing PDU, so inform the transmitter
*/
sscop_send_ustat(sop, ns);
/*
* Update high-water mark
*/
sop->so_rcvhigh = SEQ_ADD(ns, 1);
}
return;
}
/**
* \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);
}
static INLINE void ApplyFlowDepth(HTTPINSPECT_CONF *ServerConf, Packet *p,
HttpSessionData *sd, int resp_header_size, int expected, uint32_t seq_num)
{
if(!ServerConf->server_flow_depth)
{
SetDetectLimit(p, p->dsize);
}
else if(ServerConf->server_flow_depth == -1)
{
SetDetectLimit(p, resp_header_size);
}
else
{
if(sd != NULL)
{
if(sd->resp_state.is_max_seq )
{
if(SEQ_GEQ((sd->resp_state.max_seq), seq_num))
{
if(p->dsize > (sd->resp_state.max_seq- seq_num))
{
SetDetectLimit(p, (sd->resp_state.max_seq-seq_num));
return;
}
else
{
SetDetectLimit(p, p->dsize);
return;
}
}
else
{
SetDetectLimit(p, resp_header_size);
return;
}
}
else
{
if(expected)
{
if(p->dsize > ServerConf->server_flow_depth)
{
SetDetectLimit(p, ServerConf->server_flow_depth);
return;
}
else
{
SetDetectLimit(p, p->dsize);
return;
}
}
else
{
SetDetectLimit(p, 0);
return;
}
}
}
else
{
SetDetectLimit(p, p->dsize);
}
}
}
void
tcp_timer_rexmt(void *arg)
{
struct tcpcb *tp = arg;
uint32_t rto;
#ifdef TCP_DEBUG
struct socket *so = NULL;
short ostate;
#endif
mutex_enter(softnet_lock);
if ((tp->t_flags & TF_DEAD) != 0) {
mutex_exit(softnet_lock);
return;
}
if (!callout_expired(&tp->t_timer[TCPT_REXMT])) {
mutex_exit(softnet_lock);
return;
}
KERNEL_LOCK(1, NULL);
if ((tp->t_flags & TF_PMTUD_PEND) && tp->t_inpcb &&
SEQ_GEQ(tp->t_pmtud_th_seq, tp->snd_una) &&
SEQ_LT(tp->t_pmtud_th_seq, (int)(tp->snd_una + tp->t_ourmss))) {
extern struct sockaddr_in icmpsrc;
struct icmp icmp;
tp->t_flags &= ~TF_PMTUD_PEND;
/* XXX create fake icmp message with relevant entries */
icmp.icmp_nextmtu = tp->t_pmtud_nextmtu;
icmp.icmp_ip.ip_len = tp->t_pmtud_ip_len;
icmp.icmp_ip.ip_hl = tp->t_pmtud_ip_hl;
icmpsrc.sin_addr = tp->t_inpcb->inp_faddr;
icmp_mtudisc(&icmp, icmpsrc.sin_addr);
/*
* Notify all connections to the same peer about
* new mss and trigger retransmit.
*/
in_pcbnotifyall(&tcbtable, icmpsrc.sin_addr, EMSGSIZE,
tcp_mtudisc);
KERNEL_UNLOCK_ONE(NULL);
mutex_exit(softnet_lock);
return;
}
#ifdef TCP_DEBUG
#ifdef INET
if (tp->t_inpcb)
so = tp->t_inpcb->inp_socket;
#endif
#ifdef INET6
if (tp->t_in6pcb)
so = tp->t_in6pcb->in6p_socket;
#endif
ostate = tp->t_state;
#endif /* TCP_DEBUG */
/*
* Clear the SACK scoreboard, reset FACK estimate.
*/
tcp_free_sackholes(tp);
tp->snd_fack = tp->snd_una;
/*
* Retransmission timer went off. Message has not
* been acked within retransmit interval. Back off
* to a longer retransmit interval and retransmit one segment.
*/
if (++tp->t_rxtshift > TCP_MAXRXTSHIFT) {
tp->t_rxtshift = TCP_MAXRXTSHIFT;
TCP_STATINC(TCP_STAT_TIMEOUTDROP);
tp = tcp_drop(tp, tp->t_softerror ?
tp->t_softerror : ETIMEDOUT);
goto out;
}
TCP_STATINC(TCP_STAT_REXMTTIMEO);
rto = TCP_REXMTVAL(tp);
if (rto < tp->t_rttmin)
rto = tp->t_rttmin;
TCPT_RANGESET(tp->t_rxtcur, rto * tcp_backoff[tp->t_rxtshift],
tp->t_rttmin, TCPTV_REXMTMAX);
TCP_TIMER_ARM(tp, TCPT_REXMT, tp->t_rxtcur);
/*
* If we are losing and we are trying path MTU discovery,
* try turning it off. This will avoid black holes in
* the network which suppress or fail to send "packet
* too big" ICMP messages. We should ideally do
* lots more sophisticated searching to find the right
* value here...
*/
if (tp->t_mtudisc && tp->t_rxtshift > TCP_MAXRXTSHIFT / 6) {
TCP_STATINC(TCP_STAT_PMTUBLACKHOLE);
#ifdef INET
/* try turning PMTUD off */
if (tp->t_inpcb)
tp->t_mtudisc = 0;
#endif
#ifdef INET6
/* try using IPv6 minimum MTU */
if (tp->t_in6pcb)
tp->t_mtudisc = 0;
#endif
/* XXX: more sophisticated Black hole recovery code? */
}
/*
* If losing, let the lower level know and try for
* a better route. Also, if we backed off this far,
* our srtt estimate is probably bogus. Clobber it
* so we'll take the next rtt measurement as our srtt;
* move the current srtt into rttvar to keep the current
* retransmit times until then.
*/
if (tp->t_rxtshift > TCP_MAXRXTSHIFT / 4) {
#ifdef INET
if (tp->t_inpcb)
in_losing(tp->t_inpcb);
#endif
#ifdef INET6
if (tp->t_in6pcb)
in6_losing(tp->t_in6pcb);
#endif
/*
* This operation is not described in RFC2988. The
* point is to keep srtt+4*rttvar constant, so we
* should shift right 2 bits to divide by 4, and then
* shift right one bit because the storage
* representation of rttvar is 1/16s vs 1/32s for
* srtt.
*/
tp->t_rttvar += (tp->t_srtt >> TCP_RTT_SHIFT);
tp->t_srtt = 0;
}
tp->snd_nxt = tp->snd_una;
tp->snd_high = tp->snd_max;
/*
* If timing a segment in this window, stop the timer.
*/
tp->t_rtttime = 0;
/*
* Remember if we are retransmitting a SYN, because if
* we do, set the initial congestion window must be set
* to 1 segment.
*/
if (tp->t_state == TCPS_SYN_SENT)
tp->t_flags |= TF_SYN_REXMT;
/*
* Adjust congestion control parameters.
*/
tp->t_congctl->slow_retransmit(tp);
(void) tcp_output(tp);
out:
#ifdef TCP_DEBUG
if (tp && so->so_options & SO_DEBUG)
tcp_trace(TA_USER, ostate, tp, NULL,
PRU_SLOWTIMO | (TCPT_REXMT << 8));
#endif
KERNEL_UNLOCK_ONE(NULL);
mutex_exit(softnet_lock);
}
/*
* 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;
}
static int tcpup_state_send(struct tcpup_info *upp, struct tcpiphdr *tcp, size_t dlen)
{
int xflags = 0;
if (tcp->th_flags & TH_RST) {
upp->t_state = TCPS_CLOSED;
return 0;
}
if (upp->x_state != upp->t_state
&& (tcp->th_flags & TH_ACK)
&& SEQ_GEQ(htonl(tcp->th_ack), upp->rcv_una)) {
tcp_state_update(upp, upp->x_state);
upp->t_state = upp->x_state;
}
switch (upp->t_state) {
case TCPS_CLOSED:
xflags = TH_SYN| TH_ACK;
if ((tcp->th_flags & xflags) == TH_SYN) {
upp->snd_nxt = htonl(tcp->th_seq) + 1;
upp->snd_max = htonl(tcp->th_seq) + 1;
upp->snd_una = htonl(tcp->th_seq) + 1;
tcp_state_update(upp, TCPS_SYN_SENT);
return 0;
}
break;
case TCPS_SYN_RECEIVED:
assert((tcp->th_flags & TH_FIN) != TH_FIN);
xflags = TH_SYN| TH_ACK;
if ((tcp->th_flags & xflags) == TH_ACK
&& SEQ_GT(htonl(tcp->th_seq), upp->snd_nxt)) {
tcp_state_update(upp, upp->x_state);
return 0;
}
break;
case TCPS_ESTABLISHED:
if ((tcp->th_flags & TH_FIN) == TH_FIN) {
upp->snd_nxt = htonl(tcp->th_seq) + dlen + 1;
tcp_state_update(upp, TCPS_FIN_WAIT_1);
return 0;
}
break;
case TCPS_CLOSE_WAIT:
if ((tcp->th_flags & TH_FIN) == TH_FIN) {
upp->snd_nxt = htonl(tcp->th_seq) + dlen + 1;
tcp_state_update(upp, TCPS_LAST_ACK);
return 0;
}
break;
case TCPS_FIN_WAIT_1:
xflags = TH_FIN| TH_ACK;
if ((tcp->th_flags & xflags) == TH_ACK) {
tcp_state_update(upp, upp->x_state);
return 0;
}
break;
}
if (dlen > 0) {
upp->snd_nxt = htonl(tcp->th_seq) + dlen;
if (SEQ_GT(upp->snd_nxt, upp->snd_max)) {
/* update snd max to nxt */
upp->snd_max = upp->snd_nxt;
}
}
return 0;
}
/*
* 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--;
}
}
enum CT_UPDATE_RES
OvsConntrackUpdateTcpEntry(OVS_CT_ENTRY* conn_,
const TCPHdr *tcp,
PNET_BUFFER_LIST nbl,
BOOLEAN reply,
UINT64 now)
{
struct conn_tcp *conn = OvsCastConntrackEntryToTcpEntry(conn_);
/* The peer that sent 'pkt' */
struct tcp_peer *src = &conn->peer[reply ? 1 : 0];
/* The peer that should receive 'pkt' */
struct tcp_peer *dst = &conn->peer[reply ? 0 : 1];
uint8_t sws = 0, dws = 0;
UINT16 tcp_flags = ntohs(tcp->flags);
uint16_t win = ntohs(tcp->window);
uint32_t ack, end, seq, orig_seq;
uint32_t p_len = OvsGetTcpPayloadLength(nbl);
int ackskew;
if (OvsCtInvalidTcpFlags(tcp_flags)) {
return CT_UPDATE_INVALID;
}
if (((tcp_flags & (TCP_SYN|TCP_ACK)) == TCP_SYN)
&& dst->state >= CT_DPIF_TCPS_FIN_WAIT_2
&& src->state >= CT_DPIF_TCPS_FIN_WAIT_2) {
src->state = dst->state = CT_DPIF_TCPS_CLOSED;
return CT_UPDATE_NEW;
}
if (src->wscale & CT_WSCALE_FLAG
&& dst->wscale & CT_WSCALE_FLAG
&& !(tcp_flags & TCP_SYN)) {
sws = src->wscale & CT_WSCALE_MASK;
dws = dst->wscale & CT_WSCALE_MASK;
} else if (src->wscale & CT_WSCALE_UNKNOWN
&& dst->wscale & CT_WSCALE_UNKNOWN
&& !(tcp_flags & TCP_SYN)) {
sws = TCP_MAX_WSCALE;
dws = TCP_MAX_WSCALE;
}
/*
* Sequence tracking algorithm from Guido van Rooij's paper:
* http://www.madison-gurkha.com/publications/tcp_filtering/
* tcp_filtering.ps
*/
orig_seq = seq = ntohl(tcp->seq);
if (src->state < CT_DPIF_TCPS_SYN_SENT) {
/* First packet from this end. Set its state */
ack = ntohl(tcp->ack_seq);
end = seq + p_len;
if (tcp_flags & TCP_SYN) {
end++;
if (dst->wscale & CT_WSCALE_FLAG) {
src->wscale = OvsTcpGetWscale(tcp);
if (src->wscale & CT_WSCALE_FLAG) {
/* Remove scale factor from initial window */
sws = src->wscale & CT_WSCALE_MASK;
win = DIV_ROUND_UP((uint32_t) win, 1 << sws);
dws = dst->wscale & CT_WSCALE_MASK;
} else {
/* fixup other window */
dst->max_win <<= dst->wscale & CT_WSCALE_MASK;
/* in case of a retrans SYN|ACK */
dst->wscale = 0;
}
}
}
if (tcp_flags & TCP_FIN) {
end++;
}
src->seqlo = seq;
src->state = CT_DPIF_TCPS_SYN_SENT;
/*
* May need to slide the window (seqhi may have been set by
* the crappy stack check or if we picked up the connection
* after establishment)
*/
if (src->seqhi == 1 ||
SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi)) {
src->seqhi = end + MAX(1, dst->max_win << dws);
}
if (win > src->max_win) {
src->max_win = win;
}
} else {
ack = ntohl(tcp->ack_seq);
end = seq + p_len;
if (tcp_flags & TCP_SYN) {
end++;
}
if (tcp_flags & TCP_FIN) {
end++;
}
}
if ((tcp_flags & TCP_ACK) == 0) {
/* Let it pass through the ack skew check */
ack = dst->seqlo;
} else if ((ack == 0
&& (tcp_flags & (TCP_ACK|TCP_RST)) == (TCP_ACK|TCP_RST))
/* broken tcp stacks do not set ack */) {
/* Many stacks (ours included) will set the ACK number in an
* FIN|ACK if the SYN times out -- no sequence to ACK. */
ack = dst->seqlo;
}
if (seq == end) {
/* Ease sequencing restrictions on no data packets */
seq = src->seqlo;
end = seq;
}
ackskew = dst->seqlo - ack;
#define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */
if (SEQ_GEQ(src->seqhi, end)
/* Last octet inside other's window space */
&& SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws))
/* Retrans: not more than one window back */
&& (ackskew >= -MAXACKWINDOW)
/* Acking not more than one reassembled fragment backwards */
&& (ackskew <= (MAXACKWINDOW << sws))
/* Acking not more than one window forward */
&& ((tcp_flags & TCP_RST) == 0 || orig_seq == src->seqlo
|| (orig_seq == src->seqlo + 1)
|| (orig_seq + 1 == src->seqlo))) {
/* Require an exact/+1 sequence match on resets when possible */
/* update max window */
if (src->max_win < win) {
src->max_win = win;
}
/* synchronize sequencing */
if (SEQ_GT(end, src->seqlo)) {
src->seqlo = end;
}
/* slide the window of what the other end can send */
if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) {
dst->seqhi = ack + MAX((win << sws), 1);
}
/* update states */
if (tcp_flags & TCP_SYN && src->state < CT_DPIF_TCPS_SYN_SENT) {
src->state = CT_DPIF_TCPS_SYN_SENT;
}
if (tcp_flags & TCP_FIN && src->state < CT_DPIF_TCPS_CLOSING) {
src->state = CT_DPIF_TCPS_CLOSING;
}
if (tcp_flags & TCP_ACK) {
if (dst->state == CT_DPIF_TCPS_SYN_SENT) {
dst->state = CT_DPIF_TCPS_ESTABLISHED;
} else if (dst->state == CT_DPIF_TCPS_CLOSING) {
dst->state = CT_DPIF_TCPS_FIN_WAIT_2;
}
}
if (tcp_flags & TCP_RST) {
src->state = dst->state = CT_DPIF_TCPS_TIME_WAIT;
}
if (src->state >= CT_DPIF_TCPS_FIN_WAIT_2
&& dst->state >= CT_DPIF_TCPS_FIN_WAIT_2) {
OvsConntrackUpdateExpiration(conn, now, 30 * CT_INTERVAL_SEC);
} else if (src->state >= CT_DPIF_TCPS_CLOSING
&& dst->state >= CT_DPIF_TCPS_CLOSING) {
OvsConntrackUpdateExpiration(conn, now, 45 * CT_INTERVAL_SEC);
} else if (src->state < CT_DPIF_TCPS_ESTABLISHED
|| dst->state < CT_DPIF_TCPS_ESTABLISHED) {
OvsConntrackUpdateExpiration(conn, now, 30 * CT_INTERVAL_SEC);
} else if (src->state >= CT_DPIF_TCPS_CLOSING
|| dst->state >= CT_DPIF_TCPS_CLOSING) {
OvsConntrackUpdateExpiration(conn, now, 15 * 60 * CT_INTERVAL_SEC);
} else {
OvsConntrackUpdateExpiration(conn, now, 24 * 60 * 60 * CT_INTERVAL_SEC);
}
} else if ((dst->state < CT_DPIF_TCPS_SYN_SENT
|| dst->state >= CT_DPIF_TCPS_FIN_WAIT_2
|| src->state >= CT_DPIF_TCPS_FIN_WAIT_2)
&& SEQ_GEQ(src->seqhi + MAXACKWINDOW, end)
/* Within a window forward of the originating packet */
&& SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) {
/* Within a window backward of the originating packet */
/*
* This currently handles three situations:
* 1) Stupid stacks will shotgun SYNs before their peer
* replies.
* 2) When PF catches an already established stream (the
* firewall rebooted, the state table was flushed, routes
* changed...)
* 3) Packets get funky immediately after the connection
* closes (this should catch Solaris spurious ACK|FINs
* that web servers like to spew after a close)
*
* This must be a little more careful than the above code
* since packet floods will also be caught here. We don't
* update the TTL here to mitigate the damage of a packet
* flood and so the same code can handle awkward establishment
* and a loosened connection close.
* In the establishment case, a correct peer response will
* validate the connection, go through the normal state code
* and keep updating the state TTL.
*/
/* update max window */
if (src->max_win < win) {
src->max_win = win;
}
/* synchronize sequencing */
if (SEQ_GT(end, src->seqlo)) {
src->seqlo = end;
}
/* slide the window of what the other end can send */
if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) {
dst->seqhi = ack + MAX((win << sws), 1);
}
/*
* Cannot set dst->seqhi here since this could be a shotgunned
* SYN and not an already established connection.
*/
if (tcp_flags & TCP_FIN && src->state < CT_DPIF_TCPS_CLOSING) {
src->state = CT_DPIF_TCPS_CLOSING;
}
if (tcp_flags & TCP_RST) {
src->state = dst->state = CT_DPIF_TCPS_TIME_WAIT;
}
} else {
return CT_UPDATE_INVALID;
}
return CT_UPDATE_VALID;
}
/* Callback function executed when something is received on fd */
int receive_callback(int file, void *arg) {
char buf[sizeof(struct rudp_packet)];
struct sockaddr_in sender;
size_t sender_length = sizeof(struct sockaddr_in);
recvfrom(file, &buf, sizeof(struct rudp_packet), 0, (struct sockaddr *)&sender, &sender_length);
struct rudp_packet *received_packet = malloc(sizeof(struct rudp_packet));
if(received_packet == NULL) {
fprintf(stderr, "receive_callback: Error allocating packet\n");
return -1;
}
memcpy(received_packet, &buf, sizeof(struct rudp_packet));
struct rudp_hdr rudpheader = received_packet->header;
char type[5];
short t = rudpheader.type;
if(t == 1)
strcpy(type, "DATA");
else if(t == 2)
strcpy(type, "ACK");
else if(t == 4)
strcpy(type, "SYN");
else if(t==5)
strcpy(type, "FIN");
else
strcpy(type, "BAD");
printf("Received %s packet from %s:%d seq number=%u on socket=%d\n",type,
inet_ntoa(sender.sin_addr), ntohs(sender.sin_port),rudpheader.seqno,file);
/* Locate the correct socket in the socket list */
if(socket_list_head == NULL) {
fprintf(stderr, "Error: attempt to receive on invalid socket. No sockets in the list\n");
return -1;
}
else {
/* We have sockets to check */
struct rudp_socket_list *curr_socket = socket_list_head;
while(curr_socket != NULL) {
if((int)curr_socket->rsock == file) {
break;
}
curr_socket = curr_socket->next;
}
if((int)curr_socket->rsock == file) {
/* We found the correct socket, now see if a session already exists for this peer */
if(curr_socket->sessions_list_head == NULL) {
/* The list is empty, so we check if the sender has initiated the protocol properly (by sending a SYN) */
if(rudpheader.type == RUDP_SYN) {
/* SYN Received. Create a new session at the head of the list */
u_int32_t seqno = rudpheader.seqno + 1;
create_receiver_session(curr_socket, seqno, &sender);
/* Respond with an ACK */
struct rudp_packet *p = create_rudp_packet(RUDP_ACK, seqno, 0, NULL);
send_packet(true, (rudp_socket_t)file, p, &sender);
free(p);
}
else {
/* No sessions exist and we got a non-SYN, so ignore it */
}
}
else {
/* Some sessions exist to be checked */
bool_t session_found = false;
struct session *curr_session = curr_socket->sessions_list_head;
struct session *last_session;
while(curr_session != NULL) {
if(curr_session->next == NULL) {
last_session = curr_session;
}
if(compare_sockaddr(&curr_session->address, &sender) == 1) {
/* Found an existing session */
session_found = true;
break;
}
curr_session = curr_session->next;
}
if(session_found == false) {
/* No session was found for this peer */
if(rudpheader.type == RUDP_SYN) {
/* SYN Received. Send an ACK and create a new session */
u_int32_t seqno = rudpheader.seqno + 1;
create_receiver_session(curr_socket, seqno, &sender);
struct rudp_packet *p = create_rudp_packet(RUDP_ACK, seqno, 0, NULL);
send_packet(true, (rudp_socket_t)file, p, &sender);
free(p);
}
else {
/* Session does not exist and non-SYN received - ignore it */
}
}
else {
/* We found a matching session */
if(rudpheader.type == RUDP_SYN) {
if(curr_session->receiver == NULL || curr_session->receiver->status == OPENING) {
/* Create a new receiver session and ACK the SYN*/
struct receiver_session *new_receiver_session = malloc(sizeof(struct receiver_session));
if(new_receiver_session == NULL) {
fprintf(stderr, "receive_callback: Error allocating receiver session\n");
return -1;
}
new_receiver_session->expected_seqno = rudpheader.seqno + 1;
new_receiver_session->status = OPENING;
new_receiver_session->session_finished = false;
curr_session->receiver = new_receiver_session;
u_int32_t seqno = curr_session->receiver->expected_seqno;
struct rudp_packet *p = create_rudp_packet(RUDP_ACK, seqno, 0, NULL);
send_packet(true, (rudp_socket_t)file, p, &sender);
free(p);
}
else {
/* Received a SYN when there is already an active receiver session, so we ignore it */
}
}
if(rudpheader.type == RUDP_ACK) {
u_int32_t ack_sqn = received_packet->header.seqno;
if(curr_session->sender->status == SYN_SENT) {
/* This an ACK for a SYN */
u_int32_t syn_sqn = curr_session->sender->seqno;
if( (ack_sqn - 1) == syn_sqn) {
/* Delete the retransmission timeout */
event_timeout_delete(timeout_callback, curr_session->sender->syn_timeout_arg);
struct timeoutargs *t = (struct timeoutargs *)curr_session->sender->syn_timeout_arg;
free(t->packet);
free(t->recipient);
free(t);
curr_session->sender->status = OPEN;
while(curr_session->sender->data_queue != NULL) {
/* Check if the window is already full */
if(curr_session->sender->sliding_window[RUDP_WINDOW-1] != NULL) {
break;
}
else {
int index;
int i;
/* Find the first unused window slot */
for(i = RUDP_WINDOW-1; i >= 0; i--) {
if(curr_session->sender->sliding_window[i] == NULL) {
index = i;
}
}
/* Send packet, add to window and remove from queue */
u_int32_t seqno = ++syn_sqn;
int len = curr_session->sender->data_queue->len;
char *payload = curr_session->sender->data_queue->item;
struct rudp_packet *datap = create_rudp_packet(RUDP_DATA, seqno, len, payload);
curr_session->sender->seqno += 1;
curr_session->sender->sliding_window[index] = datap;
curr_session->sender->retransmission_attempts[index] = 0;
struct data *temp = curr_session->sender->data_queue;
curr_session->sender->data_queue = curr_session->sender->data_queue->next;
free(temp->item);
free(temp);
send_packet(false, (rudp_socket_t)file, datap, &sender);
}
}
}
}
else if(curr_session->sender->status == OPEN) {
/* This is an ACK for DATA */
if(curr_session->sender->sliding_window[0] != NULL) {
if(curr_session->sender->sliding_window[0]->header.seqno == (rudpheader.seqno-1)) {
/* Correct ACK received. Remove the first window item and shift the rest left */
event_timeout_delete(timeout_callback, curr_session->sender->data_timeout_arg[0]);
struct timeoutargs *args = (struct timeoutargs *)curr_session->sender->data_timeout_arg[0];
free(args->packet);
free(args->recipient);
free(args);
free(curr_session->sender->sliding_window[0]);
int i;
if(RUDP_WINDOW == 1) {
curr_session->sender->sliding_window[0] = NULL;
curr_session->sender->retransmission_attempts[0] = 0;
curr_session->sender->data_timeout_arg[0] = NULL;
}
else {
for(i = 0; i < RUDP_WINDOW - 1; i++) {
curr_session->sender->sliding_window[i] = curr_session->sender->sliding_window[i+1];
curr_session->sender->retransmission_attempts[i] = curr_session->sender->retransmission_attempts[i+1];
curr_session->sender->data_timeout_arg[i] = curr_session->sender->data_timeout_arg[i+1];
if(i == RUDP_WINDOW-2) {
curr_session->sender->sliding_window[i+1] = NULL;
curr_session->sender->retransmission_attempts[i+1] = 0;
curr_session->sender->data_timeout_arg[i+1] = NULL;
}
}
}
while(curr_session->sender->data_queue != NULL) {
if(curr_session->sender->sliding_window[RUDP_WINDOW-1] != NULL) {
break;
}
else {
int index;
int i;
/* Find the first unused window slot */
for(i = RUDP_WINDOW-1; i >= 0; i--) {
if(curr_session->sender->sliding_window[i] == NULL) {
index = i;
}
}
/* Send packet, add to window and remove from queue */
curr_session->sender->seqno = curr_session->sender->seqno + 1;
u_int32_t seqno = curr_session->sender->seqno;
int len = curr_session->sender->data_queue->len;
char *payload = curr_session->sender->data_queue->item;
struct rudp_packet *datap = create_rudp_packet(RUDP_DATA, seqno, len, payload);
curr_session->sender->sliding_window[index] = datap;
curr_session->sender->retransmission_attempts[index] = 0;
struct data *temp = curr_session->sender->data_queue;
curr_session->sender->data_queue = curr_session->sender->data_queue->next;
free(temp->item);
free(temp);
send_packet(false, (rudp_socket_t)file, datap, &sender);
}
}
if(curr_socket->close_requested) {
/* Can the socket be closed? */
struct session *head_sessions = curr_socket->sessions_list_head;
while(head_sessions != NULL) {
if(head_sessions->sender->session_finished == false) {
if(head_sessions->sender->data_queue == NULL &&
head_sessions->sender->sliding_window[0] == NULL &&
head_sessions->sender->status == OPEN) {
head_sessions->sender->seqno += 1;
struct rudp_packet *p = create_rudp_packet(RUDP_FIN, head_sessions->sender->seqno, 0, NULL);
send_packet(false, (rudp_socket_t)file, p, &head_sessions->address);
free(p);
head_sessions->sender->status = FIN_SENT;
}
}
head_sessions = head_sessions->next;
}
}
}
}
}
else if(curr_session->sender->status == FIN_SENT) {
/* Handle ACK for FIN */
if( (curr_session->sender->seqno + 1) == received_packet->header.seqno) {
event_timeout_delete(timeout_callback, curr_session->sender->fin_timeout_arg);
struct timeoutargs *t = curr_session->sender->fin_timeout_arg;
free(t->packet);
free(t->recipient);
free(t);
curr_session->sender->session_finished = true;
if(curr_socket->close_requested) {
/* See if we can close the socket */
struct session *head_sessions = curr_socket->sessions_list_head;
bool_t all_done = true;
while(head_sessions != NULL) {
if(head_sessions->sender->session_finished == false) {
all_done = false;
}
else if(head_sessions->receiver != NULL && head_sessions->receiver->session_finished == false) {
all_done = false;
}
else {
free(head_sessions->sender);
if(head_sessions->receiver) {
free(head_sessions->receiver);
}
}
struct session *temp = head_sessions;
head_sessions = head_sessions->next;
free(temp);
}
if(all_done) {
if(curr_socket->handler != NULL) {
curr_socket->handler((rudp_socket_t)file, RUDP_EVENT_CLOSED, &sender);
event_fd_delete(receive_callback, (rudp_socket_t)file);
close(file);
free(curr_socket);
}
}
}
}
else {
/* Received incorrect ACK for FIN - ignore it */
}
}
}
else if(rudpheader.type == RUDP_DATA) {
/* Handle DATA packet. If the receiver is OPENING, it can transition to OPEN */
if(curr_session->receiver->status == OPENING) {
if(rudpheader.seqno == curr_session->receiver->expected_seqno) {
curr_session->receiver->status = OPEN;
}
}
if(rudpheader.seqno == curr_session->receiver->expected_seqno) {
/* Sequence numbers match - ACK the data */
u_int32_t seqno = rudpheader.seqno + 1;
curr_session->receiver->expected_seqno = seqno;
struct rudp_packet *p = create_rudp_packet(RUDP_ACK, seqno, 0, NULL);
send_packet(true, (rudp_socket_t)file, p, &sender);
free(p);
/* Pass the data up to the application */
if(curr_socket->recv_handler != NULL)
curr_socket->recv_handler((rudp_socket_t)file, &sender,
(void*)&received_packet->payload, received_packet->payload_length);
}
/* Handle the case where an ACK was lost */
else if(SEQ_GEQ(rudpheader.seqno, (curr_session->receiver->expected_seqno - RUDP_WINDOW)) &&
SEQ_LT(rudpheader.seqno, curr_session->receiver->expected_seqno)) {
u_int32_t seqno = rudpheader.seqno + 1;
struct rudp_packet *p = create_rudp_packet(RUDP_ACK, seqno, 0, NULL);
send_packet(true, (rudp_socket_t)file, p, &sender);
free(p);
}
}
else if(rudpheader.type == RUDP_FIN) {
if(curr_session->receiver->status == OPEN) {
if(rudpheader.seqno == curr_session->receiver->expected_seqno) {
/* If the FIN is correct, we can ACK it */
u_int32_t seqno = curr_session->receiver->expected_seqno + 1;
struct rudp_packet *p = create_rudp_packet(RUDP_ACK, seqno, 0, NULL);
send_packet(true, (rudp_socket_t)file, p, &sender);
free(p);
curr_session->receiver->session_finished = true;
if(curr_socket->close_requested) {
/* Can we close the socket now? */
struct session *head_sessions = curr_socket->sessions_list_head;
int all_done = true;
while(head_sessions != NULL) {
if(head_sessions->sender->session_finished == false) {
all_done = false;
}
else if(head_sessions->receiver != NULL && head_sessions->receiver->session_finished == false) {
all_done = false;
}
else {
free(head_sessions->sender);
if(head_sessions->receiver) {
free(head_sessions->receiver);
}
}
struct session *temp = head_sessions;
head_sessions = head_sessions->next;
free(temp);
}
if(all_done) {
if(curr_socket->handler != NULL) {
curr_socket->handler((rudp_socket_t)file, RUDP_EVENT_CLOSED, &sender);
event_fd_delete(receive_callback, (rudp_socket_t)file);
close(file);
free(curr_socket);
}
}
}
}
else {
/* FIN received with incorrect sequence number - ignore it */
}
}
}
}
}
}
}
free(received_packet);
return 0;
}