int
tcp_rcv(PACKET pkt) /* NOTE: pkt has nb_prot pointing to IP header */
{
struct mbuf * m_in;
struct ip * bip; /* IP header, berkeley version */
struct tcphdr * tcpp;
unshort len; /* scratch length holder */
/* For TCP, the netport IP layer is modified to set nb_prot to the
* start of the IP header (not TCP). We need to do some further
* mods which the BSD code expects:
*/
bip = (struct ip *)pkt->nb_prot; /* get ip header */
len = ntohs(bip->ip_len); /* get length in local endian */
/* verify checksum of received packet */
tcpp = (struct tcphdr *)ip_data(bip);
if (tcp_cksum(bip) != tcpp->th_sum)
{
TCP_MIB_INC(tcpInErrs); /* keep MIB stats */
tcpstat.tcps_rcvbadsum++; /* keep BSD stats */
LOCK_NET_RESOURCE(FREEQ_RESID);
pk_free(pkt); /* punt packet */
UNLOCK_NET_RESOURCE(FREEQ_RESID);
return ENP_BAD_HEADER;
}
m_in = m_getnbuf(MT_RXDATA, 0);
if (!m_in){
LOCK_NET_RESOURCE(FREEQ_RESID);
pk_free(pkt);
UNLOCK_NET_RESOURCE(FREEQ_RESID);
return ENP_RESOURCE;
}
IN_PROFILER(PF_TCP, PF_ENTRY); /* measure time in TCP */
/* subtract IP header length from total IP packet length */
len -= ((unshort)(bip->ip_ver_ihl & 0x0f) << 2);
bip->ip_len = len; /* put TCP length in struct for TCP code to use */
/* set mbuf to point to start of IP header (not TCP) */
m_in->pkt = pkt;
m_in->m_data = pkt->nb_prot;
m_in->m_len = pkt->nb_plen;
m_in->m_base = pkt->nb_buff; /* ??? */
m_in->m_memsz = pkt->nb_blen; /* ??? */
tcp_input(m_in, pkt->net);
IN_PROFILER(PF_TCP, PF_EXIT); /* measure time in TCP */
return 0;
}
int
tcp_output(struct tcpcb * tp)
{
struct socket * so = tp->t_inpcb->inp_socket;
int len;
long win;
int off, flags, error;
struct mbuf * m;
struct tcpiphdr * ti;
unsigned optlen = 0;
int idle, sendalot;
struct mbuf * sendm; /* mbuf which contains data to send */
struct mbuf * tcp_mbuf; /* mbuf containing TCP header */
int bufoff; /* offset of data in sendm->m_data */
#ifdef TCP_SACK
int sack_resend;
int sack_hole = 0; /* next sack hole to fill */
if(tp->t_flags & TF_SACKREPLY)
{
/* we are resending based on a received SACK header */
sack_resend = TRUE;
tp->t_flags &= ~TF_SACKREPLY; /* clear flag */
}
else
sack_resend = FALSE;
#endif /* TCP_SACK */
/*
* 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->snd_max == tp->snd_una);
again:
sendalot = 0;
off = (int)(tp->snd_nxt - tp->snd_una);
win = (long)tp->snd_wnd; /* set basic send window */
if (win > (long)tp->snd_cwnd) /* see if we need congestion control */
{
win = (int)(tp->snd_cwnd & ~(ALIGN_TYPE-1)); /* keep data aligned */
}
/*
* 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 (win == 0)
win = 1;
else
{
tp->t_timer[TCPT_PERSIST] = 0;
tp->t_rxtshift = 0;
}
}
#ifdef TCP_SACK
/* See if we need to adjust the offset for a sack resend */
if(sack_resend)
{
off = (int)(tp->sack_hole_start[sack_hole] - tp->snd_una);
/* if this hole's already been acked then punt and move to next hole */
if(off < 0)
{
/* clear out the acked hole */
tp->sack_hole_start[sack_hole] = tp->sack_hole_end[sack_hole] = 0;
/* see if we're done with SACK hole list (2 tests) */
if(++sack_hole >= SACK_BLOCKS)
return 0;
if(tp->sack_hole_start[sack_hole] == tp->sack_hole_end[sack_hole])
return 0;
goto again;
}
tp->snd_nxt = tp->sack_hole_start[sack_hole];
len = (int)(tp->sack_hole_end[sack_hole] - tp->sack_hole_start[sack_hole]);
len = (int)MIN(len, (int)win);
}
else
#endif /* TCP_SACK */
{
/* set length of packets which are not sack resends */
len = (int)MIN(so->so_snd.sb_cc, (unsigned)win) - off;
}
flags = tcp_outflags[tp->t_state];
/* See if we need to build TCP options field. This test should be fast. */
#if (defined(TCP_TIMESTAMP) | defined(TCP_SACK))
if((flags & TH_SYN) ||
/* !!!??? (so->so_options & SO_TIMESTAMP) || */
(tp->t_flags & TF_SACKNOW)
)
{
optlen = bld_options(tp, &tcp_optionbuf[optlen], flags, so);
}
#else
/* If other options not defined this build then don't bother to call bld_options() except
* on SYN packets
*/
if(flags & TH_SYN)
{
optlen = bld_options(tp, &tcp_optionbuf[optlen], flags, so);
}
#endif
if (len < 0)
{
/*
* If FIN has been sent but not acked,
* but we haven't been called to retransmit,
* len will be -1. Otherwise, window shrank
* after we sent into it. If window shrank to 0,
* cancel pending retransmit and pull snd_nxt
* back to (closed) window. We will enter persist
* state below. If the window didn't close completely,
* just wait for an ACK.
*/
len = 0;
if (win == 0)
{
tp->t_timer[TCPT_REXMT] = 0;
tp->snd_nxt = tp->snd_una;
}
}
if (len > (int)tp->t_maxseg)
{
len = tp->t_maxseg;
sendalot = 1;
}
#ifdef IP_V4
#ifdef IP_PMTU
{
int pmtu = tp->t_inpcb->inp_pmtu - 40;
if (len > pmtu)
{
len = pmtu - 40;
sendalot = 1;
}
}
#endif /* IP_PMTU */
/* We don't need a pmtu test for IPv6. V6 code limits t_maxseg to
* the Path MTU, so the test above the v4 ifdef above covers us.
*/
#endif /* IP_V4 */
if (SEQ_LT(tp->snd_nxt + len, tp->snd_una + so->so_snd.sb_cc))
flags &= ~TH_FIN;
win = (long)(sbspace(&so->so_rcv));
/*
* If our state indicates that FIN should be sent
* and we have not yet done so, or we're retransmitting the FIN,
* then we need to send.
*/
if ((flags & TH_FIN) &&
(so->so_snd.sb_cc == 0) &&
((tp->t_flags & TF_SENTFIN) == 0 || tp->snd_nxt == tp->snd_una))
{
goto send;
}
/*
* Send if we owe peer an ACK.
*/
if (tp->t_flags & TF_ACKNOW)
goto send;
if (flags & (TH_SYN|TH_RST))
goto send;
if (SEQ_GT(tp->snd_up, tp->snd_una))
goto send;
/*
* Sender silly window avoidance. If connection is idle
* and can send all data, a maximum segment,
* at least a maximum default-size segment do it,
* or are forced, do it; otherwise don't bother.
* If peer's buffer is tiny, then send
* when window is at least half open.
* If retransmitting (possibly after persist timer forced us
* to send into a small window), then must resend.
*/
if (len)
{
if (len == (int)tp->t_maxseg)
goto send;
if ((idle || tp->t_flags & TF_NODELAY) &&
len + off >= (int)so->so_snd.sb_cc)
{
goto send;
}
if (tp->t_force)
goto send;
if (len >= (int)(tp->max_sndwnd / 2))
goto send;
if (SEQ_LT(tp->snd_nxt, tp->snd_max))
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 35% of the maximum possible
* window, then want to send a window update to peer.
*/
if (win > 0)
{
int adv = (int)win - (int)(tp->rcv_adv - tp->rcv_nxt);
if (so->so_rcv.sb_cc == 0 && adv >= (int)(tp->t_maxseg * 2))
goto send;
if (100 * (u_int)adv / so->so_rcv.sb_hiwat >= 35)
goto send;
}
/*
* 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
*
* tp->t_timer[TCPT_PERSIST]
* is set when we are in persist state.
* tp->t_force
* is set when we are called to send a persist packet.
* tp->t_timer[TCPT_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 && tp->t_timer[TCPT_REXMT] == 0 &&
tp->t_timer[TCPT_PERSIST] == 0)
{
tp->t_rxtshift = 0;
tcp_setpersist(tp);
}
/*
* No reason to send a segment, just return.
*/
return (0);
send:
ENTER_CRIT_SECTION(tp);
/* Limit send length to the current buffer so as to
* avoid doing the "mbuf shuffle" in m_copy().
*/
bufoff = off;
sendm = so->so_snd.sb_mb;
if (len)
{
/* find mbuf containing data to send (at "off") */
while (sendm) /* loop through socket send list */
{
bufoff -= sendm->m_len;
if (bufoff < 0) /* if off is in this buffer, break */
break;
sendm = sendm->m_next;
}
if (!sendm) { dtrap(); /* shouldn't happen */ }
bufoff += sendm->m_len; /* index to next data to send in msend */
/* if socket has multiple unsent mbufs, set flag for send to loop */
if ((sendm->m_next) && (len > (int)sendm->m_len))
{
flags &= ~TH_FIN; /* don't FIN on segment prior to last */
sendalot = 1; /* set to send more segments */
}
if((flags & TH_FIN) && (so->so_snd.sb_cc > (unsigned)len))
{
/* This can happen on slow links (PPP) which retry the last
* segment - the one with the FIN bit attached to data.
*/
flags &= ~TH_FIN; /* don't FIN on segment prior to last */
}
/* only send the rest of msend */
len = min(len, (int)sendm->m_len);
/* if we're not sending starting at sendm->m_data (in which
* case bufoff != 0), then we will copy the data; else we would
* write IP/TCP headers over sent but un-ack'ed data in sendm.
* Similarly, if sendm->m_data is not aligned with respect to
* sendm->m_base and ALIGN_TYPE, we will copy the data to
* ensure that it (and the then-prepended IP/TCP headers) will
* be aligned according to ALIGN_TYPE.
*/
if ((bufoff != 0) || /* data not front aligned in send mbuf? */
(((sendm->m_data - sendm->m_base) & (ALIGN_TYPE - 1)) != 0))
{
len = min(len, (int)(sendm->m_len - bufoff)); /* limit len again */
/* One more test - if this data is not aligned with the front
* of the m_data buffer then we can't use it in place, else we
* might write the IP/TCP header over data that has not yet
* been acked. In this case we must make sure our send
* fits into a little buffer and send what we can.
*/
if ((len > (int)(lilbufsiz - HDRSLEN)) && /* length is bigger the small buffer? */
(bigfreeq.q_len < 2)) /* and we are low on big buffers */
{
len = lilbufsiz - HDRSLEN;
}
}
}
/* if send data is sufficiently aligned in packet, prepend TCP/IP header
* in the space provided.
*/
if (len && (bufoff == 0) &&
(sendm->pkt->inuse == 1) &&
(((sendm->m_data - sendm->m_base) & (ALIGN_TYPE - 1)) == 0) &&
(optlen == 0))
{
/* get an empty mbuf to "clone" the data */
m = m_getnbuf(MT_TXDATA, 0);
if (!m)
{
EXIT_CRIT_SECTION(tp);
return (ENOBUFS);
}
m->pkt = sendm->pkt; /* copy packet location in new mbuf */
m->pkt->inuse++; /* bump packet's use count */
m->m_base = sendm->m_base; /* clone mbuf members */
m->m_memsz = sendm->m_memsz;
m->m_len = len + TCPIPHDRSZ; /* adjust clone for header */
m->m_data = sendm->m_data - TCPIPHDRSZ;
}
else /* either no data or data is not front aligned in mbuf */
{
/* Grab a header mbuf, attaching a copy of data to be
* transmitted, and initialize the header from
* the template for sends on this connection.
*/
m = m_getwithdata (MT_HEADER, IFNETHDR_SIZE + TCPIPHDRSZ);
if (m ==(struct mbuf *)NULL)
{
EXIT_CRIT_SECTION(tp);
return ENOBUFS;
}
m->m_len = TCPIPHDRSZ;
m->m_data += IFNETHDR_SIZE;/* Move this to sizeof tcpip hdr leave*/
/* 14 bytes for ethernet header */
if (len) /* attach any data to send */
{
m->m_next = m_copy(so->so_snd.sb_mb, off, (int) len);
if (m->m_next == 0)
{
m_freem(m);
EXIT_CRIT_SECTION(tp);
return ENOBUFS;
}
}
}
EXIT_CRIT_SECTION(tp);
if (len)
{
if (tp->t_force && len == 1)
tcpstat.tcps_sndprobe++;
else if (SEQ_LT(tp->snd_nxt, tp->snd_max))
{
tcpstat.tcps_sndrexmitpack++;
tcpstat.tcps_sndrexmitbyte += len;
#ifdef TCP_SACK
if(sack_resend)
tcpstat.tcps_sackresend++;
#endif
}
else
{
tcpstat.tcps_sndpack++;
tcpstat.tcps_sndbyte += len;
}
}
else if (tp->t_flags & TF_ACKNOW)
{
tcpstat.tcps_sndacks++;
}
else if (flags & (TH_SYN|TH_FIN|TH_RST))
tcpstat.tcps_sndctrl++;
else if (SEQ_GT(tp->snd_up, tp->snd_una))
tcpstat.tcps_sndurg++;
else
tcpstat.tcps_sndwinup++;
ti = (struct tcpiphdr *)(m->m_data+sizeof(struct ip)-sizeof(struct ipovly));
if ((char *)ti < m->pkt->nb_buff)
{
panic("tcp_out- packet ptr underflow\n");
}
tcp_mbuf = m; /* flag TCP header mbuf */
#ifdef IP_V6 /* Dual mode code */
if(so->so_domain == AF_INET6)
{
m = mbuf_prepend(m, sizeof(struct ipv6));
if(m == NULL)
{
/* this can happen when we run out of mbufs or pkt buffers
* That is, mfreeq is empty or (lilfreeq, bigfreeq) are empty.
* One solution is to find out which one is getting full and
* then increase them.
*/
dtrap(); /* This is really rare... */
m_freem(tcp_mbuf); /* Free TCP/data chain */
return ENOBUFS;
}
/* strip overlay from front of TCP header */
tcp_mbuf->m_data += sizeof(struct ipovly);
tcp_mbuf->m_len -= sizeof(struct ipovly);
}
#endif /* end IP_V6 */
if (tp->t_template == 0)
panic("tcp_output");
MEMCPY((char*)ti, (char*)tp->t_template, sizeof(struct tcpiphdr));
/*
* Fill in fields, remembering maximum advertised
* window for use in delaying messages about window sizes.
* If resending a FIN, be sure not to use a new sequence number.
*/
if (flags & TH_FIN && tp->t_flags & TF_SENTFIN &&
tp->snd_nxt == tp->snd_max)
{
tp->snd_nxt--;
}
ti->ti_seq = htonl(tp->snd_nxt);
ti->ti_ack = htonl(tp->rcv_nxt);
/*
* If we're sending a SYN, check the IP address of the interface
* that we will (likely) use to send the IP datagram -- if it's
* changed from what is in the template (as it might if this is
* a retransmission, and the original SYN caused PPP to start
* bringing the interface up, and PPP has got a new IP address
* via IPCP), update the template and the inpcb with the new
* address.
*/
if (flags & TH_SYN)
{
struct inpcb * inp;
inp = (struct inpcb *)so->so_pcb;
switch(so->so_domain)
{
#ifdef IP_V4
case AF_INET:
{
ip_addr src;
#ifdef INCLUDE_PPP
if(((flags & TH_ACK) == 0) && /* SYN only, not SYN/ACK */
(inp->ifp) && /* Make sure we have iface */
(inp->ifp->mib.ifType == PPP)) /* only PPP type */
{
dtrap(); /* remove after confirmed to work in PPP */
src = ip_mymach(ti->ti_dst.s_addr);
if (src != ti->ti_src.s_addr)
{
ti->ti_src.s_addr = src;
tp->t_template->ti_src.s_addr = src;
tp->t_inpcb->inp_laddr.s_addr = src;
}
}
#endif /* INCLUDE_PPP */
/* If this is a SYN (not a SYN/ACK) then set the pmtu */
if((flags & TH_ACK) == 0)
{
#ifdef IP_PMTU
inp->inp_pmtu = pmtucache_get(inp->inp_faddr.s_addr);
#else /* not compiled for pathmtu, guess based on iface */
{
NET ifp;
/* find iface for route. Pass "src" as nexthop return */
ifp = iproute(ti->ti_dst.s_addr, &src);
if(ifp)
inp->inp_pmtu = ifp->n_mtu - (ifp->n_lnh + 40);
else
inp->inp_pmtu = 580; /* Ugh. */
}
#endif /* IP_PMTU */
}
break;
}
#endif /* IP_V4 */
#ifdef IP_V6
case AF_INET6:
{
struct ip6_inaddr * local;
local = ip6_myaddr(&tp->t_inpcb->ip6_faddr, inp->ifp);
/* If we got a local address & it's not the one in the pcb, then
* we assume it changed at the iface and fix it in the pcb. Unlike
* v4, we don't have an IP header yet, not do we have a template
* to worry about.
*/
if((local) &&
(!IP6EQ(&local->addr, &tp->t_inpcb->ip6_laddr)))
{
IP6CPY(&tp->t_inpcb->ip6_laddr, &local->addr);
}
/* If this is a SYN (not a SYN/ACK) then set the pmtu */
if((flags & TH_ACK) == 0)
{
inp->inp_pmtu = ip6_pmtulookup(&inp->ip6_laddr, inp->ifp);
}
break;
}
#endif /* IP_V6 */
default:
dtrap(); /* bad domain setting */
}
}
/* fill in options if any are set */
if (optlen)
{
struct mbuf * mopt;
mopt = m_getwithdata(MT_TXDATA, MAXOPTLEN);
if (mopt == NULL)
{
m_freem(m);
return (ENOBUFS);
}
/* insert options mbuf after after tmp_mbuf */
mopt->m_next = tcp_mbuf->m_next;
tcp_mbuf->m_next = mopt;
/* extend options to aligned address */
while(optlen & 0x03)
tcp_optionbuf[optlen++] = TCPOPT_EOL;
MEMCPY(mtod(mopt, char *), tcp_optionbuf, optlen);
mopt->m_len = optlen;
/* use portable macro to set tcp data offset bits */
SET_TH_OFF(ti->ti_t, ((sizeof (struct tcphdr) + optlen) >> 2));
}
ti->ti_flags = (u_char)flags;
/*
* Calculate receive window. Don't shrink window,
* but avoid silly window syndrome.
*/
if (win < (long)(so->so_rcv.sb_hiwat / 4) && win < (long)tp->t_maxseg)
win = 0;
if (win < (long)(tp->rcv_adv - tp->rcv_nxt))
win = (long)(tp->rcv_adv - tp->rcv_nxt);
/* do check for Iniche buffer limits -JB- */
if (bigfreeq.q_len == 0) /* If queue length is 0, set window to 0 */
{
win = 0;
}
else if(win > (((long)bigfreeq.q_len - 1) * (long)bigbufsiz))
{
win = ((long)bigfreeq.q_len - 1) * bigbufsiz;
}
#ifdef TCP_WIN_SCALE
if(tp->t_flags & TF_WINSCALE)
{
ti->ti_win = htons((u_short)(win >> tp->rcv_wind_scale)); /* apply scale */
}
int
rawip_soinput(PACKET pkt, void * so_ptr)
{
struct mbuf * m_in; /* packet/data mbuf */
struct socket * so = (struct socket *)so_ptr;
struct sockaddr_in sin;
LOCK_NET_RESOURCE(NET_RESID);
/* make sure we're not flooding input buffers */
if ((so->so_rcv.sb_cc + pkt->nb_plen) >= so->so_rcv.sb_hiwat)
{
UNLOCK_NET_RESOURCE(NET_RESID);
return ENOBUFS;
}
/* alloc mbuf for received data */
m_in = m_getnbuf(MT_RXDATA, 0);
if (!m_in)
{
UNLOCK_NET_RESOURCE(NET_RESID);
return ENOBUFS;
}
/* set data mbuf to point to start of IP header */
m_in->pkt = pkt;
m_in->m_base = pkt->nb_buff;
m_in->m_memsz = pkt->nb_blen;
m_in->m_data = pkt->nb_prot;
m_in->m_len = pkt->nb_plen;
/* if this socket doesn't have IP_HDRINCL set, adjust the
* mbuf to skip past the IP header
*/
if (!(so->so_options & SO_HDRINCL))
{
unsigned int ihl =
(((struct ip *)(pkt->nb_prot))->ip_ver_ihl & 0x0f) << 2;
m_in->m_data += ihl;
m_in->m_len -= ihl;
}
/* fill in net address info for pass to socket append()ers */
sin.sin_addr.s_addr = pkt->fhost;
sin.sin_port = 0;
sin.sin_family = AF_INET;
/* attempt to append address information to mbuf */
if (!sbappendaddr(&so->so_rcv, (struct sockaddr *)&sin, m_in))
{
/* set the pkt field in the mbuf to NULL so m_free() below wont
* free the packet buffer, because that is left to the
* underlying stack
*/
m_in->pkt = NULL;
/* free only the mbuf itself */
m_free(m_in);
/* return error condition so caller can free the packet buffer */
UNLOCK_NET_RESOURCE(NET_RESID);
return ENOBUFS;
}
tcp_wakeup(&so->so_rcv); /* wake anyone waiting for this */
sorwakeup(so); /* wake up selects too */
UNLOCK_NET_RESOURCE(NET_RESID);
return 0;
}
