int tcp_bind(struct tcp_pcb * __tp, in_addr_t __addr, uint16_t __port)
{
if (__tp == NULL) {
DCC_LOG(LOG_WARNING, "NULL pointer");
/* FIXME: not a socket? The semantic here is not exactly the same
as the sockets API. */
return -ENOTSOCK;
}
if (pcb_find((struct pcb *)__tp, &__tcp__.closed) < 0) {
DCC_LOG1(LOG_ERROR, "<%04x> pcb_find()", (int)__tp);
return -ENOTSOCK;
}
tcpip_net_lock();
if (__port == 0) {
do {
/* generate an ephemeral port number from 1024 to 33791 */
__port = ntohs(((__tcp__.port_seq++) & 0x7fff) + 1024);
} while (!can_bind(__addr, __port));
} else {
if (!can_bind(__addr, __port)) {
DCC_LOG3(LOG_WARNING, "<%04x> %I:%d in use",
(int)__tp, __addr, ntohs(__port));
return -EADDRINUSE;
}
}
DCC_LOG3(LOG_TRACE, "<%05x> %I:%d", (int)__tp, __addr, ntohs(__port));
__tp->t_lport = __port;
__tp->t_laddr = __addr;
tcpip_net_unlock();
return 0;
}
int tcp_close(struct tcp_pcb * __tp)
{
int ret;
if (__tp == NULL) {
DCC_LOG(LOG_WARNING, "NULL pointer");
return -1;
}
tcpip_net_lock();
#ifdef ENABLE_SANITY
if ((pcb_find((struct pcb *)__tp, &__tcp__.active) < 0) &&
(pcb_find((struct pcb *)__tp, &__tcp__.listen) < 0) &&
pcb_find((struct pcb *)__tp, &__tcp__.closed) < 0) {
DBG(DBG_ERROR, "<%05x> pcb_find()", (int)__tp);
tcpip_net_unlock();
return -1;
}
#endif
switch(__tp->t_state) {
case TCPS_LISTEN: {
ret = tcp_pcb_free(__tp);
tcpip_net_unlock();
return ret;
}
case TCPS_TIME_WAIT:
case TCPS_CLOSED:
case TCPS_SYN_SENT:
DCC_LOG2(LOG_TRACE, "<%05x> [%s]", (int)__tp,
__tcp_state[__tp->t_state]);
if (__tp->t_cond >= 0) {
__os_cond_free(__tp->t_cond);
__tp->t_cond = -1;
}
ret = tcp_pcb_free(__tp);
tcpip_net_unlock();
return ret;
/* active close */
case TCPS_SYN_RCVD:
case TCPS_ESTABLISHED:
/* Close the receive window */
/*
* XXX: if we close the receive window we may stuck at
* FIN_WAIT_2 state...
*/
// __tp->rcv_wnd = 0;
__tp->t_state = TCPS_FIN_WAIT_1;
DCC_LOG1(LOG_TRACE, "<%05x> [FIN_WAIT_1]", (int)__tp);
break;
/* passive close */
case TCPS_CLOSE_WAIT:
__tp->t_state = TCPS_LAST_ACK;
DCC_LOG1(LOG_TRACE, "<%05x> [LAST_ACK]", (int)__tp);
/* discard the data
* TODO: check whether both buffers must be
* released or not. Probably they where released already.
*/
/* discards unsent data */
__tp->snd_off -= __tp->snd_q.len;
__tp->snd_max -= __tp->snd_q.len;
mbuf_queue_free(&__tp->snd_q);
mbuf_queue_free(&__tp->rcv_q);
/* notify the upper layer that we are closed */
break;
default: {
DCC_LOG2(LOG_ERROR, "<%05x> state=[%s]", (int)__tp,
__tcp_state[__tp->t_state]);
tcpip_net_unlock();
return -1;
}
}
/* ACK now */
__tp->t_flags |= TF_ACKNOW;
/* schedule output */
tcp_output_sched(__tp);
tcpip_net_unlock();
return 0;
}
int tcp_recv(struct tcp_pcb * __tp, void * __buf, int __len)
{
int n;
if (__tp == NULL) {
DCC_LOG(LOG_WARNING, "NULL pointer");
return -1;
}
if (__len == 0) {
/* invalid argument */
DCC_LOG(LOG_WARNING, "invalid argument");
return -1;
}
tcpip_net_lock();
#ifdef ENABLE_SANITY_CHECK
if (pcb_find((struct pcb *)__tp, &__tcp__.active) < 0) {
DCC_LOG1(LOG_ERROR, "<%05x> pcb_find()", (int)__tp);
tcpip_net_unlock();
return -1;
}
#endif
for (;;) {
if ((__tp->t_state == TCPS_CLOSED)) {
DCC_LOG(LOG_WARNING, "closed!");
tcpip_net_unlock();
return -1;
}
if ((__tp->t_state == TCPS_TIME_WAIT) ||
(__tp->t_state == TCPS_CLOSING) ||
(__tp->t_state == TCPS_LAST_ACK)) {
tcpip_net_unlock();
return 0;
}
if (__tp->rcv_q.len)
break;
if (__tp->t_state == TCPS_CLOSE_WAIT) {
tcpip_net_unlock();
return 0;
}
DCC_LOG2(LOG_MSG, "<%05x> wait [%d]", (int)__tp, __tp->t_cond);
thinkos_cond_wait(__tp->t_cond, net_mutex);
}
n = mbuf_queue_remove(&__tp->rcv_q, __buf, __len);
DCC_LOG1(LOG_INFO, "len=%d", n);
/* Half close: don't reopen the receiver window, i'm not sure
whether it is a rule break or not, but it may prevent
resources been consumed by an about to die connection! */
if ((__tp->t_state == TCPS_FIN_WAIT_1) ||
(__tp->t_state == TCPS_FIN_WAIT_2)) {
DCC_LOG1(LOG_TRACE, "<%05x> FIN_WAIT", (int)__tp);
tcpip_net_unlock();
return n;
}
/* XXX: revisit this ... */
// if ((__tp->rcv_q.len == 0) || (__tp->t_flags & TF_DELACK)) {
if ((__tp->rcv_q.len == 0)) {
if (__tp->t_flags & TF_DELACK) {
__tp->t_flags |= TF_ACKNOW;
}
DCC_LOG(LOG_INFO, "empty queue, call tcp_out.");
tcp_output_sched(__tp);
}
tcpip_net_unlock();
return n;
}
int tcp_send(struct tcp_pcb * __tp, const void * __buf,
int __len, int __flags)
{
uint8_t * src;
int rem;
int n;
int m;
if (__tp == NULL) {
DCC_LOG(LOG_WARNING, "NULL pointer");
return -1;
}
#ifdef ENABLE_SANITY
if (__buf == NULL) {
DCC_LOG1(LOG_WARNING, "<%04x> NULL pointer:", (int)__tp);
return -1;
}
if (__len < 0) {
DCC_LOG2(LOG_WARNING, "<%04x> invalid length: %d", (int)__tp, __len);
return -1;
}
#endif
tcpip_net_lock();
#ifdef ENABLE_SANITY
if (pcb_find((struct pcb *)__tp, &__tcp__.active) < 0) {
DCC_LOG(LOG_ERROR, "<%04x> pcb_find()", (int)__tp);
tcpip_net_unlock();
return -1;
}
#endif
DCC_LOG3(LOG_INFO, "<%05x> buf=%05x len=%d", (int)__tp, (int)__buf, __len);
src = (uint8_t *)__buf;
rem = __len;
again:
if (__tp->t_state != TCPS_ESTABLISHED) {
/*
if ((__tp->t_state != TCPS_ESTABLISHED) &&
(__tp->t_state != TCPS_CLOSE_WAIT)) {
*/
DCC_LOG2(LOG_WARNING, "<%05x> [%s]", (int)__tp,
__tcp_state[__tp->t_state]);
if (__tp->t_state == TCPS_SYN_RCVD) {
DCC_LOG1(LOG_TRACE, "<%05x> wait", (int)__tp);
__os_cond_wait(__tp->t_cond, net_mutex);
DCC_LOG2(LOG_TRACE, "<%05x> again [%s]",
(int)__tp, __tcp_state[__tp->t_state]);
goto again;
}
DCC_LOG(LOG_TRACE, "done.");
tcpip_net_unlock();
return -1;
}
while (rem) {
/* buffer limit ... */
m = tcp_maxsnd - __tp->snd_q.len;
if (m <= 0) {
DCC_LOG1(LOG_INFO, "<%05x> queue limit", (int)__tp);
__tp->t_flags |= TF_ACKNOW;
DCC_LOG(LOG_INFO, "output request.");
tcp_output_sched(__tp);
DCC_LOG(LOG_INFO, "waiting for buffer space.");
__os_cond_wait(__tp->t_cond, net_mutex);
goto again;
}
m = MIN(m, rem);
if ((n = mbuf_queue_add(&__tp->snd_q, src, m)) == 0) {
DCC_LOG(LOG_TRACE, "mbuf_wait...");
mbuf_wait(net_mutex);
goto again;
}
rem -= n;
src += n;
}
#if 0
/* FIXME: Set retransmit timer if not currently set,
and not doing an ack or a keepalive probe.
Initial value for retransmit is smoothed
round-trip time + 2 * round-trip time variance.
Initialize counter which is used for backoff
:of retransmit time. */
if ((__tp->t_rxmt_tmr == 0) && (__tp->snd_una != 0)) {
__tp->t_rxmt_tmr = tcp_rxmtintvl[0];
__tp->t_rxmt_cnt = 0;
/* tp->t_flags &= ~TF_IDLE; */
}
#endif
if (__len > 0) {
/* TCP_SEND_NOWAIT flag set or one maximum segment size pending for
send then send now */
if ((__flags & TCP_SEND_NOWAIT) ||
((__tp->snd_q.len - (int)__tp->snd_q.offs) >= __tp->t_maxseg)) {
DCC_LOG(LOG_INFO, "output request.");
tcp_output_sched(__tp);
// if (tcp_output(__tp) < 0) {
/* if the reason to fail was an arp failure
try query an address pending for resolution ... */
// arp_query_pending();
// }
} else {
__tp->t_flags |= TF_DELACK;
}
}
DCC_LOG(LOG_INFO, "done.");
tcpip_net_unlock();
return __len;
}
int udp_sendto(struct udp_pcb * __up, void * __buf, int __len,
const struct sockaddr_in * __sin)
{
struct iphdr * ip;
struct udphdr * uh;
in_addr_t daddr;
int dport;
in_addr_t saddr;
struct route * rt;
#if (ENABLE_NET_UDP_CHECKSUM)
unsigned int sum;
#endif
uint8_t * ptr;
int mtu;
struct ifnet * ifn;
int ret;
int retry = 0;
DCC_LOG2(LOG_INFO, "<%05x> len=%d", (int)__up, __len);
if (__up == NULL) {
DCC_LOG1(LOG_WARNING, "<%05x> invalid pcb", (int)__up);
return -EFAULT;
}
if (__buf == NULL) {
DCC_LOG1(LOG_WARNING, "<%05x> invalid buffer", (int)__up);
return -EFAULT;
}
tcpip_net_lock();
#if (ENABLE_NET_SANITY_CHECK)
if (pcb_find((struct pcb *)__up, &__udp__.pcb) < 0) {
DCC_LOG1(LOG_ERROR, "<%05x> pcb_find()", (int)__up);
tcpip_net_unlock();
/* TODO: errno */
return -1;
}
#endif
if ((__up->u_lport) == 0) {
DCC_LOG1(LOG_WARNING, "<%05x> not bound", (int)__up);
tcpip_net_unlock();
/* TODO: errno */
return -3;
}
if (__sin == NULL) {
if ((dport = __up->u_fport) == 0) {
DCC_LOG1(LOG_WARNING, "<%05x> connection refused", (int)__up);
tcpip_net_unlock();
return -ECONNREFUSED;
}
if ((daddr = __up->u_faddr) == INADDR_ANY) {
DCC_LOG1(LOG_WARNING, "<%05x> not connected", (int)__up);
tcpip_net_unlock();
return -ENOTCONN;
}
} else {
if ((dport = __sin->sin_port) == 0) {
DCC_LOG1(LOG_WARNING, "<%05x> invalid port", (int)__up);
tcpip_net_unlock();
return -ECONNREFUSED;
}
if ((daddr = __sin->sin_addr.s_addr) == INADDR_ANY) {
DCC_LOG1(LOG_WARNING, "<%05x> invalid address", (int)__up);
tcpip_net_unlock();
return -EDESTADDRREQ;
}
}
if ((rt = __route_lookup(daddr)) == NULL) {
DCC_LOG2(LOG_WARNING, "<%05x> no route to host: %I", (int)__up, daddr);
tcpip_net_unlock();
UDP_PROTO_STAT_ADD(tx_drop, 1);
UDP_PROTO_STAT_ADD(tx_err, 1);
return -EHOSTUNREACH;
}
ifn = (struct ifnet *)rt->rt_ifn;
mtu = ifn->if_mtu - sizeof(struct iphdr);
if ((__len <= 0) || (__len > mtu)) {
DCC_LOG3(LOG_WARNING, "<%04x> invalid length %d (max: %d)",
(int)__up, __len, mtu);
tcpip_net_unlock();
/* TODO: errno */
return -7;
}
/* get the source address */
if ((saddr = __up->u_laddr) == INADDR_ANY) {
saddr = ifn->if_ipv4_addr;
}
again:
ip = (struct iphdr *)ifn_mmap(ifn, sizeof(struct iphdr) +
sizeof(struct udphdr) + __len);
if (ip == NULL) {
DCC_LOG1(LOG_WARNING, "<%04x> ifn_mmap() fail", (int)__up);
tcpip_net_unlock();
/* TODO: errno */
return -1;
}
DCC_LOG2(LOG_TRACE, "<%05x> ip=%p", (int)__up, ip);
iph_template(ip, IPPROTO_UDP, udp_def_ttl, udp_def_tos);
uh = (struct udphdr *)ip->opt;
/* build the ip header */
ip = mk_iphdr(ip, saddr, daddr, sizeof(struct udphdr) + __len);
/* fill the udp header fields */
uh = (struct udphdr *)ip->opt;
uh->dport = dport;
uh->sport = __up->u_lport;
uh->len = htons(__len + sizeof(struct udphdr));
#if (ENABLE_NET_UDP_CHECKSUM)
/* initialize the udp checksum */
sum = uh->len << 1;
sum += uh->dport;
sum += uh->sport;
sum += (IPPROTO_UDP << 8);
sum += ((uint16_t *)(void *)&(ip->saddr))[0] +
((uint16_t *)(void *)&(ip->saddr))[1];
sum += ((uint16_t *)(void *)&(ip->daddr))[0] +
((uint16_t *)(void *)&(ip->daddr))[1];
#endif
ptr = (uint8_t *)uh + sizeof(struct udphdr);
memcpy(ptr, __buf, __len);
#if (ENABLE_NET_UDP_CHECKSUM)
if (__len) {
sum = in_chksum(sum, ptr, __len);
}
uh->chksum = ~sum;
#else
uh->chksum = 0;
#endif
#if 0
DCC_LOG(LOG_INFO, "IP %d.%d.%d.%d:%d > %d.%d.%d.%d:%d: %d",
IP4_ADDR1(ip->saddr), IP4_ADDR2(ip->saddr), IP4_ADDR3(ip->saddr),
IP4_ADDR4(ip->saddr), ntohs(uh->sport), IP4_ADDR1(ip->daddr),
IP4_ADDR2(ip->daddr), IP4_ADDR3(ip->daddr), IP4_ADDR4(ip->daddr),
ntohs(uh->dport), ntohs(uh->len));
#endif
if ((ret = ip_output(ifn, rt, ip)) < 0) {
ifn_munmap(ifn, ip);
/* if the reason to fail was an arp failure
try query an address pending for resolution ... */
if ((ret == -EAGAIN) && (retry < 10)) {
etharp_query_pending();
tcpip_net_unlock();
DCC_LOG2(LOG_WARNING, "<%05x> again, retry=%d!", (int)__up, retry);
thinkos_sleep(10 + retry * 10);
retry++;
tcpip_net_lock();
goto again;
}
DCC_LOG1(LOG_ERROR, "<%05x> ip_output() fail!", (int)__up);
UDP_PROTO_STAT_ADD(tx_drop, 1);
} else {
UDP_PROTO_STAT_ADD(tx_ok, 1);
DCC_LOG5(LOG_INFO, "IP %I:%d > %I:%d: %d",
ip->saddr, ntohs(uh->sport), ip->daddr,
ntohs(uh->dport), ntohs(uh->len));
#if (LOG_LEVEL < LOG_INFO)
DCC_LOG(LOG_INFO, "sent.");
#endif
ret = __len;
}
tcpip_net_unlock();
return ret;
}