osd-contiki/core/net/tcpip.c

779 lines
21 KiB
C
Raw Normal View History

/*
* Copyright (c) 2004, Swedish Institute of Computer Science.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the Institute nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE INSTITUTE AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE INSTITUTE OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* This file is part of the Contiki operating system.
*
*
* $Id: tcpip.c,v 1.24 2010/02/09 12:58:53 adamdunkels Exp $
2008-10-14 11:40:56 +02:00
*/
/**
* \file
* Code for tunnelling uIP packets over the Rime mesh routing module
*
2008-10-14 11:40:56 +02:00
* \author Adam Dunkels <adam@sics.se>\author
* \author Mathilde Durvy <mdurvy@cisco.com> (IPv6 related code)
* \author Julien Abeille <jabeille@cisco.com> (IPv6 related code)
*/
#include "contiki-net.h"
#include "net/uip-split.h"
#include <string.h>
2008-10-14 11:40:56 +02:00
#if UIP_CONF_IPV6
#include "net/uip-nd6.h"
#include "net/uip-netif.h"
#endif
#define DEBUG 0
#if DEBUG
#include <stdio.h>
#define PRINTF(...) printf(__VA_ARGS__)
#define PRINT6ADDR(addr) PRINTF(" %02x%02x:%02x%02x:%02x%02x:%02x%02x:%02x%02x:%02x%02x:%02x%02x:%02x%02x ", ((u8_t *)addr)[0], ((u8_t *)addr)[1], ((u8_t *)addr)[2], ((u8_t *)addr)[3], ((u8_t *)addr)[4], ((u8_t *)addr)[5], ((u8_t *)addr)[6], ((u8_t *)addr)[7], ((u8_t *)addr)[8], ((u8_t *)addr)[9], ((u8_t *)addr)[10], ((u8_t *)addr)[11], ((u8_t *)addr)[12], ((u8_t *)addr)[13], ((u8_t *)addr)[14], ((u8_t *)addr)[15])
#define PRINTLLADDR(lladdr) PRINTF(" %02x:%02x:%02x:%02x:%02x:%02x ",lladdr->addr[0], lladdr->addr[1], lladdr->addr[2], lladdr->addr[3],lladdr->addr[4], lladdr->addr[5])
#else
#define PRINTF(...)
#define PRINT6ADDR(addr)
#endif
#if UIP_LOGGING
#include <stdio.h>
void uip_log(char *msg);
#define UIP_LOG(m) uip_log(m)
#else
#define UIP_LOG(m)
#endif
#define UIP_ICMP_BUF ((struct uip_icmp_hdr *)&uip_buf[UIP_LLIPH_LEN + uip_ext_len])
#define UIP_IP_BUF ((struct uip_ip_hdr *)&uip_buf[UIP_LLH_LEN])
#define UIP_TCP_BUF ((struct uip_tcpip_hdr *)&uip_buf[UIP_LLH_LEN])
2008-10-14 11:40:56 +02:00
process_event_t tcpip_event;
#if UIP_CONF_ICMP6
process_event_t tcpip_icmp6_event;
#endif /* UIP_CONF_ICMP6 */
2008-10-14 11:40:56 +02:00
/*static struct tcpip_event_args ev_args;*/
/*periodic check of active connections*/
static struct etimer periodic;
#if UIP_CONF_IPV6 && UIP_CONF_IPV6_REASSEMBLY
/*timer for reassembly*/
extern struct etimer uip_reass_timer;
#endif
#if UIP_TCP
/**
* \internal Structure for holding a TCP port and a process ID.
*/
struct listenport {
u16_t port;
struct process *p;
};
static struct internal_state {
struct listenport listenports[UIP_LISTENPORTS];
struct process *p;
} s;
2008-10-14 11:40:56 +02:00
#endif
enum {
TCP_POLL,
UDP_POLL,
PACKET_INPUT
};
2008-10-14 11:40:56 +02:00
/* Called on IP packet output. */
#if UIP_CONF_IPV6
static u8_t (* outputfunc)(uip_lladdr_t *a);
u8_t
tcpip_output(uip_lladdr_t *a)
{
int ret;
if(outputfunc != NULL) {
ret = outputfunc(a);
return ret;
}
UIP_LOG("tcpip_output: Use tcpip_set_outputfunc() to set an output function");
return 0;
}
void
tcpip_set_outputfunc(u8_t (*f)(uip_lladdr_t *))
{
outputfunc = f;
}
2008-10-14 11:40:56 +02:00
#else
static u8_t (* outputfunc)(void);
u8_t
tcpip_output(void)
{
if(outputfunc != NULL) {
return outputfunc();
}
UIP_LOG("tcpip_output: Use tcpip_set_outputfunc() to set an output function");
return 0;
}
void
tcpip_set_outputfunc(u8_t (*f)(void))
{
outputfunc = f;
}
2008-10-14 11:40:56 +02:00
#endif
#if UIP_CONF_IP_FORWARD
unsigned char tcpip_is_forwarding; /* Forwarding right now? */
#endif /* UIP_CONF_IP_FORWARD */
PROCESS(tcpip_process, "TCP/IP stack");
/*---------------------------------------------------------------------------*/
static void
start_periodic_tcp_timer(void)
{
if(etimer_expired(&periodic)) {
etimer_restart(&periodic);
}
}
/*---------------------------------------------------------------------------*/
static void
check_for_tcp_syn(void)
{
/* This is a hack that is needed to start the periodic TCP timer if
an incoming packet contains a SYN: since uIP does not inform the
application if a SYN arrives, we have no other way of starting
this timer. This function is called for every incoming IP packet
to check for such SYNs. */
#define TCP_SYN 0x02
if(UIP_IP_BUF->proto == UIP_PROTO_TCP &&
(UIP_TCP_BUF->flags & TCP_SYN) == TCP_SYN) {
start_periodic_tcp_timer();
}
}
/*---------------------------------------------------------------------------*/
static void
packet_input(void)
{
#if UIP_CONF_IP_FORWARD
if(uip_len > 0) {
tcpip_is_forwarding = 1;
if(uip_fw_forward() == UIP_FW_LOCAL) {
tcpip_is_forwarding = 0;
check_for_tcp_syn();
uip_input();
if(uip_len > 0) {
#if UIP_CONF_TCP_SPLIT
2008-10-14 11:40:56 +02:00
uip_split_output();
#else /* UIP_CONF_TCP_SPLIT */
2008-10-14 11:40:56 +02:00
#if UIP_CONF_IPV6
tcpip_ipv6_output();
#else
PRINTF("tcpip packet_input forward output len %d\n", uip_len);
2008-10-14 11:40:56 +02:00
tcpip_output();
#endif
#endif /* UIP_CONF_TCP_SPLIT */
}
}
tcpip_is_forwarding = 0;
}
#else /* UIP_CONF_IP_FORWARD */
if(uip_len > 0) {
check_for_tcp_syn();
uip_input();
if(uip_len > 0) {
#if UIP_CONF_TCP_SPLIT
uip_split_output();
#else /* UIP_CONF_TCP_SPLIT */
2008-10-14 11:40:56 +02:00
#if UIP_CONF_IPV6
tcpip_ipv6_output();
#else
PRINTF("tcpip packet_input output len %d\n", uip_len);
tcpip_output();
2008-10-14 11:40:56 +02:00
#endif
#endif /* UIP_CONF_TCP_SPLIT */
}
}
#endif /* UIP_CONF_IP_FORWARD */
}
/*---------------------------------------------------------------------------*/
2008-10-14 11:40:56 +02:00
#if UIP_TCP
#if UIP_ACTIVE_OPEN
struct uip_conn *
2006-08-10 18:43:32 +02:00
tcp_connect(uip_ipaddr_t *ripaddr, u16_t port, void *appstate)
{
struct uip_conn *c;
2006-08-10 18:43:32 +02:00
c = uip_connect(ripaddr, port);
if(c == NULL) {
return NULL;
}
c->appstate.p = PROCESS_CURRENT();
c->appstate.state = appstate;
tcpip_poll_tcp(c);
return c;
}
#endif /* UIP_ACTIVE_OPEN */
/*---------------------------------------------------------------------------*/
void
tcp_unlisten(u16_t port)
{
static unsigned char i;
struct listenport *l;
l = s.listenports;
for(i = 0; i < UIP_LISTENPORTS; ++i) {
if(l->port == port &&
l->p == PROCESS_CURRENT()) {
l->port = 0;
uip_unlisten(port);
break;
}
++l;
}
}
/*---------------------------------------------------------------------------*/
void
tcp_listen(u16_t port)
{
static unsigned char i;
struct listenport *l;
l = s.listenports;
for(i = 0; i < UIP_LISTENPORTS; ++i) {
if(l->port == 0) {
l->port = port;
l->p = PROCESS_CURRENT();
uip_listen(port);
break;
}
++l;
}
}
/*---------------------------------------------------------------------------*/
void
tcp_attach(struct uip_conn *conn,
void *appstate)
{
register uip_tcp_appstate_t *s;
s = &conn->appstate;
s->p = PROCESS_CURRENT();
s->state = appstate;
}
2008-10-14 11:40:56 +02:00
#endif /* UIP_TCP */
/*---------------------------------------------------------------------------*/
#if UIP_UDP
void
udp_attach(struct uip_udp_conn *conn,
void *appstate)
{
register uip_udp_appstate_t *s;
s = &conn->appstate;
s->p = PROCESS_CURRENT();
s->state = appstate;
}
/*---------------------------------------------------------------------------*/
struct uip_udp_conn *
2007-03-15 20:41:29 +01:00
udp_new(const uip_ipaddr_t *ripaddr, u16_t port, void *appstate)
{
struct uip_udp_conn *c;
uip_udp_appstate_t *s;
c = uip_udp_new(ripaddr, port);
if(c == NULL) {
return NULL;
}
s = &c->appstate;
s->p = PROCESS_CURRENT();
s->state = appstate;
return c;
}
/*---------------------------------------------------------------------------*/
struct uip_udp_conn *
udp_broadcast_new(u16_t port, void *appstate)
{
uip_ipaddr_t addr;
struct uip_udp_conn *conn;
#if UIP_CONF_IPV6
uip_create_linklocal_allnodes_mcast(&addr);
#else
uip_ipaddr(&addr, 255,255,255,255);
#endif /* UIP_CONF_IPV6 */
conn = udp_new(&addr, port, appstate);
if(conn != NULL) {
udp_bind(conn, port);
}
return conn;
}
#endif /* UIP_UDP */
2008-10-14 11:40:56 +02:00
/*---------------------------------------------------------------------------*/
#if UIP_CONF_ICMP6
u8_t
2008-10-14 11:40:56 +02:00
icmp6_new(void *appstate) {
if(uip_icmp6_conns.appstate.p == PROCESS_NONE) {
uip_icmp6_conns.appstate.p = PROCESS_CURRENT();
uip_icmp6_conns.appstate.state = appstate;
return 0;
}
return 1;
}
void
tcpip_icmp6_call(u8_t type)
{
if(uip_icmp6_conns.appstate.p != PROCESS_NONE) {
/* XXX: This is a hack that needs to be updated. Passing a pointer (&type)
like this only works with process_post_synch. */
process_post_synch(uip_icmp6_conns.appstate.p, tcpip_icmp6_event, &type);
}
2008-10-14 11:40:56 +02:00
return;
}
#endif /* UIP_CONF_ICMP6 */
/*---------------------------------------------------------------------------*/
static void
eventhandler(process_event_t ev, process_data_t data)
{
#if UIP_TCP
static unsigned char i;
register struct listenport *l;
#endif /*UIP_TCP*/
struct process *p;
2008-10-14 11:40:56 +02:00
switch(ev) {
2008-10-14 11:40:56 +02:00
case PROCESS_EVENT_EXITED:
/* This is the event we get if a process has exited. We go through
the TCP/IP tables to see if this process had any open
connections or listening TCP ports. If so, we'll close those
connections. */
p = (struct process *)data;
#if UIP_TCP
2008-10-14 11:40:56 +02:00
l = s.listenports;
for(i = 0; i < UIP_LISTENPORTS; ++i) {
if(l->p == p) {
uip_unlisten(l->port);
l->port = 0;
l->p = PROCESS_NONE;
}
++l;
}
2008-10-14 11:40:56 +02:00
{
register struct uip_conn *cptr;
for(cptr = &uip_conns[0]; cptr < &uip_conns[UIP_CONNS]; ++cptr) {
if(cptr->appstate.p == p) {
cptr->appstate.p = PROCESS_NONE;
cptr->tcpstateflags = UIP_CLOSED;
}
}
}
#endif /* UIP_TCP */
#if UIP_UDP
2008-10-14 11:40:56 +02:00
{
register struct uip_udp_conn *cptr;
for(cptr = &uip_udp_conns[0];
cptr < &uip_udp_conns[UIP_UDP_CONNS]; ++cptr) {
if(cptr->appstate.p == p) {
cptr->lport = 0;
}
}
2008-10-14 11:40:56 +02:00
}
#endif /* UIP_UDP */
2008-10-14 11:40:56 +02:00
break;
case PROCESS_EVENT_TIMER:
/* We get this event if one of our timers have expired. */
{
/* Check the clock so see if we should call the periodic uIP
processing. */
2008-10-14 11:40:56 +02:00
if(data == &periodic &&
etimer_expired(&periodic)) {
#if UIP_TCP
for(i = 0; i < UIP_CONNS; ++i) {
if(uip_conn_active(i)) {
/* Only restart the timer if there are active
connections. */
etimer_restart(&periodic);
uip_periodic(i);
#if UIP_CONF_IPV6
tcpip_ipv6_output();
#else
if(uip_len > 0) {
PRINTF("tcpip_output from periodic len %d\n", uip_len);
2008-10-14 11:40:56 +02:00
tcpip_output();
PRINTF("tcpip_output after periodic len %d\n", uip_len);
2008-10-14 11:40:56 +02:00
}
#endif /* UIP_CONF_IPV6 */
2008-10-14 11:40:56 +02:00
}
}
#endif /* UIP_TCP */
#if UIP_CONF_IP_FORWARD
2008-10-14 11:40:56 +02:00
uip_fw_periodic();
#endif /* UIP_CONF_IP_FORWARD */
2008-10-14 11:40:56 +02:00
}
#if UIP_CONF_IPV6
#if UIP_CONF_IPV6_REASSEMBLY
/*
* check the timer for reassembly
*/
if(data == &uip_reass_timer &&
etimer_expired(&uip_reass_timer)) {
uip_reass_over();
tcpip_ipv6_output();
}
#endif /* UIP_CONF_IPV6_REASSEMBLY */
/*
* check the different timers for neighbor discovery and
* stateless autoconfiguration
*/
if(data == &uip_nd6_timer_periodic &&
etimer_expired(&uip_nd6_timer_periodic)) {
uip_nd6_periodic();
tcpip_ipv6_output();
}
if(data == &uip_netif_timer_dad &&
etimer_expired(&uip_netif_timer_dad)){
uip_netif_dad();
tcpip_ipv6_output();
}
if(data == &uip_netif_timer_rs &&
etimer_expired(&uip_netif_timer_rs)){
uip_netif_send_rs();
tcpip_ipv6_output();
}
2008-10-14 11:40:56 +02:00
if(data == &uip_netif_timer_periodic &&
etimer_expired(&uip_netif_timer_periodic)){
uip_netif_periodic();
}
#endif /* UIP_CONF_IPV6 */
}
2008-10-14 11:40:56 +02:00
break;
#if UIP_TCP
case TCP_POLL:
if(data != NULL) {
uip_poll_conn(data);
#if UIP_CONF_IPV6
tcpip_ipv6_output();
#else /* UIP_CONF_IPV6 */
2008-10-14 11:40:56 +02:00
if(uip_len > 0) {
PRINTF("tcpip_output from tcp poll len %d\n", uip_len);
2008-10-14 11:40:56 +02:00
tcpip_output();
}
#endif /* UIP_CONF_IPV6 */
/* Start the periodic polling, if it isn't already active. */
start_periodic_tcp_timer();
2008-10-14 11:40:56 +02:00
}
break;
#endif /* UIP_TCP */
#if UIP_UDP
2008-10-14 11:40:56 +02:00
case UDP_POLL:
if(data != NULL) {
uip_udp_periodic_conn(data);
#if UIP_CONF_IPV6
tcpip_ipv6_output();
#else
if(uip_len > 0) {
tcpip_output();
}
#endif /* UIP_UDP */
}
2008-10-14 11:40:56 +02:00
break;
#endif /* UIP_UDP */
2008-10-14 11:40:56 +02:00
case PACKET_INPUT:
packet_input();
break;
};
}
/*---------------------------------------------------------------------------*/
void
tcpip_input(void)
{
process_post_synch(&tcpip_process, PACKET_INPUT, NULL);
uip_len = 0;
2008-10-14 11:40:56 +02:00
#if UIP_CONF_IPV6
uip_ext_len = 0;
#endif /*UIP_CONF_IPV6*/
}
/*---------------------------------------------------------------------------*/
2008-10-14 11:40:56 +02:00
#if UIP_CONF_IPV6
void
tcpip_ipv6_output(void)
{
struct uip_nd6_neighbor *nbc = NULL;
struct uip_nd6_defrouter *dr = NULL;
if(uip_len == 0)
return;
if(uip_len > UIP_LINK_MTU){
UIP_LOG("tcpip_ipv6_output: Packet to big");
uip_len = 0;
return;
}
if(uip_is_addr_unspecified(&UIP_IP_BUF->destipaddr)){
UIP_LOG("tcpip_ipv6_output: Destination address unspecified");
uip_len = 0;
return;
2008-10-14 11:40:56 +02:00
}
if(!uip_is_addr_mcast(&UIP_IP_BUF->destipaddr)) {
/*If destination is on link */
nbc = NULL;
2008-10-14 11:40:56 +02:00
if(uip_nd6_is_addr_onlink(&UIP_IP_BUF->destipaddr)){
nbc = uip_nd6_nbrcache_lookup(&UIP_IP_BUF->destipaddr);
} else {
#if UIP_CONF_ROUTER
2008-10-14 11:40:56 +02:00
/*destination is not on link*/
uip_ipaddr_t ipaddr;
uip_ipaddr_t *next_hop;
/* Try to find the next hop address in the local routing table. */
next_hop = uip_router != NULL ?
uip_router->lookup(&UIP_IP_BUF->destipaddr, &ipaddr) : NULL;
if(next_hop != NULL) {
/* Look for the next hop of the route in the neighbor cache.
Add a cache entry if we can't find it. */
nbc = uip_nd6_nbrcache_lookup(next_hop);
if(nbc == NULL) {
nbc = uip_nd6_nbrcache_add(next_hop, NULL, 1, NO_STATE);
}
2008-10-14 11:40:56 +02:00
} else {
#endif /* UIP_CONF_ROUTER */
/* No route found, check if a default router exists and use it then. */
dr = uip_nd6_choose_defrouter();
if(dr != NULL){
nbc = dr->nb;
} else {
/* shall we send a icmp error message destination unreachable ?*/
UIP_LOG("tcpip_ipv6_output: Destination off-link but no router");
uip_len = 0;
return;
}
#if UIP_CONF_ROUTER
}
#endif /* UIP_CONF_ROUTER */
2008-10-14 11:40:56 +02:00
}
/* there are two cases where the entry logically does not exist:
* 1 it really does not exist. 2 it is in the NO_STATE state */
if (nbc == NULL || nbc->state == NO_STATE) {
if (nbc == NULL) {
/* create neighbor cache entry, original packet is replaced by NS*/
nbc = uip_nd6_nbrcache_add(&UIP_IP_BUF->destipaddr, NULL, 0, INCOMPLETE);
} else {
nbc->state = INCOMPLETE;
}
#if UIP_CONF_IPV6_QUEUE_PKT
2008-10-14 11:40:56 +02:00
/* copy outgoing pkt in the queuing buffer for later transmmit */
memcpy(nbc->queue_buf, UIP_IP_BUF, uip_len);
nbc->queue_buf_len = uip_len;
#endif
/* RFC4861, 7.2.2:
* "If the source address of the packet prompting the solicitation is the
* same as one of the addresses assigned to the outgoing interface, that
* address SHOULD be placed in the IP Source Address of the outgoing
* solicitation. Otherwise, any one of the addresses assigned to the
* interface should be used."*/
if(uip_netif_is_addr_my_unicast(&UIP_IP_BUF->srcipaddr)){
uip_nd6_io_ns_output(&UIP_IP_BUF->srcipaddr, NULL, &nbc->ipaddr);
} else {
uip_nd6_io_ns_output(NULL, NULL, &nbc->ipaddr);
}
stimer_set(&(nbc->last_send),
uip_netif_physical_if.retrans_timer / 1000);
2008-10-14 11:40:56 +02:00
nbc->count_send = 1;
} else {
if (nbc->state == INCOMPLETE){
PRINTF("tcpip_ipv6_output: neighbor cache entry incomplete\n");
#if UIP_CONF_IPV6_QUEUE_PKT
2008-10-14 11:40:56 +02:00
/* copy outgoing pkt in the queuing buffer for later transmmit and set
the destination neighbor to nbc */
memcpy(nbc->queue_buf, UIP_IP_BUF, uip_len);
nbc->queue_buf_len = uip_len;
uip_len = 0;
#endif /*UIP_CONF_IPV6_QUEUE_PKT*/
return;
}
/* if running NUD (nbc->state == STALE, DELAY, or PROBE ) keep
sending in parallel see rfc 4861 Node behavior in section 7.7.3*/
if (nbc->state == STALE){
nbc->state = DELAY;
stimer_set(&(nbc->reachable),
UIP_ND6_DELAY_FIRST_PROBE_TIME);
2008-10-14 11:40:56 +02:00
PRINTF("tcpip_ipv6_output: neighbor cache entry stale moving to delay\n");
}
stimer_set(&(nbc->last_send),
uip_netif_physical_if.retrans_timer / 1000);
2008-10-14 11:40:56 +02:00
tcpip_output(&(nbc->lladdr));
#if UIP_CONF_IPV6_QUEUE_PKT
/* Send the queued packets from here, may not be 100% perfect though.
* This happens in a few cases, for example when instead of receiving a
* NA after sendiong a NS, you receive a NS with SLLAO: the entry moves
*to STALE, and you must both send a NA and the queued packet
*/
if(nbc->queue_buf_len != 0) {
uip_len = nbc->queue_buf_len;
2008-10-14 11:40:56 +02:00
memcpy(UIP_IP_BUF, nbc->queue_buf, uip_len);
nbc->queue_buf_len = 0;
2008-10-14 11:40:56 +02:00
tcpip_output(&(nbc->lladdr));
}
#endif /*UIP_CONF_IPV6_QUEUE_PKT*/
uip_len = 0;
return;
}
}
/*multicast IP destination address */
2008-10-14 11:40:56 +02:00
tcpip_output(NULL);
uip_len = 0;
uip_ext_len = 0;
}
#endif
/*---------------------------------------------------------------------------*/
#if UIP_UDP
void
tcpip_poll_udp(struct uip_udp_conn *conn)
{
process_post(&tcpip_process, UDP_POLL, conn);
}
#endif /* UIP_UDP */
/*---------------------------------------------------------------------------*/
2008-10-14 11:40:56 +02:00
#if UIP_TCP
void
tcpip_poll_tcp(struct uip_conn *conn)
{
process_post(&tcpip_process, TCP_POLL, conn);
}
2008-10-14 11:40:56 +02:00
#endif /* UIP_TCP */
/*---------------------------------------------------------------------------*/
void
tcpip_uipcall(void)
{
register uip_udp_appstate_t *ts;
2008-10-14 11:40:56 +02:00
#if UIP_UDP
if(uip_conn != NULL) {
ts = &uip_conn->appstate;
} else {
ts = &uip_udp_conn->appstate;
}
#else /* UIP_UDP */
ts = &uip_conn->appstate;
#endif /* UIP_UDP */
#if UIP_TCP
{
static unsigned char i;
register struct listenport *l;
/* If this is a connection request for a listening port, we must
mark the connection with the right process ID. */
if(uip_connected()) {
l = &s.listenports[0];
for(i = 0; i < UIP_LISTENPORTS; ++i) {
if(l->port == uip_conn->lport &&
l->p != PROCESS_NONE) {
ts->p = l->p;
ts->state = NULL;
break;
}
++l;
}
/* Start the periodic polling, if it isn't already active. */
start_periodic_tcp_timer();
}
}
#endif /* UIP_TCP */
2008-10-14 11:40:56 +02:00
if(ts->p != NULL) {
process_post_synch(ts->p, tcpip_event, ts->state);
}
}
/*---------------------------------------------------------------------------*/
PROCESS_THREAD(tcpip_process, ev, data)
{
PROCESS_BEGIN();
2008-10-14 11:40:56 +02:00
#if UIP_TCP
{
static unsigned char i;
for(i = 0; i < UIP_LISTENPORTS; ++i) {
s.listenports[i].port = 0;
}
s.p = PROCESS_CURRENT();
}
2008-10-14 11:40:56 +02:00
#endif
tcpip_event = process_alloc_event();
#if UIP_CONF_ICMP6
tcpip_icmp6_event = process_alloc_event();
#endif /* UIP_CONF_ICMP6 */
etimer_set(&periodic, CLOCK_SECOND / 2);
uip_init();
while(1) {
PROCESS_YIELD();
eventhandler(ev, data);
}
PROCESS_END();
}
/*---------------------------------------------------------------------------*/