osd-contiki/core/net/rpl/rpl-timers.c

237 lines
8 KiB
C

/**
* \addtogroup uip6
* @{
*/
/*
* Copyright (c) 2010, 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.
*/
/**
* \file
* RPL timer management.
*
* \author Joakim Eriksson <joakime@sics.se>, Nicolas Tsiftes <nvt@sics.se>
*/
#include "contiki-conf.h"
#include "net/rpl/rpl-private.h"
#include "lib/random.h"
#include "sys/ctimer.h"
#define DEBUG DEBUG_NONE
#include "net/uip-debug.h"
/************************************************************************/
static struct ctimer periodic_timer;
static void handle_periodic_timer(void *ptr);
static void new_dio_interval(rpl_instance_t *instance);
static void handle_dio_timer(void *ptr);
static uint16_t next_dis;
/* dio_send_ok is true if the node is ready to send DIOs */
static uint8_t dio_send_ok;
/************************************************************************/
static void
handle_periodic_timer(void *ptr)
{
rpl_purge_routes();
rpl_recalculate_ranks();
/* handle DIS */
#ifdef RPL_DIS_SEND
next_dis++;
if(rpl_get_any_dag() == NULL && next_dis >= RPL_DIS_INTERVAL) {
next_dis = 0;
dis_output(NULL);
}
#endif
ctimer_reset(&periodic_timer);
}
/************************************************************************/
static void
new_dio_interval(rpl_instance_t *instance)
{
uint32_t time;
/* TODO: too small timer intervals for many cases */
time = 1UL << instance->dio_intcurrent;
/* Convert from milliseconds to CLOCK_TICKS. */
time = (time * CLOCK_SECOND) / 1000;
instance->dio_next_delay = time;
/* random number between I/2 and I */
time = time >> 1;
time += (time * random_rand()) / RANDOM_RAND_MAX;
/*
* The intervals must be equally long among the nodes for Trickle to
* operate efficiently. Therefore we need to calculate the delay between
* the randomized time and the start time of the next interval.
*/
instance->dio_next_delay -= time;
instance->dio_send = 1;
#if RPL_CONF_STATS
/* keep some stats */
instance->dio_totint++;
instance->dio_totrecv += instance->dio_counter;
ANNOTATE("#A rank=%u.%u(%u),stats=%d %d %d %d,color=%s\n",
DAG_RANK(instance->current_dag->rank, instance),
(10 * (instance->current_dag->rank % instance->min_hoprankinc)) / instance->min_hoprankinc,
instance->current_dag->version,
instance->dio_totint, instance->dio_totsend,
instance->dio_totrecv,instance->dio_intcurrent,
instance->current_dag->rank == ROOT_RANK(instance) ? "BLUE" : "ORANGE");
#endif /* RPL_CONF_STATS */
/* reset the redundancy counter */
instance->dio_counter = 0;
/* schedule the timer */
PRINTF("RPL: Scheduling DIO timer %lu ticks in future (Interval)\n", time);
ctimer_set(&instance->dio_timer, time, &handle_dio_timer, instance);
}
/************************************************************************/
static void
handle_dio_timer(void *ptr)
{
rpl_instance_t *instance;
instance = (rpl_instance_t *)ptr;
PRINTF("RPL: DIO Timer triggered\n");
if(!dio_send_ok) {
if(uip_ds6_get_link_local(ADDR_PREFERRED) != NULL) {
dio_send_ok = 1;
} else {
PRINTF("RPL: Postponing DIO transmission since link local address is not ok\n");
ctimer_set(&instance->dio_timer, CLOCK_SECOND, &handle_dio_timer, instance);
return;
}
}
if(instance->dio_send) {
/* send DIO if counter is less than desired redundancy */
if(instance->dio_counter < instance->dio_redundancy) {
#if RPL_CONF_STATS
instance->dio_totsend++;
#endif /* RPL_CONF_STATS */
dio_output(instance, NULL);
} else {
PRINTF("RPL: Supressing DIO transmission (%d >= %d)\n",
instance->dio_counter, instance->dio_redundancy);
}
instance->dio_send = 0;
PRINTF("RPL: Scheduling DIO timer %"PRIu32" ticks in future (sent)\n",
instance->dio_next_delay);
ctimer_set(&instance->dio_timer, instance->dio_next_delay, handle_dio_timer, instance);
} else {
/* check if we need to double interval */
if(instance->dio_intcurrent < instance->dio_intmin + instance->dio_intdoubl) {
instance->dio_intcurrent++;
PRINTF("RPL: DIO Timer interval doubled %d\n", instance->dio_intcurrent);
}
new_dio_interval(instance);
}
}
/************************************************************************/
void
rpl_reset_periodic_timer(void)
{
next_dis = RPL_DIS_INTERVAL - RPL_DIS_START_DELAY;
ctimer_set(&periodic_timer, CLOCK_SECOND, handle_periodic_timer, NULL);
}
/************************************************************************/
/* Resets the DIO timer in the instance to its minimal interval. */
void
rpl_reset_dio_timer(rpl_instance_t *instance, uint8_t force)
{
#if !RPL_LEAF_ONLY
/* Do not reset if we are already on the minimum interval,
unless forced to do so. */
if(force || instance->dio_intcurrent > instance->dio_intmin) {
instance->dio_counter = 0;
instance->dio_intcurrent = instance->dio_intmin;
new_dio_interval(instance);
}
#if RPL_CONF_STATS
rpl_stats.resets++;
#endif /* RPL_CONF_STATS */
#endif /* RPL_LEAF_ONLY */
}
/************************************************************************/
static void
handle_dao_timer(void *ptr)
{
rpl_instance_t *instance;
instance = (rpl_instance_t *)ptr;
if(!dio_send_ok && uip_ds6_get_link_local(ADDR_PREFERRED) == NULL) {
PRINTF("RPL: Postpone DAO transmission\n");
ctimer_set(&instance->dao_timer, CLOCK_SECOND, handle_dao_timer, instance);
return;
}
/* Send the DAO to the DAO parent set -- the preferred parent in our case. */
if(instance->current_dag->preferred_parent != NULL) {
PRINTF("RPL: handle_dao_timer - sending DAO\n");
/* Set the route lifetime to the default value. */
dao_output(instance->current_dag->preferred_parent, instance->default_lifetime);
} else {
PRINTF("RPL: No suitable DAO parent\n");
}
ctimer_stop(&instance->dao_timer);
}
/************************************************************************/
void
rpl_schedule_dao(rpl_instance_t *instance)
{
clock_time_t expiration_time;
expiration_time = etimer_expiration_time(&instance->dao_timer.etimer);
if(!etimer_expired(&instance->dao_timer.etimer)) {
PRINTF("RPL: DAO timer already scheduled\n");
} else {
expiration_time = DEFAULT_DAO_LATENCY / 2 +
(random_rand() % (DEFAULT_DAO_LATENCY));
PRINTF("RPL: Scheduling DAO timer %u ticks in the future\n",
(unsigned)expiration_time);
ctimer_set(&instance->dao_timer, expiration_time,
handle_dao_timer, instance);
}
}
/************************************************************************/