a strange behaviour when running a program that accesses process image, only when launched automatically by init.d
Symptoms
Automatically launching a program at boot (before login), it crashes silently but the process still seems to be loaded in memory.
Cause
It seems that the RevPi process image interface (reading/writing process image) doesn’t work when the program is launched automatically before login (inti.d method).
If the program is launched manually from console all works fine.
In fact, remarking calls to process image all works fine.
Note
- Access to process image has been done as in the sample program that can be found on RevPI GitHub site (derived from piControlIf.c ).
The process image is accessed from a thread.
Script code (unable to attach file)
#! /bin/sh
### BEGIN INIT INFO
# Provides: WebApp
# Required-Start:
# Required-Stop:
# Should-Start:
# X-Start-Before:
# Default-Start: 2
# Default-Stop:
# X-Interactive: true
# Short-Description: Starts WebApp daemon
# Description: Starts WebApp daemon.
### END INIT INFO
# inserire qui il path dove risiede WebApp se necessario
WebApp_DIR=/data/webhmi
# settare a 1 per abilitare il core dump; il file di trace è invece sempre generato
# COREDUMP_ENABLE=1
WebAppStart()
{
# verifica se WebApp e' già in esecuzione, e' richiesto il comando pidof !
if pidof WebApp ; then
echo "WebApp already running !"
exit 1
fi
if test "$COREDUMP_ENABLE" = "1" ; then
echo core_%e_%t > /proc/sys/kernel/core_pattern
ulimit -c unlimited
fi
cd /data/webhmi
nohup ./WebApp &
#./WebApp -daemon &
}
WaitProcessEnd()
{
cnt=0
while test -e /proc/$1 -a $cnt -lt $2 ; do
sleep 1
cnt=$((cnt + 1))
done
! test -e /proc/$1
}
WebAppStop()
{
timeout=15
# ricava il PID di WebApp, e' richiesto il comando pidof !
pid=`pidof WebApp`
if test -z $pid ; then
echo "WebApp not running !"
exit 1
fi
# tentativo di arresto con SIGTERM (arresto soft)
kill $pid
if ! WaitProcessEnd $pid $timeout ; then
# arresto soft non riuscito, tenta con SIGQUIT (arresto forzato, gestito dal signal handler)
echo "Failed to stop gracefully with SIGTERM, sending SIGQUIT ..."
kill -QUIT $pid
if ! WaitProcessEnd $pid $timeout ; then
# anche l'arresto con SIGQUIT e' fallito, manda il SIGKILL, che deve terminarlo per forza (salvo blocco nel kernel)
echo "Failed to stop with SIGQUIT, sending SIGKILL ..."
kill -KILL $pid
if ! WaitProcessEnd $pid $timeout ; then
# neanche SIGKILL ha funzionato: il processo e' bloccato nel kernel
echo "Failed to stop with SIGKILL !!! Can not stop WebApp"
fi
fi
fi
}
DESC="WebApp PLC Runtime"
case "$1" in
start)
echo -n "Starting $DESC: "
WebAppStart
echo "done."
;;
stop)
echo -n "Stopping $DESC: "
WebAppStop
echo "done."
;;
restart)
echo -n "Restarting $DESC: "
WebAppStop
sleep 1
WebAppStart
echo "done."
;;
*)
echo "Usage: WebApp {start|stop|restart}" >&2
exit 1
;;
esac
exit 0
The program works fine if manually launched by means of:
/etc/init.d/WebAppDaemon start
Or
./data/webmi/webApp
