#!/bin/bash
#
# lnet	This shell script takes care of starting and stopping
#       the lnet (Lustre networking) services.
#
# chkconfig: - 59 76
# description:  Part of the lustre file system.
# probe: true
# config: /etc/sysconfig/lnet

# Source function library.
[ -f /etc/rc.d/init.d/functions ] && . /etc/rc.d/init.d/functions

# Source networking configuration and check that networking is up.
[ -f /etc/sysconfig/network ] && . /etc/sysconfig/network && \
[ "${NETWORKING}" = "no" ] && exit 0

# Check for and source configuration file otherwise set defaults
[ -f /etc/sysconfig/lnet ] && . /etc/sysconfig/lnet

declare -r awkprog='BEGIN { rc = -1 }
		       { if ( $1 == module_name ) { rc = $3; exit; } }
		    END { print rc }'

# Usage: run_preexec_check [ start | restart | condrestart ]
# The single parameter will be passed to the PREEXEC_SCRIPT
run_preexec_check ()
{
	if [ -n "$PREEXEC_CHECK" ] && ! $PREEXEC_CHECK ; then
		echo "Pre-exec check \"$PREEXEC_CHECK\" failed.  Aborting."
		exit 1
	fi

	if [ -n "$PREEXEC_SCRIPT" ] && ! "$PREEXEC_SCRIPT" "$1" ; then
		echo "Pre-exec script \"$PREEXEC_SCRIPT\" failed.  Aborting."
		exit 1
	fi
}

# Usage: run_postexec_check [ start | restart | condrestart ]
# The single parameter will be passed to the POSTEXEC_SCRIPT
run_postexec_check ()
{
	if [ -n "$POSTEXEC_CHECK" ] && ! $POSTEXEC_CHECK ; then
		echo "Post-exec check \"$POSTEXEC_CHECK\" failed.  Aborting."
		exit 1
	fi

	if [ -n "$POSTEXEC_SCRIPT" ] && ! "$POSTEXEC_SCRIPT" "$1" ; then
		echo "Post-exec script \"$POSTEXEC_SCRIPT\" failed.  Aborting."
		exit 1
	fi
}

stop_lnet ()
{
	local errmsg=$(/usr/sbin/lctl network down 2>&1)
	if [ $? -gt 0 ]; then
		# The following error message means that lnet is already
		# unconfigured, and the modules are not loaded.
		echo $errmsg | grep "LNET unconfigure error 19" > /dev/null
		if [ $? -gt 0 ]; then
			return 0
		else
			echo "$errmsg"
			return 1
		fi
	fi
	return 0
}

status ()
{
	old_nullglob="`shopt -p nullglob`"
	shopt -u nullglob

	STATE="stopped"
	# LSB compliance - return 3 if service is not running
	# Lustre-specific returns
	# 150 - partial startup
	# 151 - health_check unhealthy
	# 152 - LBUG
	RETVAL=3
	egrep -q "lnet" /proc/modules && STATE="loaded"

	# check for any routes - on a portals router this is the only thing
	VAR=$(lctl get_param -n routes 2>&1)
	if [ $? = 0 ] ; then
		STATE="running"
		RETVAL=0
	fi

	# check if this is a router
	if [[ "$(lctl get_param -n routes)" =~ "Routing enabled" ]]; then
		STATE="running"
		RETVAL=0
	fi

	# check for error in health_check
	local health_check=$(lctl get_param -n health_check)
	if [[ "$health_check" =~ "NOT HEALTHY" ]]; then
		STATE="unhealthy"
		RETVAL=1
	fi

	if [[ "$health_check" =~ "LBUG" ]]; then
		STATE="LBUG"
		RETVAL=152
	fi

	echo $STATE
	eval $old_nullglob
}

LUSTRE_ROUTES_CONFIG_FILE="/etc/sysconfig/lnet_routes.conf"
LUSTRE_LNET_CONFIG_FILE="/etc/sysconfig/lnet.conf"
LUSTRE_LNET_CONFIG_UTILITY="/usr/sbin/lnetctl"

# See how we were called.
case "$1" in
  start)
	run_preexec_check "start"
	touch /var/lock/subsys/lnet
	modprobe lnet || exit 1
	# if lnet.conf file exists then use lnetctl lnet configure, since
	# that doesn't load the networks and routes defined in the mod
	# params.  The appropriate configuration will be picked up from
	# the lnet.conf YAML file.
	if [ -x $LUSTRE_LNET_CONFIG_UTILITY -a -f "$LUSTRE_LNET_CONFIG_FILE" ]; then
		$LUSTRE_LNET_CONFIG_UTILITY lnet configure || exit 1
	else
		lctl network up || exit 1
	fi
	# if an lnet.conf file exists then pass that to the lnetctl
	# utility for parsing.  This will configure the items defined
	# in YAML format in the config file.
	if [ -x $LUSTRE_LNET_CONFIG_UTILITY -a -f "$LUSTRE_LNET_CONFIG_FILE" ]; then
		$LUSTRE_LNET_CONFIG_UTILITY import < $LUSTRE_LNET_CONFIG_FILE
	fi
	# if a routes config file is given then use it to configure the
	# routes if not then default to LUSTRE_ROUTES_CONFIG_FILE
	if [ -f "$2" ]; then
		lustre_routes_config $2
	elif [ -f "$LUSTRE_ROUTES_CONFIG_FILE" ]; then
		lustre_routes_config $LUSTRE_ROUTES_CONFIG_FILE
	fi
	run_postexec_check "start"
	;;
  stop)
	run_preexec_check "stop"
	stop_lnet || exit 1
	lustre_rmmod || exit 1
	rm -f /var/lock/subsys/lnet
	run_postexec_check "stop"
	;;
  status)
	status
	;;
  restart)
	$0 stop
	$0 start
	;;
  reload)
	touch /var/lock/subsys/lnet
	;;
  probe)
	if [ ! -f /var/lock/subsys/lnet ] ; then
	  echo $"start"; exit 0
	fi
	;;
  condrestart)
	[ -f /var/lock/subsys/lnet ] && {
		$0 stop
		$0 start
	}
	;;
  *)
	echo $"Usage: lnet {start|stop|status|restart|reload|condrestart}"
	exit 1
esac

exit 0