#!/bin/sh

# Exit on 1st error
set -e

# NFS exports file.  Some code below keeps a cache of output derived
# from exportfs(8).  When this file is updated the cache is invalid
# and needs to be regenerated.
#
# To change the file, edit the default value below.  Do not set
# CTDB_NFS_EXPORTS_FILE - it isn't a configuration variable, just a
# hook for testing.
nfs_exports_file="${CTDB_NFS_EXPORTS_FILE:-/var/lib/nfs/etab}"

# As above, edit the default value below.  CTDB_NFS_DISTRO_STYLE is a
# test variable only.
nfs_distro_style="${CTDB_NFS_DISTRO_STYLE:-systemd-redhat}"

# As above, edit the default value below.  CTDB_SYS_ETCDIR is a
# test variable only.
etc_dir="${CTDB_SYS_ETCDIR:-/etc}"

# A value of "AUTO" for any service means that service is usually
# automatically started and stopped by one of the other services.
# Such services will still be restarted by hand on failure, if
# configured to do so.  This allows services that should not be
# running to be set to "".

case "$nfs_distro_style" in
systemd-*)
	# Defaults
	nfs_service="nfs-server"
	nfs_lock_service="rpc-statd"
	nfs_mountd_service="nfs-mountd"
	nfs_status_service="rpc-statd"
	nfs_rquotad_service="rpc-rquotad"
	nfs_config="${etc_dir}/sysconfig/nfs"
	nfs_rquotad_config="" # Not use with systemd, restart via service

	case "$nfs_distro_style" in
	*-redhat | *-suse)
		: # Defaults only
		;;
	*-debian)
		nfs_rquotad_service="quotarpc"
		;;
	*)
		echo "Internal error"
		exit 1
		;;
	esac
	;;

sysvinit-*)
	# Defaults
	nfs_service="nfs"
	nfs_lock_service="AUTO"
	nfs_mountd_service="AUTO"
	nfs_status_service="AUTO"
	nfs_rquotad_service="AUTO"
	nfs_config="${etc_dir}/sysconfig/nfs"
	nfs_rquotad_config="$nfs_config"

	case "$nfs_distro_style" in
	*-redhat)
		nfs_lock_service="nfslock"
		;;
	*-suse)
		nfs_service="nfsserver"
		;;
	*-debian)
		nfs_service="nfs-kernel-server"
		nfs_config="${etc_dir}/default/nfs-kernel-server"
		nfs_rquotad_config="${etc_dir}/default/quota"
		;;
	*)
		echo "Internal error"
		exit 1
		;;
	esac
	;;

*)
	echo "Internal error"
	exit 1
	;;
esac

# Override for unit testing
if [ -z "$PROCFS_PATH" ]; then
	PROCFS_PATH="/proc"
fi

##################################################

usage()
{
	_c=$(basename "$0")
	cat <<EOF
usage: $_c { shutdown | startup }
       $_c { stop | start } { nfs | nlockmgr }
       $_c { monitor-list-shares | monitor-post }
       $_c { register }
EOF
	exit 1
}

##################################################

nfs_load_config()
{
	_config="${1:-${nfs_config}}"

	if [ -r "$_config" ]; then
		. "$_config"
	fi
}

##################################################

service_is_auto_started()
{
	[ "$1" = "AUTO" ]
}

service_is_defined()
{
	_service="$1"

	[ -n "$_service" ] && ! service_is_auto_started "$_service"
}

service_if_defined()
{
	_service="$1"
	_action="$2"

	if service_is_defined "$_service"; then
		service "$_service" "$_action"
	fi
}

##################################################
# Overall NFS service stop and start

nfs_service_stop()
{
	service_if_defined "$nfs_rquotad_service" stop

	service "$nfs_service" stop

	service_if_defined "$nfs_lock_service" stop
}

nfs_service_start()
{
	service_if_defined "$nfs_lock_service" start

	service "$nfs_service" start

	service_if_defined "$nfs_rquotad_service" start
}

##################################################
# service "stop" and "start" options for restarting

manual_stop()
{
	case "$1" in
	mountd)
		killall -q -9 rpc.mountd
		;;
	rquotad)
		killall -q -9 rpc.rquotad
		;;
	status)
		killall -q -9 rpc.statd
		;;
	*)
		echo "$0: Internal error - invalid call to manual_stop()"
		exit 1
		;;
	esac
}

service_or_manual_stop()
{
	_rpc_service="$1"
	_system_service="$2"

	if service_is_defined "$_system_service"; then
		service "$_system_service" stop
	elif service_is_auto_started "$_system_service"; then
		manual_stop "$_rpc_service"
	fi
}

service_stop()
{
	_rpc_service="$1"

	case "$_rpc_service" in
	nfs)
		echo 0 >"${PROCFS_PATH}/fs/nfsd/threads"
		nfs_service_stop >/dev/null 2>&1 || true
		pkill -9 nfsd
		;;
	nlockmgr)
		if service_is_defined "$nfs_lock_service" ; then
			service "$nfs_lock_service" stop >/dev/null 2>&1 || true
		else
			service "$nfs_service" stop >/dev/null 2>&1 || true
		fi
		;;
	mountd)
		service_or_manual_stop "$_rpc_service" "$nfs_mountd_service"
		;;
	rquotad)
		service_or_manual_stop "$_rpc_service" "$nfs_rquotad_service"
		;;
	status)
		service_or_manual_stop "$_rpc_service" "$nfs_status_service"
		;;
	*)
		usage
		;;
	esac
}

manual_start()
{
	case "$1" in
	mountd)
		nfs_load_config
		if [ -z "$RPCMOUNTDOPTS" ]; then
			RPCMOUNTDOPTS="${MOUNTD_PORT:+-p }$MOUNTD_PORT"
		fi
		# shellcheck disable=SC2086
		rpc.mountd $RPCMOUNTDOPTS
		;;
	rquotad)
		nfs_load_config "$nfs_rquotad_config"
		if [ -z "$RPCRQUOTADOPTS" ]; then
			RPCRQUOTADOPTS="${RQUOTAD_PORT:+-p }$RQUOTAD_PORT"
		fi
		# shellcheck disable=SC2086
		rpc.rquotad $RPCRQUOTADOPTS
		;;
	status)
		nfs_load_config
		# Red Hat uses STATDARG, Debian uses STATDOPTS
		opts="${STATDARG:-${STATDOPTS:-''}}"
		if [ -z "$opts" ]; then
			# shellcheck disable=SC2086
			set -- \
				${STATD_HA_CALLOUT:+-H} $STATD_HA_CALLOUT \
				${STATD_HOSTNAME:+-n} $STATD_HOSTNAME \
				${STATD_PORT:+-p} $STATD_PORT \
				${STATD_OUTGOING_PORT:+-o} $STATD_OUTGOING_PORT
			opts="$*"
		fi
		# shellcheck disable=SC2086
		rpc.statd $opts
		;;
	*)
		echo "$0: Internal error - invalid call to manual_start()"
		exit 1
		;;
	esac
}

service_or_manual_start()
{
	_rpc_service="$1"
	_system_service="$2"

	if service_is_defined "$_system_service"; then
		service "$_system_service" start
	elif service_is_auto_started "$_system_service"; then
		manual_start "$_rpc_service"
	fi
}

service_start()
{
	_rpc_service="$1"

	case "$_rpc_service" in
	nfs)
		nfs_service_start
		;;
	nlockmgr)
		if service_is_defined "$nfs_lock_service" ; then
			service "$nfs_lock_service" start
		else
			service "$nfs_service" start
		fi
		;;
	mountd)
		service_or_manual_start "$_rpc_service" "$nfs_mountd_service"
		;;
	rquotad)
		service_or_manual_start "$_rpc_service" "$nfs_rquotad_service"
		;;
	status)
		service_or_manual_start "$_rpc_service" "$nfs_status_service"
		;;
	*)
		usage
		;;
	esac
}

##################################################
# service init startup and final shutdown

nfs_shutdown()
{
	nfs_service_stop
}

nfs_startup()
{
	nfs_service_stop || true
	nfs_service_start
	_f="${PROCFS_PATH}/sys/net/ipv4/tcp_tw_recycle"
	if [ -f "$_f" ]; then
		echo 1 >"$_f"
	fi
}

##################################################
# monitor-post support

nfs_check_thread_count()
{
	# Load NFS configuration to get desired number of threads.
	nfs_load_config

	# If $RPCNFSDCOUNT/$USE_KERNEL_NFSD_NUMBER isn't set then we could
	# guess the default from the initscript.  However, let's just
	# assume that those using the default don't care about the number
	# of threads and that they have switched on this feature in error.
	_configured_threads="${RPCNFSDCOUNT:-${USE_KERNEL_NFSD_NUMBER}}"
	if [ -z "$_configured_threads" ] && type nfsconf >/dev/null 2>&1; then
		_configured_threads=$(nfsconf --get nfsd threads) || true
	fi
	[ -n "$_configured_threads" ] || return 0

	_threads_file="${PROCFS_PATH}/fs/nfsd/threads"

	# nfsd should be running the configured number of threads.  If
	# there are a different number of threads then tell nfsd the
	# correct number.
	read -r _running_threads <"$_threads_file" || {
		echo "WARNING: Reading \"${_threads_file}\" unexpectedly failed"
		exit 0
	}

	# Intentionally not arithmetic comparison - avoids extra errors
	# when above read fails in an unexpected way...
	if [ "$_running_threads" != "$_configured_threads" ]; then
		echo "Attempting to correct number of nfsd threads from ${_running_threads} to ${_configured_threads}"
		echo "$_configured_threads" >"$_threads_file"
	fi
}

##################################################
# list share directories

nfs_monitor_list_shares()
{
	_cache_file="${CTDB_NFS_CALLOUT_STATE_DIR}/list_shares_cache"
	# -nt operator is well supported in Linux: dash, bash, ksh, ...
	# shellcheck disable=SC2039,SC3013
	if [ ! -r "$nfs_exports_file" ] || [ ! -r "$_cache_file" ] ||
		[ "$nfs_exports_file" -nt "$_cache_file" ]; then
		mkdir -p "$CTDB_NFS_CALLOUT_STATE_DIR"
		# We could just use the contents of $nfs_exports_file.
		# However, let's regard that file as internal to NFS and use
		# exportfs, which is the public API.
		if ! _exports=$(exportfs -v); then
			echo "WARNING: failed to run exportfs to list NFS shares" >&2
			return
		fi

		echo "$_exports" |
			grep '^/' |
			sed -e 's@[[:space:]][[:space:]]*[^[:space:]()][^[:space:]()]*([^[:space:]()][^[:space:]()]*)$@@' |
			sort -u >"$_cache_file"
	fi

	cat "$_cache_file"
}

##################################################

nfs_register()
{
	cat <<EOF
shutdown
startup
stop
start
monitor-list-shares
monitor-post
EOF
}

##################################################

case "$1" in
shutdown)
	nfs_shutdown
	;;
startup)
	nfs_startup
	;;
stop)
	service_stop "$2"
	;;
start)
	service_start "$2"
	;;
monitor-list-shares)
	nfs_monitor_list_shares
	;;
monitor-post)
	nfs_check_thread_count
	;;
register)
	nfs_register
	;;
monitor-pre | releaseip | takeip | releaseip-pre | takeip-pre)
	# Not required/implemented
	:
	;;
*)
	usage
	;;
esac