#!/bin/sh

#################################
# interface event script for ctdb
# this adds/removes IPs from your 
# public interface

. $CTDB_BASE/functions
loadconfig

[ -z "$CTDB_PUBLIC_ADDRESSES" ] && {
	CTDB_PUBLIC_ADDRESSES=$CTDB_BASE/public_addresses
}

[ ! -f "$CTDB_PUBLIC_ADDRESSES" ] && {
	exit 0
}

monitor_interfaces()
{
	INTERFACES=`cat $CTDB_PUBLIC_ADDRESSES |
		sed -e "s/^[^\t ]*[\t ]*//" -e "s/,/ /g" -e "s/[\t ]*$//"`

	[ "$CTDB_PUBLIC_INTERFACE" ] && INTERFACES="$CTDB_PUBLIC_INTERFACE $INTERFACES"
	[ "$CTDB_NATGW_PUBLIC_IFACE" ] && INTERFACES="$CTDB_NATGW_PUBLIC_IFACE $INTERFACES"


	# For all but the 1st line, get the 2nd last field with commas
	# changes to spaces.
	IFACES=`ctdb -Y ip -v | sed -e '1d' -e 's/:[^:]*:$//' -e 's/^.*://' -e 's/,/ /g'`

	INTERFACES=`for IFACE in $INTERFACES $IFACES ; do echo $IFACE ; done | sort | uniq`

	fail=0
	ok=0
	for IFACE in $INTERFACES ; do

	    ip addr show $IFACE 2>/dev/null >/dev/null || {
		echo Interface $IFACE does not exist but it is used by public addresses.
		continue
	    }

	    # These interfaces are sometimes bond devices
	    # When we use VLANs for bond interfaces, there will only
	    # be an entry in /proc for the underlying real interface
	    REALIFACE=`echo $IFACE |sed -e 's/\..*$//'`
	    bi=$(get_proc "net/bonding/$REALIFACE" 2>/dev/null) && {
		echo "$bi" | grep -q 'Currently Active Slave: None' && {
			echo "ERROR: No active slaves for bond device $REALIFACE"
			fail=1
			ctdb setifacelink $IFACE down >/dev/null 2>/dev/null
			continue;
		}
		echo "$bi" | grep -q '^MII Status: up' || {
			echo "ERROR: public network interface $REALIFACE is down"
			fail=1
			ctdb setifacelink $IFACE down >/dev/null 2>/dev/null
			continue;
		}
		echo "$bi" | grep -q '^Bonding Mode: IEEE 802.3ad Dynamic link aggregation' && {
			echo "$bi" | grep 'MII Status:' | tail -n +2 | grep -q '^MII Status: up' || {
				echo No active slaves for 802.ad bond device $REALIFACE
				ctdb setifacelink $IFACE down >/dev/null 2>/dev/null
				fail=1
				continue
			}
		}
		ok=1 # we only set ok for interfaces known to ctdbd
		ctdb setifacelink $IFACE up >/dev/null 2>/dev/null
		continue;
	    }

	    case $IFACE in
	    lo*)
		# loopback is always working
		ok=1 # we only set ok for interfaces known to ctdbd
		ctdb setifacelink $IFACE up >/dev/null 2>/dev/null
		;;
	    ib*)
		# we dont know how to test ib links
		ok=1 # we only set ok for interfaces known to ctdbd
		ctdb setifacelink $IFACE up >/dev/null 2>/dev/null
		;;
	    *)
		[ -z "$IFACE" ] || {
		    [ "$(basename $(readlink /sys/class/net/$IFACE/device/driver) 2>/dev/null)" = virtio_net ] ||
		    ethtool $IFACE | grep -q 'Link detected: yes' || {
			# On some systems, this is not successful when a
			# cable is plugged but the interface has not been
			# brought up previously. Bring the interface up and
			# try again...
			ip link set $IFACE up
			ethtool $IFACE | grep -q 'Link detected: yes' || {
			    echo "ERROR: No link on the public network interface $IFACE"
			    fail=1
			    ctdb setifacelink $IFACE down >/dev/null 2>/dev/null
			    continue
			}
		    }
		    ok=1 # we only set ok for interfaces known to ctdbd
		    ctdb setifacelink $IFACE up >/dev/null 2>/dev/null
		}
		;;
	    esac

	done

	test x"$fail" = x"0" && {
		return 0;
	}

	test x"$ok" = x"1" && {
		return 2;
	}

	return 1;
}

case "$1" in 
     #############################
     # called when ctdbd starts up
     init)
	# make sure that we only respond to ARP messages from the NIC where
	# a particular ip address is associated.
	get_proc sys/net/ipv4/conf/all/arp_filter >/dev/null 2>&1 && {
	    set_proc sys/net/ipv4/conf/all/arp_filter 1
	}
	;;

     #############################
     # called after ctdbd has done its initial recovery
     # and we start the services to become healthy
     startup)
	# Assume all links are good initially
	INTERFACES=`for IFACE in $INTERFACES ; do echo $IFACE ; done | sort | uniq`

	for IFACE in $INTERFACES ; do
		ctdb setifacelink $IFACE down >/dev/null 2>/dev/null
	done
	
	monitor_interfaces

	;;


     ################################################
     # called when ctdbd wants to claim an IP address
     takeip)
	if [ $# != 4 ]; then
	   echo "must supply interface, IP and maskbits"
	   exit 1
	fi
	iface=$2
	ip=$3
	maskbits=$4

	add_ip_to_iface $iface $ip $maskbits || {
		exit 1;
	}

	# cope with the script being killed while we have the interface blocked
	iptables -D INPUT -i $iface -d $ip -j DROP 2> /dev/null

	# flush our route cache
	set_proc sys/net/ipv4/route/flush 1
	;;


     ##################################################
     # called when ctdbd wants to release an IP address
     releaseip)
	if [ $# != 4 ]; then
	   echo "must supply interface, IP and maskbits"
	   exit 1
	fi

	# releasing an IP is a bit more complex than it seems. Once the IP
	# is released, any open tcp connections to that IP on this host will end
	# up being stuck. Some of them (such as NFS connections) will be unkillable
	# so we need to use the killtcp ctdb function to kill them off. We also
	# need to make sure that no new connections get established while we are 
	# doing this! So what we do is this:
	# 1) firewall this IP, so no new external packets arrive for it
	# 2) use netstat -tn to find existing connections, and kill them 
	# 3) remove the IP from the interface
	# 4) remove the firewall rule
	iface=$2
	ip=$3
	maskbits=$4

	failed=0
	# we do an extra delete to cope with the script being killed
	iptables -D INPUT -i $iface -d $ip -j DROP 2> /dev/null
	iptables -I INPUT -i $iface -d $ip -j DROP
	kill_tcp_connections $ip

	delete_ip_from_iface $iface $ip $maskbits || {
		iptables -D INPUT -i $iface -d $ip -j DROP 2> /dev/null
		exit 1;
	}

	iptables -D INPUT -i $iface -d $ip -j DROP 2> /dev/null

	# flush our route cache
	set_proc sys/net/ipv4/route/flush 1
	;;

     ##################################################
     # called when ctdbd wants to update an IP address
     updateip)
	if [ $# != 5 ]; then
	   echo "must supply old interface, new interface, IP and maskbits"
	   exit 1
	fi

	# moving an IP is a bit more complex than it seems.
	# First we drop all traffic on the old interface.
	# Then we try to add the ip to the new interface and before
	# we finally remove it from the old interface.
	#
	# 1) firewall this IP, so no new external packets arrive for it
	# 2) add the IP to the new interface
	# 3) remove the IP from the old interface
	# 4) remove the firewall rule
	# 5) use ctdb gratiousarp to propagate the new mac address
	# 6) use netstat -tn to find existing connections, and tickle them
	oiface=$2
	niface=$3
	ip=$4
	maskbits=$5

	failed=0
	# we do an extra delete to cope with the script being killed
	iptables -D INPUT -i $oiface -d $ip -j DROP 2> /dev/null
	iptables -I INPUT -i $oiface -d $ip -j DROP

	delete_ip_from_iface $oiface $ip $maskbits 2>/dev/null
	delete_ip_from_iface $niface $ip $maskbits 2>/dev/null

	add_ip_to_iface $niface $ip $maskbits || {
		iptables -D INPUT -i $oiface -d $ip -j DROP 2> /dev/null
		exit 1;
	}

	# cope with the script being killed while we have the interface blocked
	iptables -D INPUT -i $oiface -d $ip -j DROP 2> /dev/null

	# flush our route cache
	set_proc sys/net/ipv4/route/flush 1

	# propagate the new mac address
	ctdb gratiousarp $ip $niface

	# tickle all existing connections, so that dropped packets
	# are retransmited and the tcp streams work

	tickle_tcp_connections $ip

	;;


     ###########################################
     # called when ctdbd has finished a recovery
     recovered)
	;;

     ####################################
     # called when ctdbd is shutting down
     shutdown)
	;;

     monitor)
	monitor_interfaces
	ret=$?

	test x"$ret" = x"2" && {
		test x"$CTDB_PARTIALLY_ONLINE_INTERFACES" != x"yes" && {
			exit 1;
		}
		# as long as we have one interface available don't become
		# unhealthy
		ret=0
	}

	test x"$ret" != x"0" && {
		exit 1;
	}
	;;
    *)
	ctdb_standard_event_handler "$@"
	;;
esac

exit 0