User guide

21
following commands
switch --enable-port-block-unknown-multicast=1:1
switch --enable-port-block-unknown-multicast=2:1
switch --enable-port-block-unknown-multicast=3:1
switch --enable-port-block-unknown-multicast=4:1
switch --enable-port-block-unknown-multicast=5:1 # Important to include CPU
port, alternatively a membership to any of the multicast groups could cause the
CPU to be overloaded.
# applicable to 7/8 port models only
switch --enable-port-block-unknown-multicast-ext=2:1
switch --enable-port-block-unknown-multicast-ext=4:1
# It is recommended NOT to block unknown multicasts on port 0 (WAN), as all
client broadcasts are blocked, as broadcasts are considered as multicasts
# This will prevent unknown multicast traffic to be passed out on any of the
LAN ports. So it requires a join message on the specific LAN port, from a set-
top box, in order to have traffic passed on to that LAN port
# The IGMP can be shut off via SNMP, use dkt_ge.mib OID named “geIGMPSnooping”,
input parameter 0 = stops IGMP snooping. Alternatively in the configuration
file add the following: “mv /etc/init.d/igmp /etc/init.d/igmp2”
Note that blocking multicasts for port 5, will result in a blocking of ARPs originated for the WAN port also,
as all broadcasts are considered as multicasts. In order to ensure that ARPs from the DHCP Server is still
passed through, ARP mirroring function must be enabled.
switch --set-arp-mirroring=0:1 # Enables ARP mirroring, so ARPs received on WAN
port will be redirected to CPU port, despite the fact that multicast/broadcasts
are blocked on the CPE port
WhenanIGMPjoinisseenforamulticastgroup,thenswitchATUisconguredwithaltersettingfor
thisgroup,allowingtrafcforthegrouptobebridgedtothespecicport.
Perdefault,theltersettingsarekeptuntilappropriateIGMPleaveisseenfortheparticularmulticast
group.
This means that once a device has joined a multicast channel then the address will be present in the
switch ATU until a leave is received by the IGMP snooper.
It is possible to enable a timeout in the IGMP snooper with default timeout values, the user must put this
lineintheCPEcongurationle:
ENABLE_IGMP_TIMEOUT=1 /etc/init.d/igmp restart
#or “enable_igmp_timeout=1 /etc/init.d/igmp restart”
WhentheCPEdetectsIGMPv2/v3joinpacket,andwillestablishalterforthemulticastgroupforthe
specicportseen,andsetadefaulttime-outvalue(forIGMPv2thespeciedtime-outvalueintheRFC,
forIGMPv3thevalueextractedfromthejoinpacket)forthisspecicmulticastgroup.Theswitchlter
will start to decrease this value immediately.
It is expected that a IGMP Query is present, followed by a IGMP report from the client, saying that it is
still a member of the group
Insomecasesraceconditionscanoccur,iftheCPEltertimesoutbeforeaclienthasrespondedtoan
IGMP query, hence signal will be lost, as CPE expect that the client is no longer member of the multicast
group.
Therefore it is possible to change the default time-out values for the IGMP snooping.