[ntp:questions] NTP Limited Broadcast

Kevin D. Casey kevin at kdcasey.net
Mon Mar 19 19:47:22 UTC 2007

I found the solution to this problem. NTP (at least on FreeBSD) does not 
broadcast on interfaces not associated with a broadcast command in ntp.conf. 
So in order to broadcast on at least one network interface 
must be associated with a network that has the broadcast address In my case I accomplished this by installing a 2nd NIC and 
configuring it with the address This class E network address 
has as it's broadcast address and so NTP is able to 
broadcast on it. I'm not sure if this is THE ONLY solution, but it worked 
for me.


----- Original Message ----- 
From: <mills at udel.edu>
Newsgroups: comp.protocols.time.ntp
To: <questions at lists.ntp.isc.org>
Sent: Tuesday, March 13, 2007 9:47 AM
Subject: Re: [ntp:questions] NTP Limited Broadcast

> Danny,
> You scratch an old wound. When I wrote the first RFC on internet
> subnetting the conventional interpretation of was
> broadcast to the world without limit. I though this a dumb idea and
> intended that broadcasts never went beyond the first-hop gateway.
> In review, my document did not say that explicitly; however, Bob Braden,
> who wrote the sequel, did strongly discourage propagation of broadcasts
> beyond the first-hop gateway. Therefore, the considered engineering
> judgement would be that anything other than the subnet bits at the left
> end of the address and other than the one bits at the right end be
> discarded forthwith. If the vendor in question does not understand that,
> switch to a different vendor.
> Once upon a time a favorite bogon was a packet with all ones in the
> destination AND source addresses, preferably in an ICMP echo request
> packet. Then measure the Ethernet temperature as result. I named such as
> Chernobyl packets and the result as broadcast storms. Gateways had what
> came to be known as Martian filters that grounded packets with bogus
> source and/or destination addresses. I assume modern gateways have
> equivalent defenses.
> Dave
> Danny Mayer wrote:
>> Kevin D. Casey wrote:
>>>I am running NTP 4.2.2 on FreeBSD 6.2 broadcasting to clients on the
>>>broadcast address of an internal subnet which is
>>> The server is using on its network
>>>interface. The attached switch hardware on the subnet is able to
>>>receive the broadcasts and update its clock as an SNTP client. I have
>>>also verified the presence of UDP 123 broadcasts using Ethereal. I
>>>have a proprietary time recording device also attached to the same
>>>broadcast domain that appears to be unable to receive it's updates in
>>>this way. Although it is designed to listen for UDP 123 broadcast
>>>traffic, the manufacturer is claiming that it will only listen for
>>>broadcasts sent to However when I try to reconfigure
>>>NTPD to use instead of the broadcast
>>>stops working. Can anyone tell me how to configure NTPD to broadcast
>>>on -- Kevin D. Casey kevin at kdcasey.net
>> See Bug #779 https://ntp.isc.org/bugs/show_bug.cgi?id=779
>> This has become a difficult issue to resolve. Although a socket gets set
>> up to accept broadcast packets packets sent to only get
>> delivered to the wildcard socket where all packets get dropped. I'm not
>> sure why this is but it looks like all O/S's seem to ignore the
>> configured socket for this case in favor of the wildcard socket. I'm
>> trying to research this issue.
>> Danny
>> _______________________________________________
>> questions mailing list
>> questions at lists.ntp.isc.org
>> https://lists.ntp.isc.org/mailman/listinfo/questions
> _______________________________________________
> questions mailing list
> questions at lists.ntp.isc.org
> https://lists.ntp.isc.org/mailman/listinfo/questions

More information about the questions mailing list