[ntp:hackers] A little cleaning and tos maxdist
Brian.Utterback at Sun.COM
Wed Feb 2 20:30:58 PST 2005
Danny Mayer wrote:
> At 10:44 PM 2/2/2005, David L. Mills wrote:
>> After discovering the situation reported on the newsgroup where
>> misconfigured NTP had the Netgear disease, I put it through some
>> wickedly broken configuration tests and found and fixed a few minor
>> things in ntp-dev (not anything like the PPS oscillator).
>> 1. If a notrust error and iburst are configured, the client did not
>> back off the poll interval. It now does that under all circumstances
>> where the reply is absent or corrupt.
>> 2. An unresponsive server with iburst got hammered with eight
>> messages every poll interval until the unreach counter reaches 12,
>> then backs off sending only a single packet per burst. I changed that
>> to send a burst only the first time after initial start or loss of
>> 3. By popular demand I put in "tos maxdist N", where N is the
>> selection threshold, normally 1. Set N to 16 and the clock will be
>> set upon first receiving a response from any server. Y'know, like SNTP.
> Or ntpdate? Isn't this dangerous? So any falseticker can royally hose
>> hackers mailing list
>> hackers at support.ntp.org
> hackers mailing list
> hackers at support.ntp.org
As Dave and I have pointed out in the past, one falseticker can royally
hose you anyway. I have been
thinking about a scheme to prevent that, but in the meantime, this
change just allows you to get hosed
a bit faster than at present 8-). Remember, during the initial startup
dance, at some point in time there
will arrive a packet that will bring one server into acceptability.
Since it will be the only server left
on the island at the final choice, it will be chosen, even though it may
be a damned liar.
More information about the hackers