[ntp:questions] Linux : ntp-dev-4.2.7p84 : Local clock not getting synced.

Martin Burnicki martin.burnicki at meinberg.de
Tue Jan 4 09:28:51 UTC 2011


Harlan Stenn wrote:
> Martin wrote:
>> Those "special" 127.127.x.y addresses are somewhat sorted out when
>> parsing the configuration options and treated in a special way. I'm not
>> sure it would work, and I'm not even sure if it would be expected to
>> work. Dave Mills, Dave Hart, and Danny, what's your opinion on using
>> IPv6-style addresses to specify refclocks?
> 
> I'd vote against this because I don't see the value in it.

Oops, I meant if we should expect that the current *existing* code already
accounts for pseudo IPv6 addresses for reflocks, i.e. when parsing the
config file. I haven't looked at the relevant piece of code for quite some
time.

I agree we don't need this.

> I'm more inclined to create a new config keyword, "refclock", and
> probably have it take x.y as its primary argument.  Then, for backward
> compatibility, we'd map "server 127.127.x.y ..." to "refclock x.y ...".
> 
> But I'd also want to look much harder at this, and figure out what other
> interactions there might be.
> 
> I'm also still interested in looking in to a "refclock definition
> language", and converting *all* of our refclock .c files to this new
> format.  I think it would be a useful step forward, and it would help us
> get rid of the hardcoded and centrally-managed "refclock types"
> currently hardwired in include/ntp.h .

I also appreciate Heiko's suggestion for a new syntax. Hm his posting via
the questions list has not yet made it to my news server. I'm wondering
whether it will still appear ...

An additional enhancemment which could also be introduces with such a change
would be a simpler way to assign a PPS source to a specific refclock. AFAIK
the current way is pretty hard, e.g. if you have several refclocks with PPS
output and one of them loses synchronization the PPS signal from that
device should be discarded, but the PPS signal from a different, sane
device should still be accepted, so it is important to have a specific PPS
source assigned to a specific refclock.


Martin
-- 
Martin Burnicki

Meinberg Funkuhren
Bad Pyrmont
Germany




More information about the questions mailing list