[pkg-ntp-maintainers] Bug#513620: Bug#513620: ntp: Please check for a loopback interface

[Vince Mulhollon]

On Sat, Jan 31, 2009 at 07:46:58PM +0100, Kurt Roeckx wrote:
> How do you expect "ntpq" to connect to localhost when you didn't
> configure it?  Note that the manpage clearly states that
> localhost is the default.

OK maybe my bug report was not so well written.  Sorry about that.

If ntpd can not be contacted by ntpq, perhaps because of "/etc/init.d/ntp 
stop" then you get the following error from ntpq when it can not reach 
ntpd:
ntpq: read: Connection refused

I am familiar with that error message.

On the other hand, if ntpd is up and running perfectly, but interface lo is 
down, such as "ifdown lo", then you get the following error from ntpq when
it can not reach ntpd:
localhost: timed out, nothing received
***Request timed out.

That's odd...  It would seem more intuitive if both failures to contact 
ntpd errored out the same way.  Why two different error messages for the 
same situation of ntpq cannot contact ntpd, depending on which ip 
interfaces happen to be up or down?

> And I'm pretty sure that there are alot of things that break
> if you do not have localhost configured. 

Oh, you might be surprised if it happened to you, in my semi-embedded 
nfs-root situation, due to a misconfiguration removing lo, all that 
failed was NFS statd would hang on startup with no error or log, portmap 
would not shutdown cleanly (although that happens late in shutdown, and 
who ever notices that on an embedded machine) and ntpq gave that unusual
response when unable to talk to ntpd.

We can agree, if you'd like, that I'll work beginning from the point that 
almost nothing fails without a localhost, and you can work beginning from 
the point that alot of things break, but that all doesn't really change 
anything in this particular package.

It's only a wishlist bug because I agree it's not the end of the world.

Anyway, thanks for your work on ntp, and have a nice day.