ntpq change in behaviour
Eric S. Raymond
esr at thyrsus.com
Sun Jan 20 22:44:57 UTC 2019
Achim Gratz via devel <devel at ntpsec.org>:
> Eric S. Raymond via devel writes:
> > Well, shit. Achim, have you IDed an actual bug, or is that a guess based
> > on the timing of the report?
>
> No, I've updated two boxes yesterday and they've reported a loop
> frequency of essentially (but not exactly) zero since. Something is
> still keeping the clock on track (kernel PLL?), but less accurate than
> normal. PLL state is randomly reported as synced and unsynced.
> Reverted back to before the lockclock change and they are snapping back
> to normal.
>
> I have no teyeballed something that jumps out at me, but I notice that
> lockclock is never actively initialized. Since it's defined as a
> static, I'm not sure the compiler puts that in for you.
C is in fact supposed to zero uninitialized statics.
But it sounds like there's some other bug. Can't say I'm overly surprised;
the loopfilter code has a long history of being fiendishly fragile, I knew
this change was risky.
Your report puts me in a bad spot. You obviously know how to tell when
it's busted, but I don't. This is one of those times we get bitten by
the fact that my knowledge of NTP operation from the outside is still
very weak. In effect, I can't test the change myself - I don't know
enough about what "normal" looks like from outside the code.
I can of course just revert the change, but there's an actual reason for
it. We've (a) had a report of a distro maker setting ENABLE_LOCKCLOCK
on, disabling time sync, and (b) we want to not potentially cheese off
NIST by removing the support. Thus, the attempt to change it to a runtime
option.
Therefore...would you please take a hard look at the change and see if
you can repair it? I hate to slough off my mistakes on anybody, but the
fix needs to be applied by somebody who can test it.
--
<a href="http://www.catb.org/~esr/">Eric S. Raymond</a>
My work is funded by the Internet Civil Engineering Institute: https://icei.org
Please visit their site and donate: the civilization you save might be your own.
More information about the devel
mailing list