[ntp:questions] WARNING: someone's faking a leap second tonight

Jeffrey Lerman jeffrey.lerman at gmail.com
Sat Aug 4 18:13:10 UTC 2012



On Fri, Aug 03 2012 at 5:42PM, Harlan Stenn <stenn at ntp.org> wrote:
> Jeff wrote:
>> Is the leap bit supposed to be cleared by a client if it gets LI=00
>> from a server?  Or is the bit only *set* based on information from a
>> server, and cleared only upon application of the leap second?  If the
>> latter is the current implementation, it might well explain the bogus
>> leap second behavior many of us saw a few days ago.  Unless you have a
>> different explanation/understanding?
> I'd have to look all that up, and I know different versions behave
> differently.
>
> This topic is something that's getting a lot of recent discussion and
> scrutiny...
Yes.  The unfortunate combination of the bogus leap second and the 
newly-discovered (on July 1) Linux kernel bug related to leap-second 
handling means that bogus leap seconds have a much bigger-than-normal 
impact.

It looks like this recently-filed (and cryptically-named) ntpd bug might 
be related to the bogus leap seconds?
http://bugs.ntp.org/show_bug.cgi?id=2246   "sys_leap is stick"

If so, that bug possibly ought to be bumped up in priority.

Meantime if we can confirm that installing a current/valid "leap 
seconds" file should block bogus leap seconds, perhaps that could be a 
recommended workaround to the bogus leap-seconds issue, until the actual 
issue can be patched.  Could you comment?
> H

Thanks,
--Jeff


More information about the questions mailing list