[ntp:questions] Clock jumps when refclock used

Dave Hart davehart at gmail.com
Tue Apr 17 09:52:13 UTC 2012


On Mon, Apr 16, 2012 at 03:08, A C <agcarver+ntp at acarver.net> wrote:
> On 4/15/2012 14:16, A C wrote:
>> I have a couple more recent examples of fuzz messages (still using the
>> Internet only configuration):
>>
>> 14 Apr 08:59:56 ntpd[11057]: ts_min 1334393996.509276921 ts_prev
>> 1334393996.509171929 ts 1334393996.518274219
>> 14 Apr 08:59:56 ntpd[11057]: sys_fuzz 104992 nsec, this fuzz
>> 0.000076449, prior 0.000007028
>> 14 Apr 08:59:56 ntpd[11057]: get_systime prev result 0xd333bb0c.84e97365
>> is 0.000836508 later than 0xd333bb0c.84b2a11e

Thanks for the examples.  See comments in

http://bugs.ntp.org/2141

for some detailed comments, but basically I think these are mostly
false positives due to get_systime() being re-entered, which is unsafe
when using statics to maintain previous-call state as these checks are
doing.

Thanks also to Hal Murray and others who have provided examples of
these warnings, which should become much rarer soon.

Cheers,
Dave Hart

Cheers,
Dave Hart


More information about the questions mailing list