[ntp:questions] Ntpd in uninterruptible sleep?

A C agcarver+ntp at acarver.net
Thu Nov 10 07:39:29 UTC 2011


On 11/9/2011 20:32, Dave Hart wrote:
> On Thu, Nov 10, 2011 at 04:21, A C<agcarver+ntp at acarver.net>  wrote:
>> As expected, it crashed again.  The interesting thing is that the constant
>> output from -D2 stopped in its tracks.  Nothing in the output shows anything
>> out of the ordinary.  It just halts completely.
>
> That's progress, and tells me a lot.  We need to see where ntpd is
> stuck, I think that means either attaching gdb to the hung ntpd, or
> starting it again under gdb and breaking into gdb once it hangs.  I'm
> relatively weak on gdb, but I believe if it's started under gdb, ^C
> will break into a gdb>  prompt.
>
> Once in, use "bt" to get a stack trace that hopefully goes back to
> ntpdmain (that is, is hopefully on the correct ntpd thread, and not
> any secondary or gdb-created thread).
>

I'll restart it under gdb and see what happens when it hangs again.

I did have a question about the debug data out of -D2.  What does the 
"head" parameter indicate in a poll_update?  I notice that most of the 
time head is zero but once in a while it's something else.  The log of 
the debug data shows it's gone over 500 a couple times.

This one is zero, I lost the bigger one which
> poll_update: at 3966 127.127.22.1 poll 4 burst 0 retry 0 head 0 early 2 next 16


More information about the questions mailing list