[ntp:hackers] Bug hunting opportunities

Todd Glassey tglassey at earthlink.net
Thu Oct 29 23:30:48 UTC 2009


Greg Dowd wrote:
> How would this work?  I expect the pll runs closed loop and not setting
> the clock would just send it to the rails and introduce all kinds of
> wacky behavior in the daemon.  
>
> -----Original Message-----
> From: hackers-bounces+gdowd=symmetricom.com at lists.ntp.org
> [mailto:hackers-bounces+gdowd=symmetricom.com at lists.ntp.org] On Behalf
> Of Hal Murray
> Sent: Thursday, October 29, 2009 11:36 AM
> To: Harlan Stenn
> Cc: hackers at ntp.isc.org
> Subject: Re: [ntp:hackers] Bug hunting opportunities
>
>
>   
>> I would be Thrilled if somebody would get a test suite going. 
>>     
>
> Here is a possibly crazy thought...
>
> Suppose we added a user mode switch to the command line.
>
> The idea is to allow a as much testing as we can while the system is
> still 
> running a production version of ntpd.  The ntpd under test would run,
> but not 
> do anything to the system. 
OK. That would also allow a query without time-setting or also the use 
of time-query from the local context which would be cool for cloud 
(cluster based) systems.
>  It could write log files and stuff in the
> user's 
> directory.
>   
You mean run a user-specific instance of NTP that the user process owns 
rather than being operated as a shared resource? Would that also allow 
the user to specify the port number as well so it could be run with a 
formal copy of NTP running on the same system or virtual system instance?
> I think it's only a few lines of code in several places.
>   
Yep
>   One is listening on a non-standard port number.
>   Another is not actually setting the time.
>   
That is a "run in query mode only"
>   We should probably not clutter up syslog.
>   
No actually you should allow for logging aggregation there as well and 
support either specifying a log per user-NTP instance or allowing 
standard logging elsewise.
>   Bypass the check for being root.
>   Changing the default config file from /etc/ntp.conf to ./ntp.conf
> would 
> reduce errors
>   
Config file should be specifiable for each virtual instance. The NTP 
service would sort of be a framework for NTP daemons - either shared or 
private then.
> Are there any others?
>   
Connection rules on a per instance basis but that's probably a given
> It might be handy if it wrote something to a log file any time it would
> have 
> called the system to adjust the time.  That log file might be handy for 
> non-user debugging.
>
> You would have to setup a special debugging ntp.conf.  It would put the
> log 
> files and drift file someplace where you could write.  
With the virtual instance files or the logging file if one is created.
> With a bit of
> work, 
> this could even be merged into something like flock-build.
>
> I think you could even test refclocks this way if you had one that
> wasn't in 
> use by the real system.
>
>
>
>   
> ------------------------------------------------------------------------
>
>
> No virus found in this incoming message.
> Checked by AVG - www.avg.com 
> Version: 8.5.423 / Virus Database: 270.14.38/2467 - Release Date: 10/29/09 07:38:00
>
>   



More information about the hackers mailing list