[ntp:questions] ntpd code

mayer at gis.net mayer at gis.net
Wed Mar 23 17:20:22 UTC 2005


----- Original Message Follows -----
> Hi, Thanks for reply,
> 
> <mayer at gis.net> wrote in message 
> news:mailman.48.1111595272.588.questions at lists.ntp.isc.org...
> > ----- Original Message Follows -----
> >> Hi,
> >>
> >> I am trying to modify some portion of NTP source code to maintain a
> >> high  resolution, precise (as far as NTP goes) logical clock
> instead >> of modifying  the system clock of the user (as it may not
> have >> necessary privileges doing  so)... You can view it as
> intercepting the >> system calls from NTP to  manipulate a software
> clock; I am not sure >> how successful it would be... The  development
> and testing environment >> is strictly Windows...
> >>
> >
> > I'm not sure I understand this at all. Why wouldn't you be running
> > this as a service with the right privileges? If you want to
> > intercept system calls you probably need privileges anyway (I
> > haven't looked at the code to verify this). ntpd IS a high
> > resolution precise logical clock
> > already. What is it that you are trying to achieve?
> >
> 
> OK, What I am trying to do is to prevent updating the SYSTEM clock,
> but  instead updating an APPLICATION level clock (nothing but a 64 bit
> variable).  My aim is to collect event data from various users using
> the application I  am developing, so I will need a way to synchronize
> the event timestamps to  reflect almost exact times the events
> occured. But users may not have the  privilege to update the system
> clock in their machines, so I am simulating  that through manipulating
> a high level clock. I am not defending this is the  best way to do,
> but I really couldnot think of any other way doing so. Any  comments
> are welcomed and highly appreciated!
> 

Why not just install ntpd as a service on each of the machines and
let it do its thing?

The problem is that once the clock is set it calls
SetSystemTimeAdjustment()to change the clock and that just makes
changes to the clock rather than setting the clock to a specific
value.

> >> I have observed two points in the source code by now...
> >>
> >> Is there a reason for the latest 4.2.0 version not using
> ClockThread >> (in  nt_clockstuff.c) for clock interpolation (high
> resolution clock >> read)? I  couldnot find a place calling
> StartClockThread()... >>
> >> Also, I believe developers/maintainers are already aware, but I am
> >> guessing  at method service_exit() in ntpd.c source file, the IF
> logic >> should read
> >>
> >> (!debug && NoWinService == FALSE)
> >>
> >> otherwise, it would result in error if we were to start the ntpd so
> as >> not  to run as a Windows service.
> >>
> >
> > You appear to be looking at a rather old version of the source code.
> > You should get the latest snapshot and use that. StartClockThread()
> > is in ntclockstuff.c and that logical statement is in ntservice.c in
> > service_exit. It's not in ntpd.c. The condition is: if (!foreground)
> > rather than the above.
> >
> 
> Oh, I actually downloaded the source code from 
> http://ntp.isc.org/bin/view/Main/SoftwareDownloads
> There says the latest (stable?) version is 4.2.0 as of 2003/10/15. I
> would  appreciate if you could give me a link to the latest code.
> However,  stability is also important in my case...
> 

http://www.eecis.udel.edu/~ntp/ntp_spool/ntp4/snapshots/ntp-dev/2005/03/

You can't expect stability here since you are altering the code 
anyway. However, I don't think that there's anything unstable about
this code especially for Windows.

Danny

> Thank you!
> 
> > Danny
> >> Regards,
> >>
> >> -Kemal.




More information about the questions mailing list