Re: Problems with gettimeofday_ns()

["Julio M. Merino Vidal" <jmerino at ac dot upc dot edu>]

On Feb 21, 2008, at 15:10 , Masami Hiramatsu wrote:

> Hello,
>
> Julio M. Merino Vidal wrote:
> > But the big problem is that sometimes I receive events in the
> > incorrect order!  That is, I place a marker B after a marker A in the
> > code, and I know it is impossible for B to appear before A in a
> > trace.  However, sometimes, I get B before A in the final trace with
> > the timings (supposedly) reversed.  I also print the CPU that fired
> > the events, and both are happening from the same virtual CPU, so the
> > possibility of both CPUs using different timings is not relevant.
>
> Indeed. The systemtap's gettimeofday_*() tries to adjust base-time
> (this means, change the base-time forward and backward) periodically.
> It is for synchronizing trace logs with in-kernel logs(dmesg, etc).
> We are discussing about this on bz3916,
> http://sources.redhat.com/bugzilla/show_bug.cgi?id=3916
>
> > If I change the gettimeofday_ns call to get_cycles or to a custom
> > counter instead, things go perfectly.
>
> Sure, I think it is the best way to get the monotonic counter on
> some processors.

OK, thanks for the explanations.

Now I am finding another issue slightly related to this.  I have just  
added a new marker to the code in an interrupt context, and this  
interrupt wakes up a specific thread to handle it later on.  The  
problem is that, if the interrupt is handled on one CPU and the  
thread, once awaken, is handled by the other one, I can still get the  
timestamps (cycles) reversed.  Any easy way around this?

Thank you.