Re: [PATCH v3 3/4] ftrace: add tracepoint for hrtimer

[Peter Zijlstra <peterz@infradead.org>]

On Fri, 2009-07-24 at 17:40 +0800, Xiao Guangrong wrote:
> 
> Peter Zijlstra wrote:
> > On Thu, 2009-07-23 at 18:01 +0800, Xiao Guangrong wrote:
> >> Peter Zijlstra wrote:
> >>
> >>> OK, so what you want to measure is the time of the actual callback
> >>> happening (hrtimer_entry) vs that where you would have expected it to
> >>> happen (hrtimer_start + delay), right?
> >>>
> >> Yes
> >>
> >>> So what's wrong with printing the expected expiration time in the
> >>> hrtimer_start tracepoint in the cheap clock units?
> >>>
> >> Is "cheap clock units" means jiffies time?
> > 
> > Nah, something like cpu_clock() which is monotonic per-cpu and should
> > have high resolution where available.
> > 
> 
> Hi Peter,
> 
> Thanks for your valuable suggestion.
> 
> I have do some test for it, and it very suitable for CLOCK_MONOTONIC clock,
> but if the hrtimer's clock is CLOCK_REALTIME, however xtime is needed to get
> the real time.

If you also log xtime shifts in cpu_clock() units you're good again :-)

You could of course also choose to ignore xtime shifts (they're rarely
significant on the scale of timer intervals).