Re: [PATCH 0/2] itimers: periodic timers fixes

[Ingo Molnar <mingo@elte.hu>]

* Stanislaw Gruszka <sgruszka@redhat.com> wrote:

> Hi Ingo.
> 
> On Thu, 2 Apr 2009 18:57:53 +0200
> Ingo Molnar <mingo@elte.hu> wrote:
> > > We found the periodic timers ITIMER_PROF and ITIMER_VIRT are 
> > > unreliable, they have systematic timing error. For example period 
> > > of 10000 us will not be represented by the kernel as 10 ticks, but 
> > > 11 (for HZ=1000). The reason is that the frequency of the hardware 
> > > timer can only be chosen in discrete steps and the actual 
> > > frequency is about 1000.152 Hz. So 10 ticks would take only about 
> > > 9.9985 ms, the kernel decides it must never return earlier than 
> > > requested, so it rounds the period up to 11 ticks. This results in 
> > > a systematic multiplicative timing error of -10 %. The situation 
> > > is even worse where application try to request with 1 thick 
> > > period. It will get the signal once per two kernel ticks, not on 
> > > every tick. The systematic multiplicative timing error is -50 %. 
> > > He have program [1] that shows itimers systematic error, results 
> > > are below [2].
> > > 
> > > To fix situation we wrote two patches. First one just simplify 
> > > code related with itimers. Second is fix, it change intervals 
> > > measurement resolutions and correct times when signal is 
> > > generated. However this add some drawback, that I'm not sure if 
> > > are acceptable:
> > > 
> > > - the time between two consecutive tics can be smaller than 
> > >   requested interval
> > > 
> > > - intervals values which are returned to user by getitimer() are 
> > >   not rounded up
> > > 
> > > Second drawback mean that applications which first call 
> > > setitimer() then call getitimer() to see if interval was round up 
> > > and to correct timings, will potentially stop works. However this 
> > > can be only problem with requested interval smaller than 1/HZ, as 
> > > for intervals > 1/Hz we can generate signals with proper 
> > > resolution.
> > 
> > Converting those to GTOD sampling instead of jiffies sampling is a 
> > worthwile change IMO and a good concept.
> > 
> > The unificaton of ITIMER_PROF and ITIMER_VIRT is a nice observation 
> > and a good patch.
> > 
> > The second one, changing all the sampling from cputime to ktime_t is 
> > nicely done too:
> > 
> > We could do more though, there's still a bit of cputime legacies 
> > around:
> > 
> > +	cputime_t cval, nval;
> > 
> > Couldnt all of that go over into the ktime_t space as well, phasing 
> > out cputime logic from the itimer code?
> > 
> > The user ABI is struct timeval based, so there's no need to have 
> > cputime anywhere. The scheduler does nanoseconds accurate stats so 
> > it can be connected up there too.
> 
> Could the patches be merged and possible other work done in later 
> time? Or perhaps I should rework on them?

It's up to Thomas - but they certainly looked good to me.

	Ingo