Re: [RFC PATCH 16/22 -v2] add get_monotonic_cycles

[Mathieu Desnoyers <compudj@krystal.dyndns.org>]

* Mathieu Desnoyers (mathieu.desnoyers@polymtl.ca) wrote:
> * john stultz (johnstul@us.ibm.com) wrote:
> > 
> > On Wed, 2008-01-16 at 18:33 -0500, Steven Rostedt wrote:
> > > Thanks John for doing this!
> > > 
> > > (comments imbedded)
> > > 
> > > On Wed, 16 Jan 2008, john stultz wrote:
> > > > +	int num = !cs->base_num;
> > > > +	cycle_t offset = (now - cs->base[!num].cycle_base_last);
> > > > +	offset &= cs->mask;
> > > > +	cs->base[num].cycle_base = cs->base[!num].cycle_base + offset;
> > > > +	cs->base[num].cycle_base_last = now;
> > > 
> > > I would think that we would need some sort of barrier here. Otherwise,
> > > base_num could be updated before all the cycle_base. I'd expect a smp_wmb
> > > is needed.
> > 
> > Hopefully addressed in the current version.
> > 
> > 
> > > > Index: monotonic-cleanup/kernel/time/timekeeping.c
> > > > ===================================================================
> > > > --- monotonic-cleanup.orig/kernel/time/timekeeping.c	2008-01-16 12:21:46.000000000 -0800
> > > > +++ monotonic-cleanup/kernel/time/timekeeping.c	2008-01-16 14:15:31.000000000 -0800
> > > > @@ -71,10 +71,12 @@
> > > >   */
> > > >  static inline s64 __get_nsec_offset(void)
> > > >  {
> > > > -	cycle_t cycle_delta;
> > > > +	cycle_t now, cycle_delta;
> > > >  	s64 ns_offset;
> > > >
> > > > -	cycle_delta = clocksource_get_cycles(clock, clocksource_read(clock));
> > > > +	now = clocksource_read(clock);
> > > > +	cycle_delta = (now - clock->cycle_last) & clock->mask;
> > > > +	cycle_delta += clock->cycle_accumulated;
> > > 
> > > Is the above just to decouple the two methods?
> > 
> > Yep. clocksource_get_cycles() ended up not being as useful as an helper
> > function (I was hoping the arch vsyscall implementations could use it,
> > but they've done too much optimization - although that may reflect a
> > need up the chain to the clocksource structure).
> > 
> 
> The problem with vsyscall is that we will have a hard time disabling
> preemption :( Therefore, insuring that the read of the data is done in a
> timely manner is hard to do.
> 

Sorry for self-reply, but I thought, in the past, of a way to make this
possible.

It would imply the creation of a new vsyscall : vgetschedperiod

It would read a counter that would increment each time the thread is
scheduled out (or in). It would be a per thread counter (not a per cpu
counter) so we can deal appropriately with a stopped thread that would
happen to come back running a loooong time afterward (if we do per-cpu
counters, we could get the same 32 bits counter value falsely if it is
shared with other thread activity).

Then, the clocksource read code would look like :

int period;

do {
  period = vgetschedperiod();

  perform the clocksource read..

} while (period != vgetschedperiod());

Therefore, we would be sure that we have not been scheduled out while
reading the value. I think this new vsyscall could be useful for others.
Actually, it would make implementation of RCU in user-space possible (as
long as the read-side can retry the read operation).

Mathieu

> 
> > thanks
> > -john
> > 
> > 
> 
> -- 
> Mathieu Desnoyers
> Computer Engineering Ph.D. Student, Ecole Polytechnique de Montreal
> OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68

-- 
Mathieu Desnoyers
Computer Engineering Ph.D. Student, Ecole Polytechnique de Montreal
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68