On Wed, 15 Jun 2016, Paul E. McKenney wrote:
> On Mon, Jun 13, 2016 at 09:15:14AM -0700, Paul E. McKenney wrote:
> > On Mon, Jun 13, 2016 at 08:40:50AM -0000, Thomas Gleixner wrote:
> > > The current timer wheel has some drawbacks:
> > >
> > > 1) Cascading
> > >
> > > Cascading can be an unbound operation and is completely pointless in
> > > most
> > > cases because the vast majority of the timer wheel timers are canceled
> > > or
> > > rearmed before expiration.
> > >
> > > 2) No fast lookup of the next expiring timer
> > >
> > > In NOHZ scenarios the first timer soft interrupt after a long NOHZ
> > > period
> > > must fast forward the base time to current jiffies. As we have no way
> > > to
> > > find the next expiring timer fast, the code loops and increments the
> > > base
> > > time by one and checks for expired timers in each step. I've observed
> > > loops
> > > lasting 1 ms!
> > >
> > > There are some other issues caused by the above, but they are minor
> > > compare to
> > > those.
> >
> > For SMP configurations, this passes light rcutorture testing. For UP
> > builds, it complains about undefined symbols. Builds succeed with
> > the following kneejerk patch. Am retesting rcutorture.
>
> And with the patch below, testing goes as well with your patch stack as
> it does without it. So, with that patch (or equivalent):
>
> Tested-by: Paul E. McKenney <paul...@linux.vnet.ibm.com>
>
> There were some complaints about increasing the size of the tiny
> configuration, FYI.
I know. The extra storage space for the deferrable stuff makes it larger along
with the extra code for avoiding all the crap which the current wheel suffers
from :) Do the tiny people need NOHZ?
> So, just out of curiosity, does anyone still run -rt on single-CPU systems?
Of course :)
Thanks,
tglx
