Re: [GIT PULL] dynticks updates for 3.13

[Frederic Weisbecker <fweisbec@gmail.com>]

On Sat, Nov 30, 2013 at 02:59:07PM +0100, Ingo Molnar wrote:
> 
> * Frederic Weisbecker <fweisbec@gmail.com> wrote:
> 
> > Ingo,
> > 
> > Please pull the timers/core branch that can be found at:
> > 
> > git://git.kernel.org/pub/scm/linux/kernel/git/frederic/linux-dynticks.git
> > 	timers/core
> > 
> > It passed through 0-day testing. I'm proposing it for 3.13 just in case but I'm ok
> > as well if you think it's too late and should be delayed for 3.14
> 
> Pulled into tip:timers/core, thanks Frederic!
> 
> In the future it would be nice to write a higher level description 
> that I can put into the merge commit. I wrote this:
> 
>   Pull dynticks updates from Frederic Weisbecker: smaller fixes and 
>   cleanups.

Right, I'll add more detailed summary next times.

> 
> Btw., how does the bug here:
> 
>   92e73cf831ba posix-timers: Fix full dynticks CPUs kick on timer rescheduling
> 
> manifest itself to users? Is there any real bug, or suboptimal nohz 
> behavior? (more wakeups than expected) This kind of information should 
> really be in the changelog.

Agreed, it fixes a real bug and that's missing from the changelog, I'll make it
clearer on the next take.

> 
> > 
> > Thanks,
> > 	Frederic
> > ---
> > 
> > Frederic Weisbecker (5):
> >       nohz: Convert a few places to use local per cpu accesses
> >       context_tracking: Wrap static key check into more intuitive function name
> >       context_tracking: Rename context_tracking_active() to context_tracking_cpu_is_enabled()
> >       posix-timers: Spare workqueue if there is no full dynticks CPU to kick
> >       posix-timers: Fix full dynticks CPUs kick on timer rescheduling
> > 
> > Paul Gortmaker (1):
> >       trivial: fix spelling in CONTEXT_TRACKING_FORCE help text
> > 
> > 
> >  include/linux/context_tracking.h       | 10 ++++-----
> >  include/linux/context_tracking_state.h | 11 +++++++---
> >  include/linux/tick.h                   |  8 ++++----
> >  include/linux/vtime.h                  |  4 ++--
> >  init/Kconfig                           |  2 +-
> >  kernel/context_tracking.c              |  8 ++++----
> >  kernel/posix-cpu-timers.c              | 21 +++++++++----------
> >  kernel/softirq.c                       |  4 +---
> >  kernel/time/tick-broadcast.c           |  6 +++---
> >  kernel/time/tick-internal.h            |  4 ++--
> >  kernel/time/tick-sched.c               | 37 ++++++++++++++--------------------
> >  11 files changed, 54 insertions(+), 61 deletions(-)
> 
> Btw., I noticed one other detail:
> 
> -       tick_check_oneshot_broadcast(cpu);
> -       tick_check_nohz(cpu);
> +       tick_check_oneshot_broadcast();
> +       tick_check_nohz();
> 
> it would be nice to name such functions that lose a 'cpu' argument 
> with the this_cpu pattern, i.e.:
> 
> 	- tick_check_oneshot_broadcast_this_cpu()

Hmm, that's going to be quite a long name. But I'm ok with the whole idea, so
I'll handle that too.

> 	- tick_check_nohz_this_cpu()
> 
> This makes it more obvious at a glance that they have 'current CPU' 
> assumptions in them

Ok, thanks!