Re: [GIT PULL rcu/next] RCU commits for 3.3

["Paul E. McKenney" <paulmck@linux.vnet.ibm.com>]

On Wed, Dec 14, 2011 at 06:47:38PM +0100, Frederic Weisbecker wrote:
> On Wed, Dec 14, 2011 at 05:30:11PM +0100, Frederic Weisbecker wrote:
> > On Wed, Dec 14, 2011 at 04:47:36PM +0100, Ingo Molnar wrote:
> > > * Paul E. McKenney <paulmck@linux.vnet.ibm.com> wrote:

[ . . . ]

> > > Pulled into tip:core/rcu, thanks a lot Paul!
> > > 
> > > Note that this commit from Frederic:
> > > 
> > >   69e1e811dcc4: sched, nohz: Track nr_busy_cpus in the sched_group_power
> > > 
> > > conflicted with this commit from Suresh in sched/core:
> > > 
> > >   69e1e811dcc4: sched, nohz: Track nr_busy_cpus in the sched_group_power
> > > 
> > > I resolved it by making the set_cpu_sd_state_idle() call 
> > > unconditional within the newly decoupled 
> > > tick_nohz_stop_sched_tick() function - please double check that 
> > > it's the right resolution.
> > 
> > After a quick look, I believe this should rather be under tick_nohz_idle_enter(),
> > (This is the equivalent of the old tick_nohz_stop_sched_tick(1))
> > This wants to be set only once we enter idle, not everytime we have an idle
> > interrupt.
> 
> I don't know how you plan to fix the conflict, by redoing the merge or by
> applying a patch on tip/master.
> 
> In any case, here is a patch you can use. Feel free to apply it as is
> or to just refer to its diff to redo the merge:
> 
> (Outrageously only compile tested)

I have started up a short rcutorture test of this patch on top of
tip/master, will let you know how it goes.

							Thanx, Paul

> ---
> From: Frederic Weisbecker <fweisbec@gmail.com>
> Date: Wed, 14 Dec 2011 18:36:00 +0100
> Subject: [PATCH] sched: Only update the CPU idleness in the domain
>  hierarchy from idle loop entry
> 
> We don't need to inform the sched domain hierarchy about the
> CPU idleness everytime we call tick_nohz_stop_sched_tick() as this
> includes both idle loop entry and idle interrupt exit.
> 
> Doing it once from the idle loop entry is enough, call
> set_cpu_sd_state_idle() only from tick_nohz_idle_enter() instead
> to fix this.
> 
> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> Cc: Suresh Siddha <suresh.b.siddha@intel.com>
> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
> ---
>  kernel/time/tick-sched.c |   16 ++++++++--------
>  1 files changed, 8 insertions(+), 8 deletions(-)
> 
> diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
> index 1f6dc515..696c997 100644
> --- a/kernel/time/tick-sched.c
> +++ b/kernel/time/tick-sched.c
> @@ -289,14 +289,6 @@ static void tick_nohz_stop_sched_tick(struct tick_sched *ts)
>  	now = tick_nohz_start_idle(cpu, ts);
> 
>  	/*
> -	 * Update the idle state in the scheduler domain hierarchy
> -	 * when tick_nohz_stop_sched_tick() is called from the idle loop.
> -	 * State will be updated to busy during the first busy tick after
> -	 * exiting idle.
> -	 */
> -	set_cpu_sd_state_idle();
> -
> -	/*
>  	 * If this cpu is offline and it is the one which updates
>  	 * jiffies, then give up the assignment and let it be taken by
>  	 * the cpu which runs the tick timer next. If we don't drop
> @@ -483,6 +475,14 @@ void tick_nohz_idle_enter(void)
>  	 * update of the idle time accounting in tick_nohz_start_idle().
>  	 */
>  	ts->inidle = 1;
> +
> +	/*
> +	 * Update the idle state in the scheduler domain hierarchy
> +	 * when tick_nohz_idle_enter() is called from the idle loop.
> +	 * State will be updated to busy during the first busy tick after
> +	 * exiting idle.
> +	 */
> +	set_cpu_sd_state_idle();
>  	tick_nohz_stop_sched_tick(ts);
> 
>  	local_irq_enable();
> -- 
> 1.7.5.4
> 
>