Re: [PATCH 1/4] nohz: Split extended quiescent state handling from nohz switch

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

On Sat, Aug 20, 2011 at 07:30:49PM +0200, Frederic Weisbecker wrote:
> It is assumed that rcu won't be used once we switch to tickless
> mode and until we restart the tick. However this is not always
> true, as in x86-64 where we dereference the idle notifiers after
> the tick is stopped.
> 
> To prepare for fixing this, split the tickless mode switching and
> RCU extended quiescent state logics.
> Make tick_nohz_stop/restart_sched_tick() RCU agnostic but provide
> a new pair of APIs tick_nohz_enter/exit_idle() that keep the
> old behaviour by handling both the nohz mode and RCU extended
> quiescent states, then convert every archs to use these.
> 
> Archs that want to switch to RCU extended QS to some custom points
> can do it later by changing the parameter in tick_nohz_enter,exit_idle()
> to false and call rcu_enter,exit() separately.

This approach looks quite good to me!  A few comments below.

> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
> Cc: David Miller <davem@davemloft.net>
> Cc: Chris Metcalf <cmetcalf@tilera.com>
> Cc: Guan Xuetao <gxt@mprc.pku.edu.cn>
> Cc: Hans-Christian Egtvedt <hans-christian.egtvedt@atmel.com>
> Cc: Mike Frysinger <vapier@gentoo.org>
> Cc: Ralf Baechle <ralf@linux-mips.org>
> Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> Cc: Ingo Molnar <mingo@elte.hu>
> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: H. Peter Anvin <hpa@zytor.com>
> Cc: Russell King <linux@arm.linux.org.uk>
> Cc: Paul Mackerras <paulus@samba.org>
> Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
> Cc: Paul Mundt <lethal@linux-sh.org>
> ---

[ . . . ]

> diff --git a/arch/powerpc/kernel/idle.c b/arch/powerpc/kernel/idle.c
> index 39a2baa..b30ddf1 100644
> --- a/arch/powerpc/kernel/idle.c
> +++ b/arch/powerpc/kernel/idle.c
> @@ -56,7 +56,7 @@ void cpu_idle(void)
> 
>  	set_thread_flag(TIF_POLLING_NRFLAG);
>  	while (1) {
> -		tick_nohz_stop_sched_tick(1);
> +		tick_nohz_enter_idle(true);

This needs to pass false, as some variants of powerpc use event tracing
(which in turn uses RCU) just before returning to the hypervisor.

>  		while (!need_resched() && !cpu_should_die()) {
>  			ppc64_runlatch_off();
> 
> @@ -93,7 +93,7 @@ void cpu_idle(void)
> 
>  		HMT_medium();
>  		ppc64_runlatch_on();
> -		tick_nohz_restart_sched_tick();
> +		tick_nohz_exit_idle(true);

As does this, as some variants of powerpc use event tracing just after
the hypervisor returns control to the OS.

>  		preempt_enable_no_resched();
>  		if (cpu_should_die())
>  			cpu_die();
> diff --git a/arch/powerpc/platforms/iseries/setup.c b/arch/powerpc/platforms/iseries/setup.c
> index c25a081..aa2be7d 100644
> --- a/arch/powerpc/platforms/iseries/setup.c
> +++ b/arch/powerpc/platforms/iseries/setup.c
> @@ -562,7 +562,7 @@ static void yield_shared_processor(void)
>  static void iseries_shared_idle(void)
>  {
>  	while (1) {
> -		tick_nohz_stop_sched_tick(1);
> +		tick_nohz_enter_idle(true);

But I don't know enough about iseries to know whether or not this works.

>  		while (!need_resched() && !hvlpevent_is_pending()) {
>  			local_irq_disable();
>  			ppc64_runlatch_off();
> @@ -576,7 +576,7 @@ static void iseries_shared_idle(void)
>  		}
> 
>  		ppc64_runlatch_on();
> -		tick_nohz_restart_sched_tick();
> +		tick_nohz_exit_idle(true);
> 
>  		if (hvlpevent_is_pending())
>  			process_iSeries_events();
> @@ -592,7 +592,7 @@ static void iseries_dedicated_idle(void)
>  	set_thread_flag(TIF_POLLING_NRFLAG);
> 
>  	while (1) {
> -		tick_nohz_stop_sched_tick(1);
> +		tick_nohz_enter_idle(true);
>  		if (!need_resched()) {
>  			while (!need_resched()) {
>  				ppc64_runlatch_off();
> @@ -609,7 +609,7 @@ static void iseries_dedicated_idle(void)
>  		}
> 
>  		ppc64_runlatch_on();
> -		tick_nohz_restart_sched_tick();
> +		tick_nohz_exit_idle(true);
>  		preempt_enable_no_resched();
>  		schedule();
>  		preempt_disable();

[ . . . ]

> diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
> index eb98e55..609bb20 100644
> --- a/kernel/time/tick-sched.c
> +++ b/kernel/time/tick-sched.c
> @@ -253,12 +253,13 @@ EXPORT_SYMBOL_GPL(get_cpu_iowait_time_us);
>   * Called either from the idle loop or from irq_exit() when an idle period was
>   * just interrupted by an interrupt which did not cause a reschedule.
>   */
> -void tick_nohz_stop_sched_tick(int inidle)
> +bool tick_nohz_stop_sched_tick(int inidle)
>  {
>  	unsigned long seq, last_jiffies, next_jiffies, delta_jiffies, flags;
>  	struct tick_sched *ts;
>  	ktime_t last_update, expires, now;
>  	struct clock_event_device *dev = __get_cpu_var(tick_cpu_device).evtdev;
> +	int stopped = false;
>  	u64 time_delta;
>  	int cpu;
> 
> @@ -405,7 +406,7 @@ void tick_nohz_stop_sched_tick(int inidle)
>  			ts->idle_tick = hrtimer_get_expires(&ts->sched_timer);
>  			ts->tick_stopped = 1;
>  			ts->idle_jiffies = last_jiffies;
> -			rcu_enter_nohz();
> +			stopped = true;
>  		}
> 
>  		ts->idle_sleeps++;
> @@ -445,6 +446,24 @@ out:
>  	ts->sleep_length = ktime_sub(dev->next_event, now);
>  end:
>  	local_irq_restore(flags);
> +
> +	return stopped;
> +}
> +
> +
> +/**
> + * tick_nohz_enter_idle - stop the tick from the idle task
> + * @rcu_ext_qs: enter rcu dynticks idle mode
> + *
> + * When an arch doesn't make any use of rcu read side critical section
> + * between tick_nohz_enter_idle() and tick_nohz_exit_idle() it can set
> + * rcu_ext_qs to 1. Otherwise it needs to call rcu_enter_nohz() itself
> + * later before the CPU goes to sleep.

How about something like this?

+ * When an arch doesn't make any use of rcu read side critical section
+ * between tick_nohz_enter_idle() and the time the CPU is put to sleep,
+ * it can set rcu_ext_qs to true. Otherwise it needs set rcu_ext_qs to
+ * false, and then to call rcu_enter_nohz() itself later, but before the
+ * CPU goes to sleep and after its last use of RCU.

> + */
> +void tick_nohz_enter_idle(bool rcu_ext_qs)
> +{
> +	if (tick_nohz_stop_sched_tick(1) && rcu_ext_qs)
> +		rcu_enter_nohz();
>  }
> 
>  /**
> @@ -486,11 +505,12 @@ static void tick_nohz_restart(struct tick_sched *ts, ktime_t now)
>  }
> 
>  /**
> - * tick_nohz_restart_sched_tick - restart the idle tick from the idle task
> + * tick_nohz_exit_idle - restart the idle tick from the idle task
> + * @rcu_ext_qs: exit rcu dynticks idle mode
>   *
>   * Restart the idle tick when the CPU is woken up from idle

How about something like the following?

+ * When an arch doesn't make any use of rcu read side critical section
+ * between the time the CPU wakes up and tick_nohz_exit_idle(), it can set
+ * rcu_ext_qs to true. Otherwise it needs set rcu_ext_qs to false, and it
+ * must also have called rcu_enter_nohz() itself earlier, after the CPU
+ * was awakened, but before its first use of RCU.

>   */
> -void tick_nohz_restart_sched_tick(void)
> +void tick_nohz_exit_idle(bool rcu_ext_qs)
>  {
>  	int cpu = smp_processor_id();
>  	struct tick_sched *ts = &per_cpu(tick_cpu_sched, cpu);
> @@ -514,7 +534,8 @@ void tick_nohz_restart_sched_tick(void)
> 
>  	ts->inidle = 0;
> 
> -	rcu_exit_nohz();
> +	if (rcu_ext_qs)
> +		rcu_exit_nohz();
> 
>  	/* Update jiffies first */
>  	select_nohz_load_balancer(0);
> -- 
> 1.7.5.4
>