Re: [PATCH RFC tip/core/rcu 3/3] cpuidle: Inform RCU of read-side critical sections

[Deepthi Dharwar <deepthi@linux.vnet.ibm.com>]

On 02/03/2012 06:42 AM, Paul E. McKenney wrote:

> From: "Paul E. McKenney" <paul.mckenney@linaro.org>
> 
> The cpuidle_idle_call() function is invoked in the inner idle loop,
> after the call to rcu_idle_enter() and before the call to rcu_idle_exit().
> This means that RCU is ignoring the CPU at this point.  Unfortunately,
> cpuidle_idle_call() nevertheless contains tracepoints (important ones
> used by powertop) that expect RCU to be paying attention.  The consequences
> of using RCU read-side critical sections on CPUs that RCU is ignoring
> can be severe, including random corruption of random memory.
> 
> Therefore, this commit uses the new RCU_NONIDLE() macro to let RCU
> do its job with respect to the tracepoints.
> 
> Suggested-by: Nicolas Pitre <nicolas.pitre@linaro.org>
> Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> Cc: Kevin Hilman <khilman@ti.com>
> Cc: Len Brown <len.brown@intel.com>
> Cc: Trinabh Gupta <g.trinabh@gmail.com>
> Cc: Arjan van de Ven <arjan@linux.intel.com>
> Cc: Deepthi Dharwar <deepthi@linux.vnet.ibm.com>
> ---
>  drivers/cpuidle/cpuidle.c |   12 ++++++++----
>  1 files changed, 8 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/cpuidle/cpuidle.c b/drivers/cpuidle/cpuidle.c
> index 59f4261..cd8a553 100644
> --- a/drivers/cpuidle/cpuidle.c
> +++ b/drivers/cpuidle/cpuidle.c
> @@ -94,13 +94,17 @@ int cpuidle_idle_call(void)
> 
>  	target_state = &drv->states[next_state];
> 
> -	trace_power_start(POWER_CSTATE, next_state, dev->cpu);
> -	trace_cpu_idle(next_state, dev->cpu);
> +	RCU_NONIDLE(
> +		trace_power_start(POWER_CSTATE, next_state, dev->cpu);
> +		trace_cpu_idle(next_state, dev->cpu)
> +	);
> 
>  	entered_state = target_state->enter(dev, drv, next_state);
> 
> -	trace_power_end(dev->cpu);
> -	trace_cpu_idle(PWR_EVENT_EXIT, dev->cpu);
> +	RCU_NONIDLE(
> +		trace_power_end(dev->cpu);
> +		trace_cpu_idle(PWR_EVENT_EXIT, dev->cpu);
> +	);
> 
>  	if (entered_state >= 0) {
>  		/* Update cpuidle counters */


Cleaner design esp by moving away from architecture specific
tweaking is much appreciated :)

Acked-by: Deepthi Dharwar <deepthi@linux.vnet.ibm.com>