Re: [PATCH tip/core/rcu 01/26] rcu: New rcu_user_enter() and rcu_user_exit() APIs

[Josh Triplett <josh@joshtriplett.org>]

On Thu, Aug 30, 2012 at 02:05:18PM -0700, Paul E. McKenney wrote:
> From: Frederic Weisbecker <fweisbec@gmail.com>
> 
> RCU currently insists that only idle tasks can enter RCU idle mode, which
> prohibits an adaptive tickless kernel (AKA nohz cpusets), which in turn
> would mean that usermode execution would always take scheduling-clock
> interrupts, even when there is only one task runnable on the CPU in
> question.
> 
> This commit therefore adds rcu_user_enter() and rcu_user_exit(), which
> allow non-idle tasks to enter RCU idle mode.  These are quite similar
> to rcu_idle_enter() and rcu_idle_exit(), respectively, except that they
> omit the idle-task checks.
> 
> [ Updated to use "user" flag rather than separate check functions. ]
> 
> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
> Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> Cc: Alessio Igor Bogani <abogani@kernel.org>
> Cc: Andrew Morton <akpm@linux-foundation.org>
> Cc: Avi Kivity <avi@redhat.com>
> Cc: Chris Metcalf <cmetcalf@tilera.com>
> Cc: Christoph Lameter <cl@linux.com>
> Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
> Cc: Geoff Levand <geoff@infradead.org>
> Cc: Gilad Ben Yossef <gilad@benyossef.com>
> Cc: Hakan Akkan <hakanakkan@gmail.com>
> Cc: Ingo Molnar <mingo@kernel.org>
> Cc: Kevin Hilman <khilman@ti.com>
> Cc: Max Krasnyansky <maxk@qualcomm.com>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Stephen Hemminger <shemminger@vyatta.com>
> Cc: Steven Rostedt <rostedt@goodmis.org>
> Cc: Sven-Thorsten Dietrich <thebigcorporation@gmail.com>
> Cc: Thomas Gleixner <tglx@linutronix.de>

A few suggestions below: an optional microoptimization and some bugfixes.
With the bugfixes, and with or without the microoptimization:

Reviewed-by: Josh Triplett <josh@joshtriplett.org>

> --- a/kernel/rcutree.c
> +++ b/kernel/rcutree.c
[...]
> -static void rcu_idle_enter_common(struct rcu_dynticks *rdtp, long long oldval)
> +static void rcu_eqs_enter_common(struct rcu_dynticks *rdtp, long long oldval,
> +				bool user)
>  {
>  	trace_rcu_dyntick("Start", oldval, 0);
> -	if (!is_idle_task(current)) {
> +	if (!is_idle_task(current) && !user) {

Microoptimization: putting the !user check first (here and in the exit
function) would allow the compiler to partially inline rcu_eqs_*_common
into the two trivial wrappers and constant-fold away the test for !user.

> +void rcu_idle_enter(void)
> +{
> +	rcu_eqs_enter(0);
> +}

s/0/false/

> +void rcu_user_enter(void)
> +{
> +	rcu_eqs_enter(1);
> +}

s/1/true/

> -static void rcu_idle_exit_common(struct rcu_dynticks *rdtp, long long oldval)
> +static void rcu_eqs_exit_common(struct rcu_dynticks *rdtp, long long oldval,
> +			       int user)
>  {
>  	smp_mb__before_atomic_inc();  /* Force ordering w/previous sojourn. */
>  	atomic_inc(&rdtp->dynticks);
> @@ -464,7 +490,7 @@ static void rcu_idle_exit_common(struct rcu_dynticks *rdtp, long long oldval)
>  	WARN_ON_ONCE(!(atomic_read(&rdtp->dynticks) & 0x1));
>  	rcu_cleanup_after_idle(smp_processor_id());
>  	trace_rcu_dyntick("End", oldval, rdtp->dynticks_nesting);
> -	if (!is_idle_task(current)) {
> +	if (!is_idle_task(current) && !user) {

Same micro-optimization as the enter function.

> +void rcu_idle_exit(void)
> +{
> +	rcu_eqs_exit(0);
> +}

s/0/false/

> +void rcu_user_exit(void)
> +{
> +	rcu_eqs_exit(1);
> +}

s/1/true/

> @@ -539,7 +586,7 @@ void rcu_irq_enter(void)
>  	if (oldval)
>  		trace_rcu_dyntick("++=", oldval, rdtp->dynticks_nesting);
>  	else
> -		rcu_idle_exit_common(rdtp, oldval);
> +		rcu_eqs_exit_common(rdtp, oldval, 1);

s/1/true/, and likewise in rcu_irq_exit.

- Josh Triplett