Re: [PATCH v2 1/2] rcu: Enable rcu_normal_wake_from_gp on small systems
On Wed, Jun 11, 2025 at 05:48:57PM -0400, Joel Fernandes wrote:
>
>
> On 6/11/2025 1:56 PM, Uladzislau Rezki (Sony) wrote:
> > Automatically enable the rcu_normal_wake_from_gp parameter on
> > systems with a small number of CPUs. The activation threshold
> > is set to 16 CPUs.
> >
> > This helps to reduce a latency of normal synchronize_rcu() API
> > by waking up GP-waiters earlier and decoupling synchronize_rcu()
> > callers from regular callback handling.
> >
> > A benchmark running 64 parallel jobs(system with 64 CPUs) invoking
> > synchronize_rcu() demonstrates a notable latency reduction with the
> > setting enabled.
>
> Reviewed-by: Joel Fernandes
>
Thank you for review!
> Just one nit below:
>
> >
> > Latency distribution (microseconds):
> >
> >
> > 0 - : 1
> > 1 - 1 : 4
> > 2 - 2 : 399
> > 3 - 3 : 3197
> > 4 - 4 : 10428
> > 5 - 5 : 17363
> > 6 - 6 : 15529
> > 7 - 7 : 9287
> > 8 - 8 : 4249
> > 9 - 9 : 1915
> > 10 - 10 : 922
> > 11 - 11 : 390
> > 12 - 12 : 187
> > ...
> >
> >
> >
> > 0 - : 1
> > 1 - 1 : 234
> > 2 - 2 : 6678
> > 3 - 3 : 33463
> > 4 - 4 : 20669
> > 5 - 5 : 2766
> > 6 - 6 : 183
> > ...
> >
> >
> > Signed-off-by: Uladzislau Rezki (Sony)
> > ---
> > kernel/rcu/tree.c | 14 +++---
> > 1 file changed, 11 insertions(+), 3 deletions(-)
> >
> > diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> > index e8a4b720d7d2..b88ceb35cebd 100644
> > --- a/kernel/rcu/tree.c
> > +++ b/kernel/rcu/tree.c
> > @@ -1625,8 +1625,10 @@ static void rcu_sr_put_wait_head(struct llist_node
> > *node)
> > atomic_set_release(&sr_wn->inuse, 0);
> > }
> >
> > -/* Disabled by default. */
> > -static int rcu_normal_wake_from_gp;
> > +/* Enable rcu_normal_wake_from_gp automatically on small systems. */
> > +#define WAKE_FROM_GP_CPU_THRESHOLD 16
> > +
> > +static int rcu_normal_wake_from_gp = -1;
> > module_param(rcu_normal_wake_from_gp, int, 0644);
> > static struct workqueue_struct *sync_wq;
> >
> > @@ -3239,7 +3241,7 @@ static void synchronize_rcu_normal(void)
> >
> > trace_rcu_sr_normal(rcu_state.name, &rs.head, TPS("request"));
> >
> > - if (!READ_ONCE(rcu_normal_wake_from_gp)) {
> > + if (READ_ONCE(rcu_normal_wake_from_gp) < 1) {
> > wait_rcu_gp(call_rcu_hurry);
> > goto trace_complete_out;
> > }
> > @@ -4843,6 +4845,12 @@ void __init rcu_init(void)
> > sync_wq = alloc_workqueue("sync_wq", WQ_MEM_RECLAIM, 0);
> > WARN_ON(!sync_wq);
> >
> > + /* Respect if explicitly disabled via a boot parameter. */
> > + if (rcu_normal_wake_from_gp < 0) {
> > + if (num_possible_cpus() <= WAKE_FROM_GP_CPU_THRESHOLD)
> > + rcu_normal_wake_from_gp = 1;
> > + }
>
> nit: better to just do:
>
> if (rcu_normal_wake_from_gp < 0 &&
> num_possible_cpus() <= WAKE_FROM_GP_CPU_THRESHOLD)
> rcu_normal_wake_from_gp = 1;
> }
>
Two lines anyway :)
--
Uladzislau Rezki
Re: [PATCH v2 1/2] rcu: Enable rcu_normal_wake_from_gp on small systems
On 6/11/2025 1:56 PM, Uladzislau Rezki (Sony) wrote:
> Automatically enable the rcu_normal_wake_from_gp parameter on
> systems with a small number of CPUs. The activation threshold
> is set to 16 CPUs.
>
> This helps to reduce a latency of normal synchronize_rcu() API
> by waking up GP-waiters earlier and decoupling synchronize_rcu()
> callers from regular callback handling.
>
> A benchmark running 64 parallel jobs(system with 64 CPUs) invoking
> synchronize_rcu() demonstrates a notable latency reduction with the
> setting enabled.
Reviewed-by: Joel Fernandes
Just one nit below:
>
> Latency distribution (microseconds):
>
>
> 0 - : 1
> 1 - 1 : 4
> 2 - 2 : 399
> 3 - 3 : 3197
> 4 - 4 : 10428
> 5 - 5 : 17363
> 6 - 6 : 15529
> 7 - 7 : 9287
> 8 - 8 : 4249
> 9 - 9 : 1915
> 10 - 10 : 922
> 11 - 11 : 390
> 12 - 12 : 187
> ...
>
>
>
> 0 - : 1
> 1 - 1 : 234
> 2 - 2 : 6678
> 3 - 3 : 33463
> 4 - 4 : 20669
> 5 - 5 : 2766
> 6 - 6 : 183
> ...
>
>
> Signed-off-by: Uladzislau Rezki (Sony)
> ---
> kernel/rcu/tree.c | 14 +++---
> 1 file changed, 11 insertions(+), 3 deletions(-)
>
> diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> index e8a4b720d7d2..b88ceb35cebd 100644
> --- a/kernel/rcu/tree.c
> +++ b/kernel/rcu/tree.c
> @@ -1625,8 +1625,10 @@ static void rcu_sr_put_wait_head(struct llist_node
> *node)
> atomic_set_release(&sr_wn->inuse, 0);
> }
>
> -/* Disabled by default. */
> -static int rcu_normal_wake_from_gp;
> +/* Enable rcu_normal_wake_from_gp automatically on small systems. */
> +#define WAKE_FROM_GP_CPU_THRESHOLD 16
> +
> +static int rcu_normal_wake_from_gp = -1;
> module_param(rcu_normal_wake_from_gp, int, 0644);
> static struct workqueue_struct *sync_wq;
>
> @@ -3239,7 +3241,7 @@ static void synchronize_rcu_normal(void)
>
> trace_rcu_sr_normal(rcu_state.name, &rs.head, TPS("request"));
>
> - if (!READ_ONCE(rcu_normal_wake_from_gp)) {
> + if (READ_ONCE(rcu_normal_wake_from_gp) < 1) {
> wait_rcu_gp(call_rcu_hurry);
> goto trace_complete_out;
> }
> @@ -4843,6 +4845,12 @@ void __init rcu_init(void)
> sync_wq = alloc_workqueue("sync_wq", WQ_MEM_RECLAIM, 0);
> WARN_ON(!sync_wq);
>
> + /* Respect if explicitly disabled via a boot parameter. */
> + if (rcu_normal_wake_from_gp < 0) {
> + if (num_possible_cpus() <= WAKE_FROM_GP_CPU_THRESHOLD)
> + rcu_normal_wake_from_gp = 1;
> + }
nit: better to just do:
if (rcu_normal_wake_from_gp < 0 &&
num_possible_cpus() <= WAKE_FROM_GP_CPU_THRESHOLD)
rcu_normal_wake_from_gp = 1;
}
> +
> /* Fill in default value for rcutree.qovld boot parameter. */
> /* -After- the rcu_node ->lock fields are initialized! */
> if (qovld < 0)
