[12/16] percpu-refcount: Use call_rcu_hurry() for atomic switch

Commit Message

Paul E. McKenney Nov. 30, 2022, 6:13 p.m. UTC

From: "Joel Fernandes (Google)" <[e-mail]>

Earlier commits in this series allow battery-powered systems to build
their kernels with the default-disabled CONFIG_RCU_LAZY=y Kconfig option.
This Kconfig option causes call_rcu() to delay its callbacks in order to
batch callbacks.  This means that a given RCU grace period covers more
callbacks, thus reducing the number of grace periods, in turn reducing
the amount of energy consumed, which increases battery lifetime which
can be a very good thing.  This is not a subtle effect: In some important
use cases, the battery lifetime is increased by more than 10%.

This CONFIG_RCU_LAZY=y option is available only for CPUs that offload
callbacks, for example, CPUs mentioned in the rcu_nocbs kernel boot
parameter passed to kernels built with CONFIG_RCU_NOCB_CPU=y.

Delaying callbacks is normally not a problem because most callbacks do
nothing but free memory.  If the system is short on memory, a shrinker
will kick all currently queued lazy callbacks out of their laziness,
thus freeing their memory in short order.  Similarly, the rcu_barrier()
function, which blocks until all currently queued callbacks are invoked,
will also kick lazy callbacks, thus enabling rcu_barrier() to complete
in a timely manner.

However, there are some cases where laziness is not a good option.
For example, synchronize_rcu() invokes call_rcu(), and blocks until
the newly queued callback is invoked.  It would not be a good for
synchronize_rcu() to block for ten seconds, even on an idle system.
Therefore, synchronize_rcu() invokes call_rcu_hurry() instead of
call_rcu().  The arrival of a non-lazy call_rcu_hurry() callback on a
given CPU kicks any lazy callbacks that might be already queued on that
CPU.  After all, if there is going to be a grace period, all callbacks
might as well get full benefit from it.

Yes, this could be done the other way around by creating a
call_rcu_lazy(), but earlier experience with this approach and
feedback at the 2022 Linux Plumbers Conference shifted the approach
to call_rcu() being lazy with call_rcu_hurry() for the few places
where laziness is inappropriate.

And another call_rcu() instance that cannot be lazy is the one on the
percpu refcounter's "per-CPU to atomic switch" code path, which
uses RCU when switching to atomic mode.  The enqueued callback
wakes up waiters waiting in the percpu_ref_switch_waitq.  Allowing
this callback to be lazy would result in unacceptable slowdowns for
users of per-CPU refcounts, such as blk_pre_runtime_suspend().

Therefore, make __percpu_ref_switch_to_atomic() use call_rcu_hurry()
in order to revert to the old behavior.

[ paulmck: Apply s/call_rcu_flush/call_rcu_hurry/ feedback from Tejun Heo. ]

Signed-off-by: Joel Fernandes (Google) <[e-mail]>
Signed-off-by: Paul E. McKenney <[e-mail]>
Cc: Dennis Zhou <[e-mail]>
Cc: Tejun Heo <[e-mail]>
Cc: Christoph Lameter <[e-mail]>
Cc: <[e-mail]>
---
 lib/percpu-refcount.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Comments

Joel Fernandes Nov. 30, 2022, 6:19 p.m. UTC | #1

Hi Tejun,

Could you give your ACK for this patch, for percpu refcount? The API
is renamed like in the workqueue one, as well.

Thanks a lot,

- Joel


On Wed, Nov 30, 2022 at 6:13 PM Paul E. McKenney <[e-mail]> wrote:
>
> From: "Joel Fernandes (Google)" <[e-mail]>
>
> Earlier commits in this series allow battery-powered systems to build
> their kernels with the default-disabled CONFIG_RCU_LAZY=y Kconfig option.
> This Kconfig option causes call_rcu() to delay its callbacks in order to
> batch callbacks.  This means that a given RCU grace period covers more
> callbacks, thus reducing the number of grace periods, in turn reducing
> the amount of energy consumed, which increases battery lifetime which
> can be a very good thing.  This is not a subtle effect: In some important
> use cases, the battery lifetime is increased by more than 10%.
>
> This CONFIG_RCU_LAZY=y option is available only for CPUs that offload
> callbacks, for example, CPUs mentioned in the rcu_nocbs kernel boot
> parameter passed to kernels built with CONFIG_RCU_NOCB_CPU=y.
>
> Delaying callbacks is normally not a problem because most callbacks do
> nothing but free memory.  If the system is short on memory, a shrinker
> will kick all currently queued lazy callbacks out of their laziness,
> thus freeing their memory in short order.  Similarly, the rcu_barrier()
> function, which blocks until all currently queued callbacks are invoked,
> will also kick lazy callbacks, thus enabling rcu_barrier() to complete
> in a timely manner.
>
> However, there are some cases where laziness is not a good option.
> For example, synchronize_rcu() invokes call_rcu(), and blocks until
> the newly queued callback is invoked.  It would not be a good for
> synchronize_rcu() to block for ten seconds, even on an idle system.
> Therefore, synchronize_rcu() invokes call_rcu_hurry() instead of
> call_rcu().  The arrival of a non-lazy call_rcu_hurry() callback on a
> given CPU kicks any lazy callbacks that might be already queued on that
> CPU.  After all, if there is going to be a grace period, all callbacks
> might as well get full benefit from it.
>
> Yes, this could be done the other way around by creating a
> call_rcu_lazy(), but earlier experience with this approach and
> feedback at the 2022 Linux Plumbers Conference shifted the approach
> to call_rcu() being lazy with call_rcu_hurry() for the few places
> where laziness is inappropriate.
>
> And another call_rcu() instance that cannot be lazy is the one on the
> percpu refcounter's "per-CPU to atomic switch" code path, which
> uses RCU when switching to atomic mode.  The enqueued callback
> wakes up waiters waiting in the percpu_ref_switch_waitq.  Allowing
> this callback to be lazy would result in unacceptable slowdowns for
> users of per-CPU refcounts, such as blk_pre_runtime_suspend().
>
> Therefore, make __percpu_ref_switch_to_atomic() use call_rcu_hurry()
> in order to revert to the old behavior.
>
> [ paulmck: Apply s/call_rcu_flush/call_rcu_hurry/ feedback from Tejun Heo. ]
>
> Signed-off-by: Joel Fernandes (Google) <[e-mail]>
> Signed-off-by: Paul E. McKenney <[e-mail]>
> Cc: Dennis Zhou <[e-mail]>
> Cc: Tejun Heo <[e-mail]>
> Cc: Christoph Lameter <[e-mail]>
> Cc: <[e-mail]>
> ---
>  lib/percpu-refcount.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/lib/percpu-refcount.c b/lib/percpu-refcount.c
> index e5c5315da2741..668f6aa6a75de 100644
> --- a/lib/percpu-refcount.c
> +++ b/lib/percpu-refcount.c
> @@ -230,7 +230,8 @@ static void __percpu_ref_switch_to_atomic(struct percpu_ref *ref,
>                 percpu_ref_noop_confirm_switch;
>
>         percpu_ref_get(ref);    /* put after confirmation */
> -       call_rcu(&ref->data->rcu, percpu_ref_switch_to_atomic_rcu);
> +       call_rcu_hurry(&ref->data->rcu,
> +                      percpu_ref_switch_to_atomic_rcu);
>  }
>
>  static void __percpu_ref_switch_to_percpu(struct percpu_ref *ref)
> --
> 2.31.1.189.g2e36527f23
>

Tejun Heo Nov. 30, 2022, 7:43 p.m. UTC | #2

On Wed, Nov 30, 2022 at 10:13:21AM -0800, Paul E. McKenney wrote:
> From: "Joel Fernandes (Google)" <[e-mail]>
> 
> Earlier commits in this series allow battery-powered systems to build
> their kernels with the default-disabled CONFIG_RCU_LAZY=y Kconfig option.
> This Kconfig option causes call_rcu() to delay its callbacks in order to
> batch callbacks.  This means that a given RCU grace period covers more
> callbacks, thus reducing the number of grace periods, in turn reducing
> the amount of energy consumed, which increases battery lifetime which
> can be a very good thing.  This is not a subtle effect: In some important
> use cases, the battery lifetime is increased by more than 10%.
> 
> This CONFIG_RCU_LAZY=y option is available only for CPUs that offload
> callbacks, for example, CPUs mentioned in the rcu_nocbs kernel boot
> parameter passed to kernels built with CONFIG_RCU_NOCB_CPU=y.
> 
> Delaying callbacks is normally not a problem because most callbacks do
> nothing but free memory.  If the system is short on memory, a shrinker
> will kick all currently queued lazy callbacks out of their laziness,
> thus freeing their memory in short order.  Similarly, the rcu_barrier()
> function, which blocks until all currently queued callbacks are invoked,
> will also kick lazy callbacks, thus enabling rcu_barrier() to complete
> in a timely manner.
> 
> However, there are some cases where laziness is not a good option.
> For example, synchronize_rcu() invokes call_rcu(), and blocks until
> the newly queued callback is invoked.  It would not be a good for
> synchronize_rcu() to block for ten seconds, even on an idle system.
> Therefore, synchronize_rcu() invokes call_rcu_hurry() instead of
> call_rcu().  The arrival of a non-lazy call_rcu_hurry() callback on a
> given CPU kicks any lazy callbacks that might be already queued on that
> CPU.  After all, if there is going to be a grace period, all callbacks
> might as well get full benefit from it.
> 
> Yes, this could be done the other way around by creating a
> call_rcu_lazy(), but earlier experience with this approach and
> feedback at the 2022 Linux Plumbers Conference shifted the approach
> to call_rcu() being lazy with call_rcu_hurry() for the few places
> where laziness is inappropriate.
> 
> And another call_rcu() instance that cannot be lazy is the one on the
> percpu refcounter's "per-CPU to atomic switch" code path, which
> uses RCU when switching to atomic mode.  The enqueued callback
> wakes up waiters waiting in the percpu_ref_switch_waitq.  Allowing
> this callback to be lazy would result in unacceptable slowdowns for
> users of per-CPU refcounts, such as blk_pre_runtime_suspend().
> 
> Therefore, make __percpu_ref_switch_to_atomic() use call_rcu_hurry()
> in order to revert to the old behavior.
> 
> [ paulmck: Apply s/call_rcu_flush/call_rcu_hurry/ feedback from Tejun Heo. ]
> 
> Signed-off-by: Joel Fernandes (Google) <[e-mail]>
> Signed-off-by: Paul E. McKenney <[e-mail]>
> Cc: Dennis Zhou <[e-mail]>
> Cc: Tejun Heo <[e-mail]>
> Cc: Christoph Lameter <[e-mail]>
> Cc: <[e-mail]>

Acked-by: Tejun Heo <[e-mail]>

Thanks.

Paul E. McKenney Nov. 30, 2022, 9:44 p.m. UTC | #3

On Wed, Nov 30, 2022 at 09:43:44AM -1000, Tejun Heo wrote:
> On Wed, Nov 30, 2022 at 10:13:21AM -0800, Paul E. McKenney wrote:
> > From: "Joel Fernandes (Google)" <[e-mail]>
> > 
> > Earlier commits in this series allow battery-powered systems to build
> > their kernels with the default-disabled CONFIG_RCU_LAZY=y Kconfig option.
> > This Kconfig option causes call_rcu() to delay its callbacks in order to
> > batch callbacks.  This means that a given RCU grace period covers more
> > callbacks, thus reducing the number of grace periods, in turn reducing
> > the amount of energy consumed, which increases battery lifetime which
> > can be a very good thing.  This is not a subtle effect: In some important
> > use cases, the battery lifetime is increased by more than 10%.
> > 
> > This CONFIG_RCU_LAZY=y option is available only for CPUs that offload
> > callbacks, for example, CPUs mentioned in the rcu_nocbs kernel boot
> > parameter passed to kernels built with CONFIG_RCU_NOCB_CPU=y.
> > 
> > Delaying callbacks is normally not a problem because most callbacks do
> > nothing but free memory.  If the system is short on memory, a shrinker
> > will kick all currently queued lazy callbacks out of their laziness,
> > thus freeing their memory in short order.  Similarly, the rcu_barrier()
> > function, which blocks until all currently queued callbacks are invoked,
> > will also kick lazy callbacks, thus enabling rcu_barrier() to complete
> > in a timely manner.
> > 
> > However, there are some cases where laziness is not a good option.
> > For example, synchronize_rcu() invokes call_rcu(), and blocks until
> > the newly queued callback is invoked.  It would not be a good for
> > synchronize_rcu() to block for ten seconds, even on an idle system.
> > Therefore, synchronize_rcu() invokes call_rcu_hurry() instead of
> > call_rcu().  The arrival of a non-lazy call_rcu_hurry() callback on a
> > given CPU kicks any lazy callbacks that might be already queued on that
> > CPU.  After all, if there is going to be a grace period, all callbacks
> > might as well get full benefit from it.
> > 
> > Yes, this could be done the other way around by creating a
> > call_rcu_lazy(), but earlier experience with this approach and
> > feedback at the 2022 Linux Plumbers Conference shifted the approach
> > to call_rcu() being lazy with call_rcu_hurry() for the few places
> > where laziness is inappropriate.
> > 
> > And another call_rcu() instance that cannot be lazy is the one on the
> > percpu refcounter's "per-CPU to atomic switch" code path, which
> > uses RCU when switching to atomic mode.  The enqueued callback
> > wakes up waiters waiting in the percpu_ref_switch_waitq.  Allowing
> > this callback to be lazy would result in unacceptable slowdowns for
> > users of per-CPU refcounts, such as blk_pre_runtime_suspend().
> > 
> > Therefore, make __percpu_ref_switch_to_atomic() use call_rcu_hurry()
> > in order to revert to the old behavior.
> > 
> > [ paulmck: Apply s/call_rcu_flush/call_rcu_hurry/ feedback from Tejun Heo. ]
> > 
> > Signed-off-by: Joel Fernandes (Google) <[e-mail]>
> > Signed-off-by: Paul E. McKenney <[e-mail]>
> > Cc: Dennis Zhou <[e-mail]>
> > Cc: Tejun Heo <[e-mail]>
> > Cc: Christoph Lameter <[e-mail]>
> > Cc: <[e-mail]>
> 
> Acked-by: Tejun Heo <[e-mail]>

I applied both, thank you very much!

							Thanx, Paul

Patch

diff --git a/lib/percpu-refcount.c b/lib/percpu-refcount.c
index e5c5315da2741..668f6aa6a75de 100644
--- a/lib/percpu-refcount.c
+++ b/lib/percpu-refcount.c
@@ -230,7 +230,8 @@  static void __percpu_ref_switch_to_atomic(struct percpu_ref *ref,
 		percpu_ref_noop_confirm_switch;
 
 	percpu_ref_get(ref);	/* put after confirmation */
-	call_rcu(&ref->data->rcu, percpu_ref_switch_to_atomic_rcu);
+	call_rcu_hurry(&ref->data->rcu,
+		       percpu_ref_switch_to_atomic_rcu);
 }
 
 static void __percpu_ref_switch_to_percpu(struct percpu_ref *ref)