[PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[Paul E. McKenney]

From: "Joel Fernandes (Google)" <joel@joelfernandes.org>

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 them.  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
in rxrpc_kill_connection(), which sometimes does a wakeup
that should not be unduly delayed.

Therefore, make rxrpc_kill_connection() 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) <joel@joelfernandes.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Marc Dionne <marc.dionne@auristor.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: <linux-afs@lists.infradead.org>
Cc: <netdev@vger.kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
---
 net/rxrpc/conn_object.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/rxrpc/conn_object.c b/net/rxrpc/conn_object.c
index 22089e37e97f0..9c5fae9ca106c 100644
--- a/net/rxrpc/conn_object.c
+++ b/net/rxrpc/conn_object.c
@@ -253,7 +253,7 @@ void rxrpc_kill_connection(struct rxrpc_connection *conn)
 	 * must carry a ref on the connection to prevent us getting here whilst
 	 * it is queued or running.
 	 */
-	call_rcu(&conn->rcu, rxrpc_destroy_connection);
+	call_rcu_hurry(&conn->rcu, rxrpc_destroy_connection);
 }
 
 /*
-- 
2.31.1.189.g2e36527f23

Re: [PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[Joel Fernandes]

Hi Eric,

Could you give your ACK for this patch?

The networking testing passed on ChromeOS and it has been in -next for
some time so has gotten testing there. The CONFIG option is default
disabled.

Thanks a lot,

- Joel

On Wed, Nov 30, 2022 at 6:13 PM Paul E. McKenney <paulmck@kernel.org> wrote:
>
> From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
>
> 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 them.  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
> in rxrpc_kill_connection(), which sometimes does a wakeup
> that should not be unduly delayed.
>
> Therefore, make rxrpc_kill_connection() 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) <joel@joelfernandes.org>
> Cc: David Howells <dhowells@redhat.com>
> Cc: Marc Dionne <marc.dionne@auristor.com>
> Cc: "David S. Miller" <davem@davemloft.net>
> Cc: Eric Dumazet <edumazet@google.com>
> Cc: Jakub Kicinski <kuba@kernel.org>
> Cc: Paolo Abeni <pabeni@redhat.com>
> Cc: <linux-afs@lists.infradead.org>
> Cc: <netdev@vger.kernel.org>
> Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
> ---
>  net/rxrpc/conn_object.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/net/rxrpc/conn_object.c b/net/rxrpc/conn_object.c
> index 22089e37e97f0..9c5fae9ca106c 100644
> --- a/net/rxrpc/conn_object.c
> +++ b/net/rxrpc/conn_object.c
> @@ -253,7 +253,7 @@ void rxrpc_kill_connection(struct rxrpc_connection *conn)
>          * must carry a ref on the connection to prevent us getting here whilst
>          * it is queued or running.
>          */
> -       call_rcu(&conn->rcu, rxrpc_destroy_connection);
> +       call_rcu_hurry(&conn->rcu, rxrpc_destroy_connection);
>  }
>
>  /*
> --
> 2.31.1.189.g2e36527f23
>

Re: [PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[Eric Dumazet]

Ah, I see a slightly better name has been chosen ;)

Reviewed-by: Eric Dumazet <edumazet@google.com>

On Wed, Nov 30, 2022 at 7:16 PM Joel Fernandes <joel@joelfernandes.org> wrote:
>
> Hi Eric,
>
> Could you give your ACK for this patch?
>
> The networking testing passed on ChromeOS and it has been in -next for
> some time so has gotten testing there. The CONFIG option is default
> disabled.
>
> Thanks a lot,
>
> - Joel
>
> On Wed, Nov 30, 2022 at 6:13 PM Paul E. McKenney <paulmck@kernel.org> wrote:
> >
> > From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
> >
> > 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 them.  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
> > in rxrpc_kill_connection(), which sometimes does a wakeup
> > that should not be unduly delayed.
> >
> > Therefore, make rxrpc_kill_connection() 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) <joel@joelfernandes.org>
> > Cc: David Howells <dhowells@redhat.com>
> > Cc: Marc Dionne <marc.dionne@auristor.com>
> > Cc: "David S. Miller" <davem@davemloft.net>
> > Cc: Eric Dumazet <edumazet@google.com>
> > Cc: Jakub Kicinski <kuba@kernel.org>
> > Cc: Paolo Abeni <pabeni@redhat.com>
> > Cc: <linux-afs@lists.infradead.org>
> > Cc: <netdev@vger.kernel.org>
> > Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
> > ---
> >  net/rxrpc/conn_object.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/net/rxrpc/conn_object.c b/net/rxrpc/conn_object.c
> > index 22089e37e97f0..9c5fae9ca106c 100644
> > --- a/net/rxrpc/conn_object.c
> > +++ b/net/rxrpc/conn_object.c
> > @@ -253,7 +253,7 @@ void rxrpc_kill_connection(struct rxrpc_connection *conn)
> >          * must carry a ref on the connection to prevent us getting here whilst
> >          * it is queued or running.
> >          */
> > -       call_rcu(&conn->rcu, rxrpc_destroy_connection);
> > +       call_rcu_hurry(&conn->rcu, rxrpc_destroy_connection);
> >  }
> >
> >  /*
> > --
> > 2.31.1.189.g2e36527f23
> >

Re: [PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[Paul E. McKenney]

On Wed, Nov 30, 2022 at 07:37:07PM +0100, Eric Dumazet wrote:
> Ah, I see a slightly better name has been chosen ;)

call_rcu_vite()?  call_rcu_tres_grande_vitesse()?  call_rcu_tgv()?

Sorry, couldn't resist!  ;-)

							Thanx, Paul

> Reviewed-by: Eric Dumazet <edumazet@google.com>
> 
> On Wed, Nov 30, 2022 at 7:16 PM Joel Fernandes <joel@joelfernandes.org> wrote:
> >
> > Hi Eric,
> >
> > Could you give your ACK for this patch?
> >
> > The networking testing passed on ChromeOS and it has been in -next for
> > some time so has gotten testing there. The CONFIG option is default
> > disabled.
> >
> > Thanks a lot,
> >
> > - Joel
> >
> > On Wed, Nov 30, 2022 at 6:13 PM Paul E. McKenney <paulmck@kernel.org> wrote:
> > >
> > > From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
> > >
> > > 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 them.  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
> > > in rxrpc_kill_connection(), which sometimes does a wakeup
> > > that should not be unduly delayed.
> > >
> > > Therefore, make rxrpc_kill_connection() 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) <joel@joelfernandes.org>
> > > Cc: David Howells <dhowells@redhat.com>
> > > Cc: Marc Dionne <marc.dionne@auristor.com>
> > > Cc: "David S. Miller" <davem@davemloft.net>
> > > Cc: Eric Dumazet <edumazet@google.com>
> > > Cc: Jakub Kicinski <kuba@kernel.org>
> > > Cc: Paolo Abeni <pabeni@redhat.com>
> > > Cc: <linux-afs@lists.infradead.org>
> > > Cc: <netdev@vger.kernel.org>
> > > Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
> > > ---
> > >  net/rxrpc/conn_object.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/net/rxrpc/conn_object.c b/net/rxrpc/conn_object.c
> > > index 22089e37e97f0..9c5fae9ca106c 100644
> > > --- a/net/rxrpc/conn_object.c
> > > +++ b/net/rxrpc/conn_object.c
> > > @@ -253,7 +253,7 @@ void rxrpc_kill_connection(struct rxrpc_connection *conn)
> > >          * must carry a ref on the connection to prevent us getting here whilst
> > >          * it is queued or running.
> > >          */
> > > -       call_rcu(&conn->rcu, rxrpc_destroy_connection);
> > > +       call_rcu_hurry(&conn->rcu, rxrpc_destroy_connection);
> > >  }
> > >
> > >  /*
> > > --
> > > 2.31.1.189.g2e36527f23
> > >

Re: [PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[Steven Rostedt]

On Wed, 30 Nov 2022 13:45:52 -0800
"Paul E. McKenney" <paulmck@kernel.org> wrote:

> On Wed, Nov 30, 2022 at 07:37:07PM +0100, Eric Dumazet wrote:
> > Ah, I see a slightly better name has been chosen ;)  
> 
> call_rcu_vite()?  call_rcu_tres_grande_vitesse()?  call_rcu_tgv()?
> 
> Sorry, couldn't resist!  ;-)
> 
>

  call_rcu_twitter_2_0()  ?

-- Steve

Re: [PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[Paul E. McKenney]

On Wed, Nov 30, 2022 at 04:49:49PM -0500, Steven Rostedt wrote:
> On Wed, 30 Nov 2022 13:45:52 -0800
> "Paul E. McKenney" <paulmck@kernel.org> wrote:
> 
> > On Wed, Nov 30, 2022 at 07:37:07PM +0100, Eric Dumazet wrote:
> > > Ah, I see a slightly better name has been chosen ;)  
> > 
> > call_rcu_vite()?  call_rcu_tres_grande_vitesse()?  call_rcu_tgv()?
> > 
> > Sorry, couldn't resist!  ;-)
> 
>   call_rcu_twitter_2_0()  ?

call_rcu_grace_period_finishes_before_it_starts() ?

							Thanx, Paul

Re: [PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[David Howells]

Note that this conflicts with my patch:

	rxrpc: Don't hold a ref for connection workqueue
	https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/commit/?h=rxrpc-next&id=450b00011290660127c2d76f5c5ed264126eb229

which should render it unnecessary.  It's a little ahead of yours in the
net-next queue, if that means anything.

David

Re: [PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[Joel Fernandes]

> On Nov 30, 2022, at 2:09 PM, David Howells <dhowells@redhat.com> wrote:
> 
> Note that this conflicts with my patch:

Oh.  I don’t see any review or Ack tags on it. Is it still under review?

Thanks,

- Joel



> 
>    rxrpc: Don't hold a ref for connection workqueue
>    https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/commit/?h=rxrpc-next&id=450b00011290660127c2d76f5c5ed264126eb229
> 
> which should render it unnecessary.  It's a little ahead of yours in the
> net-next queue, if that means anything.
> 
> David
> 

Re: [PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[Paul E. McKenney]

On Wed, Nov 30, 2022 at 02:20:52PM -0500, Joel Fernandes wrote:
> 
> 
> > On Nov 30, 2022, at 2:09 PM, David Howells <dhowells@redhat.com> wrote:
> > 
> > Note that this conflicts with my patch:
> 
> Oh.  I don’t see any review or Ack tags on it. Is it still under review?

So what I have done is to drop this patch from the series, but to also
preserve it for posterity at -rcu branch lazy-obsolete.2022.11.30a.

It looks like that wakeup is still delayed, but I could easily be
missing something.

Joel, could you please test the effects of having the current lazy branch,
but also David Howells's patch?  That way, if there is an issue, we can
work it sooner rather than later, and if it all works fine, we can stop
worrying about it.  ;-)

							Thanx, Paul

> Thanks,
> 
> - Joel
> 
> 
> 
> > 
> >    rxrpc: Don't hold a ref for connection workqueue
> >    https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/commit/?h=rxrpc-next&id=450b00011290660127c2d76f5c5ed264126eb229
> > 
> > which should render it unnecessary.  It's a little ahead of yours in the
> > net-next queue, if that means anything.
> > 
> > David
> > 

Re: [PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[Joel Fernandes]

> On Nov 30, 2022, at 4:43 PM, Paul E. McKenney <paulmck@kernel.org> wrote:
> 
> On Wed, Nov 30, 2022 at 02:20:52PM -0500, Joel Fernandes wrote:
>> 
>> 
>>>> On Nov 30, 2022, at 2:09 PM, David Howells <dhowells@redhat.com> wrote:
>>> 
>>> Note that this conflicts with my patch:
>> 
>> Oh.  I don’t see any review or Ack tags on it. Is it still under review?
> 
> So what I have done is to drop this patch from the series, but to also
> preserve it for posterity at -rcu branch lazy-obsolete.2022.11.30a.
> 
> It looks like that wakeup is still delayed, but I could easily be
> missing something.
> 
> Joel, could you please test the effects of having the current lazy branch,
> but also David Howells's patch?  That way, if there is an issue, we can
> work it sooner rather than later, and if it all works fine, we can stop
> worrying about it.  ;-)

Sure, I will kick off the failing test and see if it passes with Davids patch. Will let you know.

Thanks,

 - Joel


>                            Thanx, Paul
> 
>> Thanks,
>> 
>> - Joel
>> 
>> 
>> 
>>> 
>>>   rxrpc: Don't hold a ref for connection workqueue
>>>   https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/commit/?h=rxrpc-next&id=450b00011290660127c2d76f5c5ed264126eb229
>>> 
>>> which should render it unnecessary.  It's a little ahead of yours in the
>>> net-next queue, if that means anything.
>>> 
>>> David
>>> 

Re: [PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[Paul E. McKenney]

On Wed, Nov 30, 2022 at 07:09:04PM +0000, David Howells wrote:
> Note that this conflicts with my patch:
> 
> 	rxrpc: Don't hold a ref for connection workqueue
> 	https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/commit/?h=rxrpc-next&id=450b00011290660127c2d76f5c5ed264126eb229
> 
> which should render it unnecessary.  It's a little ahead of yours in the
> net-next queue, if that means anything.

OK, I will drop this patch in favor of yours, thank you!

							Thanx, Paul

Re: [PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[Joel Fernandes]

Hi David,

On Wed, Nov 30, 2022 at 7:09 PM David Howells <dhowells@redhat.com> wrote:
>
> Note that this conflicts with my patch:
>
>         rxrpc: Don't hold a ref for connection workqueue
>         https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/commit/?h=rxrpc-next&id=450b00011290660127c2d76f5c5ed264126eb229
>
> which should render it unnecessary.  It's a little ahead of yours in the
> net-next queue, if that means anything.

Could you clarify why it is unnecessary?

After your patch, you are still doing a wake up in your call_rcu() callback:

- ASSERTCMP(refcount_read(&conn->ref), ==, 0);
+ if (atomic_dec_and_test(&rxnet->nr_conns))
+    wake_up_var(&rxnet->nr_conns);
+}

Are you saying the code can now tolerate delays? What if the RCU
callback is invoked after arbitrarily long delays making the sleeping
process to wait?

If you agree, you can convert the call_rcu() to call_rcu_hurry() in
your patch itself. Would you be willing to do that? If not, that's
totally OK and I can send a patch later once yours is in (after
further testing).

Thanks,

 - Joel

Re: [PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[David Howells]

Joel Fernandes <joel@joelfernandes.org> wrote:

> > Note that this conflicts with my patch:
> >
> >         rxrpc: Don't hold a ref for connection workqueue
> >         https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/commit/?h=rxrpc-next&id=450b00011290660127c2d76f5c5ed264126eb229
> >
> > which should render it unnecessary.  It's a little ahead of yours in the
> > net-next queue, if that means anything.
> 
> Could you clarify why it is unnecessary?

Rather than tearing down parts of the connection it only logs a trace line,
frees the memory and decrements the counter on the namespace.  This it used to
account that all the pieces of memory allocated in that namespace are gone
before the namespace is removed to check for leaks.  The RCU cleanup used to
use some other stuff (such as the peer hash) in the rxrpc_net struct but no
longer will after the patches I submitted.

> After your patch, you are still doing a wake up in your call_rcu() callback:
>
> - ASSERTCMP(refcount_read(&conn->ref), ==, 0);
> + if (atomic_dec_and_test(&rxnet->nr_conns))
> +    wake_up_var(&rxnet->nr_conns);
> +}
> 
> Are you saying the code can now tolerate delays? What if the RCU
> callback is invoked after arbitrarily long delays making the sleeping
> process to wait?

True.  But that now only holds up the destruction of a net namespace and the
removal of the rxrpc module.

> If you agree, you can convert the call_rcu() to call_rcu_hurry() in
> your patch itself. Would you be willing to do that? If not, that's
> totally OK and I can send a patch later once yours is in (after
> further testing).

I can add it to part 4 (see my rxrpc-ringless-5 branch) if it is necessary.

David

Re: [PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[Joel Fernandes]

On Wed, Nov 30, 2022 at 11:05 PM David Howells <dhowells@redhat.com> wrote:
>
> Joel Fernandes <joel@joelfernandes.org> wrote:
>
> > > Note that this conflicts with my patch:
> > >
> > >         rxrpc: Don't hold a ref for connection workqueue
> > >         https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/commit/?h=rxrpc-next&id=450b00011290660127c2d76f5c5ed264126eb229
> > >
> > > which should render it unnecessary.  It's a little ahead of yours in the
> > > net-next queue, if that means anything.
> >
> > Could you clarify why it is unnecessary?
>
> Rather than tearing down parts of the connection it only logs a trace line,
> frees the memory and decrements the counter on the namespace.  This it used to
> account that all the pieces of memory allocated in that namespace are gone
> before the namespace is removed to check for leaks.  The RCU cleanup used to
> use some other stuff (such as the peer hash) in the rxrpc_net struct but no
> longer will after the patches I submitted.
>
> > After your patch, you are still doing a wake up in your call_rcu() callback:
> >
> > - ASSERTCMP(refcount_read(&conn->ref), ==, 0);
> > + if (atomic_dec_and_test(&rxnet->nr_conns))
> > +    wake_up_var(&rxnet->nr_conns);
> > +}
> >
> > Are you saying the code can now tolerate delays? What if the RCU
> > callback is invoked after arbitrarily long delays making the sleeping
> > process to wait?
>
> True.  But that now only holds up the destruction of a net namespace and the
> removal of the rxrpc module.
>
> > If you agree, you can convert the call_rcu() to call_rcu_hurry() in
> > your patch itself. Would you be willing to do that? If not, that's
> > totally OK and I can send a patch later once yours is in (after
> > further testing).
>
> I can add it to part 4 (see my rxrpc-ringless-5 branch) if it is necessary.

Ok sounds good, on module removal the rcu_barrier() will flush out
pending callbacks so that should not be an issue.

Based on your message, I think we can drop this patch then. Since Paul
is already dropping it, no other action is needed.

(I just realized my patch was not fixing a test failure, like the
other net ones did, but rather we found the issue by static analysis
-- i.e. programmatically auditing all callbacks in the kernel doing
wake ups).

thanks,
 - Joel

Re: [PATCH rcu 14/16] rxrpc: Use call_rcu_hurry() instead of call_rcu()

[Joel Fernandes]

On Wed, Nov 30, 2022 at 11:05:22PM +0000, David Howells wrote:
> Joel Fernandes <joel@joelfernandes.org> wrote:
[...] 
> > After your patch, you are still doing a wake up in your call_rcu() callback:
> >
> > - ASSERTCMP(refcount_read(&conn->ref), ==, 0);
> > + if (atomic_dec_and_test(&rxnet->nr_conns))
> > +    wake_up_var(&rxnet->nr_conns);
> > +}
> > 
> > Are you saying the code can now tolerate delays? What if the RCU
> > callback is invoked after arbitrarily long delays making the sleeping
> > process to wait?
> 
> True.  But that now only holds up the destruction of a net namespace and the
> removal of the rxrpc module.

I am guessing not destructing the net namespace soon enough is not an issue.
I do remember (in a different patch) that not tearing down networking things
have a weird side effect to tools that require state to disappear..

> > If you agree, you can convert the call_rcu() to call_rcu_hurry() in
> > your patch itself. Would you be willing to do that? If not, that's
> > totally OK and I can send a patch later once yours is in (after
> > further testing).
> 
> I can add it to part 4 (see my rxrpc-ringless-5 branch) if it is necessary.

I am guessing the conversion to call_rcu_hurry() is still not necessary here,
if it is then consider the conversion.

But yeah feel free to ignore this, I am just pinging here so that it did not
slip through the cracks.

thanks,

 - Joel

[PATCH rcu 15/16] net: Use call_rcu_hurry() for dst_release()

[Paul E. McKenney]

From: "Joel Fernandes (Google)" <joel@joelfernandes.org>

In a networking test on ChromeOS, kernels built with the new
CONFIG_RCU_LAZY=y Kconfig option fail a networking test in the teardown
phase.

This failure may be reproduced as follows: ip netns del <name>

The CONFIG_RCU_LAZY=y Kconfig option was introduced by earlier commits
in this series for the benefit of certain battery-powered systems.
This Kconfig option causes call_rcu() to delay its callbacks in order
to batch them.  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.

Returning to the test failure, use of ftrace showed that this failure
cause caused by the aadded delays due to this new lazy behavior of
call_rcu() in kernels built with CONFIG_RCU_LAZY=y.

Therefore, make dst_release() use call_rcu_hurry() in order to revert
to the old test-failure-free behavior.

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

Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Cc: David Ahern <dsahern@kernel.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: <netdev@vger.kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
---
 net/core/dst.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/core/dst.c b/net/core/dst.c
index bc9c9be4e0801..a4e738d321ba2 100644
--- a/net/core/dst.c
+++ b/net/core/dst.c
@@ -174,7 +174,7 @@ void dst_release(struct dst_entry *dst)
 			net_warn_ratelimited("%s: dst:%p refcnt:%d\n",
 					     __func__, dst, newrefcnt);
 		if (!newrefcnt)
-			call_rcu(&dst->rcu_head, dst_destroy_rcu);
+			call_rcu_hurry(&dst->rcu_head, dst_destroy_rcu);
 	}
 }
 EXPORT_SYMBOL(dst_release);
-- 
2.31.1.189.g2e36527f23

Re: [PATCH rcu 15/16] net: Use call_rcu_hurry() for dst_release()

[Joel Fernandes]

Hi Eric,

Could you give your ACK for this patch for this one as well? This is
the other networking one.

The networking testing passed on ChromeOS and it has been in -next for
some time so has gotten testing there. The CONFIG option is default
disabled.

Thanks a lot,

- Joel

On Wed, Nov 30, 2022 at 6:14 PM Paul E. McKenney <paulmck@kernel.org> wrote:
>
> From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
>
> In a networking test on ChromeOS, kernels built with the new
> CONFIG_RCU_LAZY=y Kconfig option fail a networking test in the teardown
> phase.
>
> This failure may be reproduced as follows: ip netns del <name>
>
> The CONFIG_RCU_LAZY=y Kconfig option was introduced by earlier commits
> in this series for the benefit of certain battery-powered systems.
> This Kconfig option causes call_rcu() to delay its callbacks in order
> to batch them.  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.
>
> Returning to the test failure, use of ftrace showed that this failure
> cause caused by the aadded delays due to this new lazy behavior of
> call_rcu() in kernels built with CONFIG_RCU_LAZY=y.
>
> Therefore, make dst_release() use call_rcu_hurry() in order to revert
> to the old test-failure-free behavior.
>
> [ paulmck: Apply s/call_rcu_flush/call_rcu_hurry/ feedback from Tejun Heo. ]
>
> Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> Cc: David Ahern <dsahern@kernel.org>
> Cc: "David S. Miller" <davem@davemloft.net>
> Cc: Eric Dumazet <edumazet@google.com>
> Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
> Cc: Jakub Kicinski <kuba@kernel.org>
> Cc: Paolo Abeni <pabeni@redhat.com>
> Cc: <netdev@vger.kernel.org>
> Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
> ---
>  net/core/dst.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/net/core/dst.c b/net/core/dst.c
> index bc9c9be4e0801..a4e738d321ba2 100644
> --- a/net/core/dst.c
> +++ b/net/core/dst.c
> @@ -174,7 +174,7 @@ void dst_release(struct dst_entry *dst)
>                         net_warn_ratelimited("%s: dst:%p refcnt:%d\n",
>                                              __func__, dst, newrefcnt);
>                 if (!newrefcnt)
> -                       call_rcu(&dst->rcu_head, dst_destroy_rcu);
> +                       call_rcu_hurry(&dst->rcu_head, dst_destroy_rcu);
>         }
>  }
>  EXPORT_SYMBOL(dst_release);
> --
> 2.31.1.189.g2e36527f23
>

Re: [PATCH rcu 15/16] net: Use call_rcu_hurry() for dst_release()

[Eric Dumazet]

Sure, thanks.

Reviewed-by: Eric Dumazet <edumazet@google.com>

I think we can work later to change how dst are freed/released to
avoid using call_rcu_hurry()

On Wed, Nov 30, 2022 at 7:17 PM Joel Fernandes <joel@joelfernandes.org> wrote:
>
> Hi Eric,
>
> Could you give your ACK for this patch for this one as well? This is
> the other networking one.
>
> The networking testing passed on ChromeOS and it has been in -next for
> some time so has gotten testing there. The CONFIG option is default
> disabled.
>
> Thanks a lot,
>
> - Joel
>
> On Wed, Nov 30, 2022 at 6:14 PM Paul E. McKenney <paulmck@kernel.org> wrote:
> >
> > From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
> >
> > In a networking test on ChromeOS, kernels built with the new
> > CONFIG_RCU_LAZY=y Kconfig option fail a networking test in the teardown
> > phase.
> >
> > This failure may be reproduced as follows: ip netns del <name>
> >
> > The CONFIG_RCU_LAZY=y Kconfig option was introduced by earlier commits
> > in this series for the benefit of certain battery-powered systems.
> > This Kconfig option causes call_rcu() to delay its callbacks in order
> > to batch them.  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.
> >
> > Returning to the test failure, use of ftrace showed that this failure
> > cause caused by the aadded delays due to this new lazy behavior of
> > call_rcu() in kernels built with CONFIG_RCU_LAZY=y.
> >
> > Therefore, make dst_release() use call_rcu_hurry() in order to revert
> > to the old test-failure-free behavior.
> >
> > [ paulmck: Apply s/call_rcu_flush/call_rcu_hurry/ feedback from Tejun Heo. ]
> >
> > Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> > Cc: David Ahern <dsahern@kernel.org>
> > Cc: "David S. Miller" <davem@davemloft.net>
> > Cc: Eric Dumazet <edumazet@google.com>
> > Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
> > Cc: Jakub Kicinski <kuba@kernel.org>
> > Cc: Paolo Abeni <pabeni@redhat.com>
> > Cc: <netdev@vger.kernel.org>
> > Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
> > ---
> >  net/core/dst.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/net/core/dst.c b/net/core/dst.c
> > index bc9c9be4e0801..a4e738d321ba2 100644
> > --- a/net/core/dst.c
> > +++ b/net/core/dst.c
> > @@ -174,7 +174,7 @@ void dst_release(struct dst_entry *dst)
> >                         net_warn_ratelimited("%s: dst:%p refcnt:%d\n",
> >                                              __func__, dst, newrefcnt);
> >                 if (!newrefcnt)
> > -                       call_rcu(&dst->rcu_head, dst_destroy_rcu);
> > +                       call_rcu_hurry(&dst->rcu_head, dst_destroy_rcu);
> >         }
> >  }
> >  EXPORT_SYMBOL(dst_release);
> > --
> > 2.31.1.189.g2e36527f23
> >

Re: [PATCH rcu 15/16] net: Use call_rcu_hurry() for dst_release()

[Joel Fernandes]

Hi Eric,

On Wed, Nov 30, 2022 at 6:39 PM Eric Dumazet <edumazet@google.com> wrote:
>
> Sure, thanks.
>
> Reviewed-by: Eric Dumazet <edumazet@google.com>
>
> I think we can work later to change how dst are freed/released to
> avoid using call_rcu_hurry()

That sounds great, if you can give me any high-level guidance (in the
future) on that and what to look for, I can give it a try as well. I
have been wanting to learn more about the networking code :-)

Thanks,

 - Joel


> On Wed, Nov 30, 2022 at 7:17 PM Joel Fernandes <joel@joelfernandes.org> wrote:
> >
> > Hi Eric,
> >
> > Could you give your ACK for this patch for this one as well? This is
> > the other networking one.
> >
> > The networking testing passed on ChromeOS and it has been in -next for
> > some time so has gotten testing there. The CONFIG option is default
> > disabled.
> >
> > Thanks a lot,
> >
> > - Joel
> >
> > On Wed, Nov 30, 2022 at 6:14 PM Paul E. McKenney <paulmck@kernel.org> wrote:
> > >
> > > From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
> > >
> > > In a networking test on ChromeOS, kernels built with the new
> > > CONFIG_RCU_LAZY=y Kconfig option fail a networking test in the teardown
> > > phase.
> > >
> > > This failure may be reproduced as follows: ip netns del <name>
> > >
> > > The CONFIG_RCU_LAZY=y Kconfig option was introduced by earlier commits
> > > in this series for the benefit of certain battery-powered systems.
> > > This Kconfig option causes call_rcu() to delay its callbacks in order
> > > to batch them.  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.
> > >
> > > Returning to the test failure, use of ftrace showed that this failure
> > > cause caused by the aadded delays due to this new lazy behavior of
> > > call_rcu() in kernels built with CONFIG_RCU_LAZY=y.
> > >
> > > Therefore, make dst_release() use call_rcu_hurry() in order to revert
> > > to the old test-failure-free behavior.
> > >
> > > [ paulmck: Apply s/call_rcu_flush/call_rcu_hurry/ feedback from Tejun Heo. ]
> > >
> > > Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> > > Cc: David Ahern <dsahern@kernel.org>
> > > Cc: "David S. Miller" <davem@davemloft.net>
> > > Cc: Eric Dumazet <edumazet@google.com>
> > > Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
> > > Cc: Jakub Kicinski <kuba@kernel.org>
> > > Cc: Paolo Abeni <pabeni@redhat.com>
> > > Cc: <netdev@vger.kernel.org>
> > > Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
> > > ---
> > >  net/core/dst.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/net/core/dst.c b/net/core/dst.c
> > > index bc9c9be4e0801..a4e738d321ba2 100644
> > > --- a/net/core/dst.c
> > > +++ b/net/core/dst.c
> > > @@ -174,7 +174,7 @@ void dst_release(struct dst_entry *dst)
> > >                         net_warn_ratelimited("%s: dst:%p refcnt:%d\n",
> > >                                              __func__, dst, newrefcnt);
> > >                 if (!newrefcnt)
> > > -                       call_rcu(&dst->rcu_head, dst_destroy_rcu);
> > > +                       call_rcu_hurry(&dst->rcu_head, dst_destroy_rcu);
> > >         }
> > >  }
> > >  EXPORT_SYMBOL(dst_release);
> > > --
> > > 2.31.1.189.g2e36527f23
> > >

Re: [PATCH rcu 15/16] net: Use call_rcu_hurry() for dst_release()

[Paul E. McKenney]

On Wed, Nov 30, 2022 at 07:39:02PM +0100, Eric Dumazet wrote:
> Sure, thanks.
> 
> Reviewed-by: Eric Dumazet <edumazet@google.com>

Applied, thank you!!!

> I think we can work later to change how dst are freed/released to
> avoid using call_rcu_hurry()

Thank you for being willing to look into that!

							Thanx, Paul

> On Wed, Nov 30, 2022 at 7:17 PM Joel Fernandes <joel@joelfernandes.org> wrote:
> >
> > Hi Eric,
> >
> > Could you give your ACK for this patch for this one as well? This is
> > the other networking one.
> >
> > The networking testing passed on ChromeOS and it has been in -next for
> > some time so has gotten testing there. The CONFIG option is default
> > disabled.
> >
> > Thanks a lot,
> >
> > - Joel
> >
> > On Wed, Nov 30, 2022 at 6:14 PM Paul E. McKenney <paulmck@kernel.org> wrote:
> > >
> > > From: "Joel Fernandes (Google)" <joel@joelfernandes.org>
> > >
> > > In a networking test on ChromeOS, kernels built with the new
> > > CONFIG_RCU_LAZY=y Kconfig option fail a networking test in the teardown
> > > phase.
> > >
> > > This failure may be reproduced as follows: ip netns del <name>
> > >
> > > The CONFIG_RCU_LAZY=y Kconfig option was introduced by earlier commits
> > > in this series for the benefit of certain battery-powered systems.
> > > This Kconfig option causes call_rcu() to delay its callbacks in order
> > > to batch them.  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.
> > >
> > > Returning to the test failure, use of ftrace showed that this failure
> > > cause caused by the aadded delays due to this new lazy behavior of
> > > call_rcu() in kernels built with CONFIG_RCU_LAZY=y.
> > >
> > > Therefore, make dst_release() use call_rcu_hurry() in order to revert
> > > to the old test-failure-free behavior.
> > >
> > > [ paulmck: Apply s/call_rcu_flush/call_rcu_hurry/ feedback from Tejun Heo. ]
> > >
> > > Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> > > Cc: David Ahern <dsahern@kernel.org>
> > > Cc: "David S. Miller" <davem@davemloft.net>
> > > Cc: Eric Dumazet <edumazet@google.com>
> > > Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
> > > Cc: Jakub Kicinski <kuba@kernel.org>
> > > Cc: Paolo Abeni <pabeni@redhat.com>
> > > Cc: <netdev@vger.kernel.org>
> > > Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
> > > ---
> > >  net/core/dst.c | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/net/core/dst.c b/net/core/dst.c
> > > index bc9c9be4e0801..a4e738d321ba2 100644
> > > --- a/net/core/dst.c
> > > +++ b/net/core/dst.c
> > > @@ -174,7 +174,7 @@ void dst_release(struct dst_entry *dst)
> > >                         net_warn_ratelimited("%s: dst:%p refcnt:%d\n",
> > >                                              __func__, dst, newrefcnt);
> > >                 if (!newrefcnt)
> > > -                       call_rcu(&dst->rcu_head, dst_destroy_rcu);
> > > +                       call_rcu_hurry(&dst->rcu_head, dst_destroy_rcu);
> > >         }
> > >  }
> > >  EXPORT_SYMBOL(dst_release);
> > > --
> > > 2.31.1.189.g2e36527f23
> > >

[PATCH rcu 16/16] net: devinet: Reduce refcount before grace period

[Paul E. McKenney]

From: Eric Dumazet <edumazet@google.com>

Currently, the inetdev_destroy() function waits for an RCU grace period
before decrementing the refcount and freeing memory. This causes a delay
with a new RCU configuration that tries to save power, which results in the
network interface disappearing later than expected. The resulting delay
causes test failures on ChromeOS.

Refactor the code such that the refcount is freed before the grace period
and memory is freed after. With this a ChromeOS network test passes that
does 'ip netns del' and polls for an interface disappearing, now passes.

Reported-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Cc: David Ahern <dsahern@kernel.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Hideaki YOSHIFUJI <yoshfuji@linux-ipv6.org>
Cc: Jakub Kicinski <kuba@kernel.org>
Cc: Paolo Abeni <pabeni@redhat.com>
Cc: <netdev@vger.kernel.org>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
---
 net/ipv4/devinet.c | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)

diff --git a/net/ipv4/devinet.c b/net/ipv4/devinet.c
index e8b9a9202fecd..b0acf6e19aed3 100644
--- a/net/ipv4/devinet.c
+++ b/net/ipv4/devinet.c
@@ -234,13 +234,20 @@ static void inet_free_ifa(struct in_ifaddr *ifa)
 	call_rcu(&ifa->rcu_head, inet_rcu_free_ifa);
 }
 
+static void in_dev_free_rcu(struct rcu_head *head)
+{
+	struct in_device *idev = container_of(head, struct in_device, rcu_head);
+
+	kfree(rcu_dereference_protected(idev->mc_hash, 1));
+	kfree(idev);
+}
+
 void in_dev_finish_destroy(struct in_device *idev)
 {
 	struct net_device *dev = idev->dev;
 
 	WARN_ON(idev->ifa_list);
 	WARN_ON(idev->mc_list);
-	kfree(rcu_dereference_protected(idev->mc_hash, 1));
 #ifdef NET_REFCNT_DEBUG
 	pr_debug("%s: %p=%s\n", __func__, idev, dev ? dev->name : "NIL");
 #endif
@@ -248,7 +255,7 @@ void in_dev_finish_destroy(struct in_device *idev)
 	if (!idev->dead)
 		pr_err("Freeing alive in_device %p\n", idev);
 	else
-		kfree(idev);
+		call_rcu(&idev->rcu_head, in_dev_free_rcu);
 }
 EXPORT_SYMBOL(in_dev_finish_destroy);
 
@@ -298,12 +305,6 @@ static struct in_device *inetdev_init(struct net_device *dev)
 	goto out;
 }
 
-static void in_dev_rcu_put(struct rcu_head *head)
-{
-	struct in_device *idev = container_of(head, struct in_device, rcu_head);
-	in_dev_put(idev);
-}
-
 static void inetdev_destroy(struct in_device *in_dev)
 {
 	struct net_device *dev;
@@ -328,7 +329,7 @@ static void inetdev_destroy(struct in_device *in_dev)
 	neigh_parms_release(&arp_tbl, in_dev->arp_parms);
 	arp_ifdown(dev);
 
-	call_rcu(&in_dev->rcu_head, in_dev_rcu_put);
+	in_dev_put(in_dev);
 }
 
 int inet_addr_onlink(struct in_device *in_dev, __be32 a, __be32 b)
-- 
2.31.1.189.g2e36527f23