diff mbox series

[5/6] rcu/nocb: Fix WARN_ON_ONCE() in the rcu_nocb_bypass_lock()

Message ID 20240201014100.3204532-6-boqun.feng@gmail.com
State New
Headers
Series RCU nocb updates for v6.9 |

Commit Message

Boqun Feng Feb. 1, 2024, 1:40 a.m. UTC 
  From: Zqiang <qiang.zhang1211@gmail.com>

For the kernels built with CONFIG_RCU_NOCB_CPU_DEFAULT_ALL=y and
CONFIG_RCU_LAZY=y, the following scenarios will trigger WARN_ON_ONCE()
in the rcu_nocb_bypass_lock() and rcu_nocb_wait_contended() functions:

        CPU2                                               CPU11
kthread
rcu_nocb_cb_kthread                                       ksys_write
rcu_do_batch                                              vfs_write
rcu_torture_timer_cb                                      proc_sys_write
__kmem_cache_free                                         proc_sys_call_handler
kmemleak_free                                             drop_caches_sysctl_handler
delete_object_full                                        drop_slab
__delete_object                                           shrink_slab
put_object                                                lazy_rcu_shrink_scan
call_rcu                                                  rcu_nocb_flush_bypass
__call_rcu_commn                                            rcu_nocb_bypass_lock
                                                            raw_spin_trylock(&rdp->nocb_bypass_lock) fail
                                                            atomic_inc(&rdp->nocb_lock_contended);
rcu_nocb_wait_contended                                     WARN_ON_ONCE(smp_processor_id() != rdp->cpu);
 WARN_ON_ONCE(atomic_read(&rdp->nocb_lock_contended))                                          |
                            |_ _ _ _ _ _ _ _ _ _same rdp and rdp->cpu != 11_ _ _ _ _ _ _ _ _ __|

Reproduce this bug with "echo 3 > /proc/sys/vm/drop_caches".

This commit therefore uses rcu_nocb_try_flush_bypass() instead of
rcu_nocb_flush_bypass() in lazy_rcu_shrink_scan().  If the nocb_bypass
queue is being flushed, then rcu_nocb_try_flush_bypass will return
directly.

Signed-off-by: Zqiang <qiang.zhang1211@gmail.com>
Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
---
 kernel/rcu/tree_nocb.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Oleksandr Natalenko Feb. 24, 2024, 4:30 p.m. UTC | #1 
Hello.

On čtvrtek 1. února 2024 2:40:57 CET Boqun Feng wrote:
> From: Zqiang <qiang.zhang1211@gmail.com>
> 
> For the kernels built with CONFIG_RCU_NOCB_CPU_DEFAULT_ALL=y and
> CONFIG_RCU_LAZY=y, the following scenarios will trigger WARN_ON_ONCE()
> in the rcu_nocb_bypass_lock() and rcu_nocb_wait_contended() functions:
> 
>         CPU2                                               CPU11
> kthread
> rcu_nocb_cb_kthread                                       ksys_write
> rcu_do_batch                                              vfs_write
> rcu_torture_timer_cb                                      proc_sys_write
> __kmem_cache_free                                         proc_sys_call_handler
> kmemleak_free                                             drop_caches_sysctl_handler
> delete_object_full                                        drop_slab
> __delete_object                                           shrink_slab
> put_object                                                lazy_rcu_shrink_scan
> call_rcu                                                  rcu_nocb_flush_bypass
> __call_rcu_commn                                            rcu_nocb_bypass_lock
>                                                             raw_spin_trylock(&rdp->nocb_bypass_lock) fail
>                                                             atomic_inc(&rdp->nocb_lock_contended);
> rcu_nocb_wait_contended                                     WARN_ON_ONCE(smp_processor_id() != rdp->cpu);
>  WARN_ON_ONCE(atomic_read(&rdp->nocb_lock_contended))                                          |
>                             |_ _ _ _ _ _ _ _ _ _same rdp and rdp->cpu != 11_ _ _ _ _ _ _ _ _ __|
> 
> Reproduce this bug with "echo 3 > /proc/sys/vm/drop_caches".
> 
> This commit therefore uses rcu_nocb_try_flush_bypass() instead of
> rcu_nocb_flush_bypass() in lazy_rcu_shrink_scan().  If the nocb_bypass
> queue is being flushed, then rcu_nocb_try_flush_bypass will return
> directly.
> 
> Signed-off-by: Zqiang <qiang.zhang1211@gmail.com>
> Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
> Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
> Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
> ---
>  kernel/rcu/tree_nocb.h | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/kernel/rcu/tree_nocb.h b/kernel/rcu/tree_nocb.h
> index 9e8052ba14b9..ffa69a5e18f4 100644
> --- a/kernel/rcu/tree_nocb.h
> +++ b/kernel/rcu/tree_nocb.h
> @@ -1391,7 +1391,7 @@ lazy_rcu_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
>  			rcu_nocb_unlock_irqrestore(rdp, flags);
>  			continue;
>  		}
> -		WARN_ON_ONCE(!rcu_nocb_flush_bypass(rdp, NULL, jiffies, false));
> +		rcu_nocb_try_flush_bypass(rdp, jiffies);
>  		rcu_nocb_unlock_irqrestore(rdp, flags);
>  		wake_nocb_gp(rdp, false);
>  		sc->nr_to_scan -= _count;
> 

Does this fix [1] [2]?

Thank you.

[1] https://bugzilla.kernel.org/show_bug.cgi?id=217948
[2] https://lore.kernel.org/lkml/8461340f-c7c8-4e1e-b7fa-a0e4b9a6c2a8@gmail.com/
Paul E. McKenney Feb. 27, 2024, 12:21 a.m. UTC | #2 
On Sat, Feb 24, 2024 at 05:30:29PM +0100, Oleksandr Natalenko wrote:
> Hello.
> 
> On čtvrtek 1. února 2024 2:40:57 CET Boqun Feng wrote:
> > From: Zqiang <qiang.zhang1211@gmail.com>
> > 
> > For the kernels built with CONFIG_RCU_NOCB_CPU_DEFAULT_ALL=y and
> > CONFIG_RCU_LAZY=y, the following scenarios will trigger WARN_ON_ONCE()
> > in the rcu_nocb_bypass_lock() and rcu_nocb_wait_contended() functions:
> > 
> >         CPU2                                               CPU11
> > kthread
> > rcu_nocb_cb_kthread                                       ksys_write
> > rcu_do_batch                                              vfs_write
> > rcu_torture_timer_cb                                      proc_sys_write
> > __kmem_cache_free                                         proc_sys_call_handler
> > kmemleak_free                                             drop_caches_sysctl_handler
> > delete_object_full                                        drop_slab
> > __delete_object                                           shrink_slab
> > put_object                                                lazy_rcu_shrink_scan
> > call_rcu                                                  rcu_nocb_flush_bypass
> > __call_rcu_commn                                            rcu_nocb_bypass_lock
> >                                                             raw_spin_trylock(&rdp->nocb_bypass_lock) fail
> >                                                             atomic_inc(&rdp->nocb_lock_contended);
> > rcu_nocb_wait_contended                                     WARN_ON_ONCE(smp_processor_id() != rdp->cpu);
> >  WARN_ON_ONCE(atomic_read(&rdp->nocb_lock_contended))                                          |
> >                             |_ _ _ _ _ _ _ _ _ _same rdp and rdp->cpu != 11_ _ _ _ _ _ _ _ _ __|
> > 
> > Reproduce this bug with "echo 3 > /proc/sys/vm/drop_caches".
> > 
> > This commit therefore uses rcu_nocb_try_flush_bypass() instead of
> > rcu_nocb_flush_bypass() in lazy_rcu_shrink_scan().  If the nocb_bypass
> > queue is being flushed, then rcu_nocb_try_flush_bypass will return
> > directly.
> > 
> > Signed-off-by: Zqiang <qiang.zhang1211@gmail.com>
> > Reviewed-by: Joel Fernandes (Google) <joel@joelfernandes.org>
> > Reviewed-by: Frederic Weisbecker <frederic@kernel.org>
> > Reviewed-by: Paul E. McKenney <paulmck@kernel.org>
> > Signed-off-by: Boqun Feng <boqun.feng@gmail.com>
> > ---
> >  kernel/rcu/tree_nocb.h | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> > 
> > diff --git a/kernel/rcu/tree_nocb.h b/kernel/rcu/tree_nocb.h
> > index 9e8052ba14b9..ffa69a5e18f4 100644
> > --- a/kernel/rcu/tree_nocb.h
> > +++ b/kernel/rcu/tree_nocb.h
> > @@ -1391,7 +1391,7 @@ lazy_rcu_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
> >  			rcu_nocb_unlock_irqrestore(rdp, flags);
> >  			continue;
> >  		}
> > -		WARN_ON_ONCE(!rcu_nocb_flush_bypass(rdp, NULL, jiffies, false));
> > +		rcu_nocb_try_flush_bypass(rdp, jiffies);
> >  		rcu_nocb_unlock_irqrestore(rdp, flags);
> >  		wake_nocb_gp(rdp, false);
> >  		sc->nr_to_scan -= _count;
> > 
> 
> Does this fix [1] [2]?
> 
> Thank you.
> 
> [1] https://bugzilla.kernel.org/show_bug.cgi?id=217948
> [2] https://lore.kernel.org/lkml/8461340f-c7c8-4e1e-b7fa-a0e4b9a6c2a8@gmail.com/

It might, but why not apply it to the exact kernel version on which the
bug appeared and see if it prevents it?

							Thanx, Paul
diff mbox series

Patch

diff --git a/kernel/rcu/tree_nocb.h b/kernel/rcu/tree_nocb.h
index 9e8052ba14b9..ffa69a5e18f4 100644
--- a/kernel/rcu/tree_nocb.h
+++ b/kernel/rcu/tree_nocb.h
@@ -1391,7 +1391,7 @@  lazy_rcu_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
 			rcu_nocb_unlock_irqrestore(rdp, flags);
 			continue;
 		}
-		WARN_ON_ONCE(!rcu_nocb_flush_bypass(rdp, NULL, jiffies, false));
+		rcu_nocb_try_flush_bypass(rdp, jiffies);
 		rcu_nocb_unlock_irqrestore(rdp, flags);
 		wake_nocb_gp(rdp, false);
 		sc->nr_to_scan -= _count;