linux-kernel - Re: [entry] 47b8ff194c: will-it-scale.per_process

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 13 May 2021 10:46:36 -0700
From:   "Paul E. McKenney" <paulmck@...nel.org>
To:     Peter Zijlstra <peterz@...radead.org>
Cc:     kernel test robot <oliver.sang@...el.com>,
        Frederic Weisbecker <frederic@...nel.org>,
        Ingo Molnar <mingo@...nel.org>,
        LKML <linux-kernel@...r.kernel.org>, lkp@...ts.01.org,
        lkp@...el.com, ying.huang@...el.com, feng.tang@...el.com,
        zhengjun.xing@...el.com
Subject: Re: [entry]  47b8ff194c:  will-it-scale.per_process_ops -3.0%
 regression

On Thu, May 13, 2021 at 10:19:21AM +0200, Peter Zijlstra wrote:
> On Wed, Apr 28, 2021 at 03:16:53PM +0800, kernel test robot wrote:
> > 
> > 
> > Greeting,
> > 
> > FYI, we noticed a -3.0% regression of will-it-scale.per_process_ops due to commit:
> > 
> > 
> > commit: 47b8ff194c1fd73d58dc339b597d466fe48c8958 ("entry: Explicitly flush pending rcuog wakeup before last rescheduling point")
> 
> So the RCU bits are in rcu_user_enter(), which is called from
> __context_tracking_enter() aka user_enter(), which is under
> context_tracking_enabled().
> 
> But the new code in entry is not; we now unconditionally call
> rcu_nocb_flush_deferred_wakeup(). Did that want to be under
> context_tracking_enabled() as well?
> 
> Frederic, Paul?

My argument in favor of the change below is that if there is no context
tracking, then scheduling-clock interrupts will happen on all non-idle
CPUs.  The next scheduling-clock interrupt will check this deferred-wakeup
state, and if set, cause rcu_core() to be invoked (for example, within the
RCU_SOFTIRQ handler).  And rcu_core() invokes do_nocb_deferred_wakeup(),
which takes care of this.

For idle CPUs, do_idle() invokes rcu_nocb_flush_deferred_wakeup().

Frederic, any cases that I am missing?

							Thanx, Paul

> ---
> 
> diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h
> index 9455476c5ba2..f4df001410fc 100644
> --- a/include/linux/rcupdate.h
> +++ b/include/linux/rcupdate.h
> @@ -114,7 +114,12 @@ static inline void rcu_user_exit(void) { }
>  void rcu_init_nohz(void);
>  int rcu_nocb_cpu_offload(int cpu);
>  int rcu_nocb_cpu_deoffload(int cpu);
> -void rcu_nocb_flush_deferred_wakeup(void);
> +void __rcu_nocb_flush_deferred_wakeup(void);
> +static inline void rcu_nocb_flush_deferred_wakeup(void)
> +{
> +	if (context_tracking_enabled())
> +		__rcu_nocb_flush_deferred_wakeup();
> +}
>  #else /* #ifdef CONFIG_RCU_NOCB_CPU */
>  static inline void rcu_init_nohz(void) { }
>  static inline int rcu_nocb_cpu_offload(int cpu) { return -EINVAL; }
> diff --git a/kernel/rcu/tree_plugin.h b/kernel/rcu/tree_plugin.h
> index ad0156b86937..3cdbbf7fba01 100644
> --- a/kernel/rcu/tree_plugin.h
> +++ b/kernel/rcu/tree_plugin.h
> @@ -2378,7 +2378,7 @@ static bool do_nocb_deferred_wakeup(struct rcu_data *rdp)
>  	return false;
>  }
>  
> -void rcu_nocb_flush_deferred_wakeup(void)
> +void __rcu_nocb_flush_deferred_wakeup(void)
>  {
>  	do_nocb_deferred_wakeup(this_cpu_ptr(&rcu_data));
>  }
>