Message-ID: <20240616223616.im3tlh6jheadlhnz@airbuntu>
Date: Sun, 16 Jun 2024 23:36:16 +0100
From: Qais Yousef <qyousef@...alina.io>
To: Peter Zijlstra <peterz@...radead.org>
Cc: John Stultz <jstultz@...gle.com>, LKML <linux-kernel@...r.kernel.org>,
	Thomas Gleixner <tglx@...utronix.de>,
	Frederic Weisbecker <frederic@...nel.org>,
	Ingo Molnar <mingo@...hat.com>, Juri Lelli <juri.lelli@...hat.com>,
	Vincent Guittot <vincent.guittot@...aro.org>,
	Dietmar Eggemann <dietmar.eggemann@....com>,
	Steven Rostedt <rostedt@...dmis.org>,
	Ben Segall <bsegall@...gle.com>, Mel Gorman <mgorman@...e.de>,
	Daniel Bristot de Oliveira <bristot@...hat.com>,
	Valentin Schneider <vschneid@...hat.com>,
	Joel Fernandes <joel@...lfernandes.org>, kernel-team@...roid.com
Subject: Re: [PATCH] RFC: sched: Rework task_sched_runtime to avoid calling
 update_rq_clock

On 06/14/24 11:48, Peter Zijlstra wrote:
> On Thu, Jun 13, 2024 at 12:51:42PM +0100, Qais Yousef wrote:
> > On 06/13/24 12:04, Peter Zijlstra wrote:
> > 
> > > ---
> > > diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> > > index 0935f9d4bb7b..d4b87539d72a 100644
> > > --- a/kernel/sched/core.c
> > > +++ b/kernel/sched/core.c
> > > @@ -724,7 +724,6 @@ static void update_rq_clock_task(struct rq *rq, s64 delta)
> > >  
> > >  	rq->prev_irq_time += irq_delta;
> > >  	delta -= irq_delta;
> > > -	psi_account_irqtime(rq->curr, irq_delta);
> > >  	delayacct_irq(rq->curr, irq_delta);
> > >  #endif
> > >  #ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING
> > > @@ -5459,6 +5458,8 @@ void sched_tick(void)
> > >  
> > >  	sched_clock_tick();
> > >  
> > > +	psi_account_irqtime(curr, &rq->psi_irq_time);
> > > +
> > 
> > If wakeup preemption causes a context switch, wouldn't we lose this
> > information then? I *think* active migration might cause this information to be
> > lost too.
> 
> I'm not sure what would be lost ?! the accounting is per cpu, not per
> task afaict. That said,...

Yes sorry I mixed this with per task accounting..

> 
> > pick_next_task() might be a better place to do the accounting?
> 
> Additionally, when there has been an effective cgroup switch. Only on
> switch doesn't work for long running tasks, then the PSI information
> will be artitrarily long out of date.
> 
> Which then gets me something like the (completely untested) below..
> 
> Hmm?
> 
> diff --git a/kernel/sched/core.c b/kernel/sched/core.c
> index 0935f9d4bb7b..36aed99d6a6c 100644
> --- a/kernel/sched/core.c
> +++ b/kernel/sched/core.c
> @@ -724,7 +724,6 @@ static void update_rq_clock_task(struct rq *rq, s64 delta)
>  
>  	rq->prev_irq_time += irq_delta;
>  	delta -= irq_delta;
> -	psi_account_irqtime(rq->curr, irq_delta);
>  	delayacct_irq(rq->curr, irq_delta);
>  #endif
>  #ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING
> @@ -5459,6 +5458,8 @@ void sched_tick(void)
>  
>  	sched_clock_tick();
>  
> +	psi_account_irqtime(curr, NULL, &rq->psi_irq_time);
> +
>  	rq_lock(rq, &rf);
>  
>  	update_rq_clock(rq);
> @@ -6521,6 +6524,7 @@ static void __sched notrace __schedule(unsigned int sched_mode)
>  		++*switch_count;
>  
>  		migrate_disable_switch(rq, prev);
> +		psi_account_irqtime(prev, next, &rq->psi_irq_time);

Hmm are prev and next swapped here? next == curr in my view if there's no
subtly I missed

>  		psi_sched_switch(prev, next, !task_on_rq_queued(prev));
>  
>  		trace_sched_switch(sched_mode & SM_MASK_PREEMPT, prev, next, prev_state);
> diff --git a/kernel/sched/psi.c b/kernel/sched/psi.c
> index 146baa91d104..65bba162408f 100644
> --- a/kernel/sched/psi.c
> +++ b/kernel/sched/psi.c
> @@ -991,22 +991,31 @@ void psi_task_switch(struct task_struct *prev, struct task_struct *next,
>  }
>  
>  #ifdef CONFIG_IRQ_TIME_ACCOUNTING
> -void psi_account_irqtime(struct task_struct *task, u32 delta)
> +void psi_account_irqtime(struct task_struct *curr, struct task_struct *prev, u64 *time)
>  {
> -	int cpu = task_cpu(task);
> +	int cpu = task_cpu(curr);
>  	struct psi_group *group;
>  	struct psi_group_cpu *groupc;
> -	u64 now;
> +	u64 now, irq;
> +	s64 delta;
>  
>  	if (static_branch_likely(&psi_disabled))
>  		return;
>  
> -	if (!task->pid)
> +	if (!curr->pid)
> +		return;
> +
> +	group = task_psi_group(curr);
> +	if( prev && task_psi_group(prev) == group)

nit: whitespace misplaced

LGTM otherwise.

Reviewed-by: Qais Yousef <qyousef@...alina.io>

>  		return;
>  
>  	now = cpu_clock(cpu);
> +	irq = irq_time_read(cpu);
> +	delta = (s64)(irq - *time);
> +	if (delta < 0)
> +		return;
> +	*time = irq;
>  
> -	group = task_psi_group(task);
>  	do {
>  		if (!group->enabled)
>  			continue;
> diff --git a/kernel/sched/sched.h b/kernel/sched/sched.h
> index 62fd8bc6fd08..a63eb546bc4a 100644
> --- a/kernel/sched/sched.h
> +++ b/kernel/sched/sched.h
> @@ -1133,6 +1133,7 @@ struct rq {
>  
>  #ifdef CONFIG_IRQ_TIME_ACCOUNTING
>  	u64			prev_irq_time;
> +	u64			psi_irq_time;
>  #endif
>  #ifdef CONFIG_PARAVIRT
>  	u64			prev_steal_time;
> diff --git a/kernel/sched/stats.h b/kernel/sched/stats.h
> index d1445410840a..1e290054c5db 100644
> --- a/kernel/sched/stats.h
> +++ b/kernel/sched/stats.h
> @@ -110,7 +110,7 @@ __schedstats_from_se(struct sched_entity *se)
>  void psi_task_change(struct task_struct *task, int clear, int set);
>  void psi_task_switch(struct task_struct *prev, struct task_struct *next,
>  		     bool sleep);
> -void psi_account_irqtime(struct task_struct *task, u32 delta);
> +void psi_account_irqtime(struct task_struct *curr, struct task_struct *prev, u64 *time);
>  
>  /*
>   * PSI tracks state that persists across sleeps, such as iowaits and
> @@ -192,7 +192,7 @@ static inline void psi_ttwu_dequeue(struct task_struct *p) {}
>  static inline void psi_sched_switch(struct task_struct *prev,
>  				    struct task_struct *next,
>  				    bool sleep) {}
> -static inline void psi_account_irqtime(struct task_struct *task, u32 delta) {}
> +static inline void psi_account_irqtime(struct task_struct *curr, struct task_struct *prev, u64 *time) {}
>  #endif /* CONFIG_PSI */
>  
>  #ifdef CONFIG_SCHED_INFO
> 
> 
> 

Powered by blists - more mailing lists