Re: [PATCH 2/6] sched/vtime: Bring all-in-one kcpustat accessor for vtime fields

[Ingo Molnar <mingo@kernel.org>]

* Frederic Weisbecker <frederic@kernel.org> wrote:

> Many callsites want to fetch the values of system, user, user_nice, guest
> or guest_nice kcpustat fields altogether or at least a pair of these.
> 
> In that case calling kcpustat_field() for each requested field brings
> unecessary overhead when we could fetch all of them in a row.
> 
> So provide kcpustat_cputime() that fetches all vtime sensitive fields
> under the same RCU and seqcount block.
> 
> Signed-off-by: Frederic Weisbecker <frederic@kernel.org>
> Cc: Yauheni Kaliuta <yauheni.kaliuta@redhat.com>
> Cc: Thomas Gleixner <tglx@linutronix.de>
> Cc: Rik van Riel <riel@surriel.com>
> Cc: Peter Zijlstra <peterz@infradead.org>
> Cc: Wanpeng Li <wanpengli@tencent.com>
> Cc: Ingo Molnar <mingo@kernel.org>
> ---
>  include/linux/kernel_stat.h |  23 ++++++
>  kernel/sched/cputime.c      | 139 ++++++++++++++++++++++++++++++------
>  2 files changed, 142 insertions(+), 20 deletions(-)
> 
> diff --git a/include/linux/kernel_stat.h b/include/linux/kernel_stat.h
> index 79781196eb25..6bd70e464c61 100644
> --- a/include/linux/kernel_stat.h
> +++ b/include/linux/kernel_stat.h
> @@ -78,15 +78,38 @@ static inline unsigned int kstat_cpu_irqs_sum(unsigned int cpu)
>  	return kstat_cpu(cpu).irqs_sum;
>  }
>  
> +
> +static inline void kcpustat_cputime_raw(u64 *cpustat, u64 *user, u64 *nice,
> +					u64 *system, u64 *guest, u64 *guest_nice)
> +{
> +	*user = cpustat[CPUTIME_USER];
> +	*nice = cpustat[CPUTIME_NICE];
> +	*system = cpustat[CPUTIME_SYSTEM];
> +	*guest = cpustat[CPUTIME_GUEST];
> +	*guest_nice = cpustat[CPUTIME_GUEST_NICE];

Could the 'cpustat' pointer be constified?

Also, please:

> +	*user	    = cpustat[CPUTIME_USER];
> +	*nice	    = cpustat[CPUTIME_NICE];
> +	*system	    = cpustat[CPUTIME_SYSTEM];
> +	*guest	    = cpustat[CPUTIME_GUEST];
> +	*guest_nice = cpustat[CPUTIME_GUEST_NICE];

More pleasing to look at and easier to verify as well.

> +static int vtime_state_check(struct vtime *vtime, int cpu)
> +{
> +	/*
> +	 * We raced against context switch, fetch the
> +	 * kcpustat task again.
> +	 */

s/against context switch
 /against a context switch

> +void kcpustat_cputime(struct kernel_cpustat *kcpustat, int cpu,
> +		      u64 *user, u64 *nice, u64 *system,
> +		      u64 *guest, u64 *guest_nice)
> +{
> +	u64 *cpustat = kcpustat->cpustat;
> +	struct rq *rq;
> +	int err;
> +
> +	if (!vtime_accounting_enabled_cpu(cpu)) {
> +		kcpustat_cputime_raw(cpustat, user, nice,
> +				     system, guest, guest_nice);
> +		return;
> +	}
> +
> +	rq = cpu_rq(cpu);
> +
> +	for (;;) {
> +		struct task_struct *curr;
> +
> +		rcu_read_lock();
> +		curr = rcu_dereference(rq->curr);
> +		if (WARN_ON_ONCE(!curr)) {
> +			rcu_read_unlock();
> +			kcpustat_cputime_raw(cpustat, user, nice,
> +					     system, guest, guest_nice);
> +			return;
> +		}
> +
> +		err = kcpustat_cputime_vtime(cpustat, curr, cpu, user,
> +					     nice, system, guest, guest_nice);
> +		rcu_read_unlock();
> +
> +		if (!err)
> +			return;
> +
> +		cpu_relax();
> +	}
> +}
> +EXPORT_SYMBOL_GPL(kcpustat_cputime);

I'm wondering whether it's worth introducing a helper structure for this 
train of parameters: user, nice, system, guest, guest_nice?

We also have similar constructs in other places:

+               u64 cpu_user, cpu_nice, cpu_sys, cpu_guest, cpu_guest_nice;

But more broadly, what do we gain by passing along a quartet of pointers, 
while we could also just use a 'struct kernel_cpustat' and store the 
values there naturally?

Yes, it's larger, because it also has 5 other fields - but we lose much 
of the space savings due to always passing along the 4 pointers already.

So I really think the parameter passing should be organized better here. 
This probably affects similar cpustat functions as well.

Thanks,

	Ingo