Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <1602510644-24536-1-git-send-email-kernelfans@gmail.com> <20201014130223.GF2628@hirez.programming.kicks-ass.net>
In-Reply-To: <20201014130223.GF2628@hirez.programming.kicks-ass.net>
From:   Pingfan Liu <kernelfans@gmail.com>
Date:   Thu, 15 Oct 2020 09:27:44 +0800
Message-ID: <CAFgQCTvrEdk6WOEpJOX1_Q64PmjvpHq2qrWo_1d_4NBnus+7Fw@mail.gmail.com>
Subject: Re: [PATCH] sched/cputime: correct account of irqtime
To:     Peter Zijlstra <peterz@infradead.org>
Cc:     LKML <linux-kernel@vger.kernel.org>,
        Ingo Molnar <mingo@redhat.com>,
        Juri Lelli <juri.lelli@redhat.com>,
        Vincent Guittot <vincent.guittot@linaro.org>,
        Dietmar Eggemann <dietmar.eggemann@arm.com>,
        Steven Rostedt <rostedt@goodmis.org>,
        Ben Segall <bsegall@google.com>, Mel Gorman <mgorman@suse.de>,
        Thomas Gleixner <tglx@linutronix.de>,
        Andy Lutomirski <luto@kernel.org>,
        Will Deacon <will@kernel.org>,
        "Paul E. McKenney" <paulmck@kernel.org>,
        Frederic Weisbecker <frederic@kernel.org>,
        Allen Pais <allen.lkml@gmail.com>,
        Romain Perier <romain.perier@gmail.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Oct 14, 2020 at 9:02 PM Peter Zijlstra <peterz@infradead.org> wrote:
>
> On Mon, Oct 12, 2020 at 09:50:44PM +0800, Pingfan Liu wrote:
> > __do_softirq() may be interrupted by hardware interrupts. In this case,
> > irqtime_account_irq() will account the time slice as CPUTIME_SOFTIRQ by
> > mistake.
> >
> > By passing irqtime_account_irq() an extra param about either hardirq or
> > softirq, irqtime_account_irq() can handle the above case.
>
> I'm not sure I see the scenario in which it goes wrong.
>
> irqtime_account_irq() is designed such that we're called with the old
> preempt_count on enter and the new preempt_count on exit. This way we'll
> accumuate the delta to the previous context.
>
Oops! You are right, the time delta between a softirq and a
interrupting hardirq should be accounted into the softrq.

Thanks for your clear explanation.

Regards,
Pingfan