Date: Tue, 13 Dec 2016 15:38:20 +0100
From: Frederic Weisbecker <fweisbec@gmail.com>
To: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: LKML <linux-kernel@vger.kernel.org>, Tony Luck <tony.luck@intel.com>,
        Wanpeng Li <wanpeng.li@hotmail.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Michael Ellerman <mpe@ellerman.id.au>,
        Heiko Carstens <heiko.carstens@de.ibm.com>,
        Benjamin Herrenschmidt <benh@kernel.crashing.org>,
        Thomas Gleixner <tglx@linutronix.de>,
        Paul Mackerras <paulus@samba.org>, Ingo Molnar <mingo@kernel.org>,
        Fenghua Yu <fenghua.yu@intel.com>, Rik van Riel <riel@redhat.com>,
        Stanislaw Gruszka <sgruszka@redhat.com>
Subject: Re: [PATCH 09/10] s390/cputime: delayed accounting of system time
Message-ID: <20161213143816.GA1272@lerouge>
References: <1480991543-6557-1-git-send-email-fweisbec@gmail.com>
 <1480991543-6557-10-git-send-email-fweisbec@gmail.com>
 <20161210014804.GA3023@lerouge>
 <20161212112754.5ad104cf@mschwideX1>
 <20161212150228.GA17032@lerouge>
 <20161213121322.5fdbbb28@mschwideX1>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20161213121322.5fdbbb28@mschwideX1>
User-Agent: Mutt/1.5.24 (2015-08-30)
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1325
Lines: 26

On Tue, Dec 13, 2016 at 12:13:22PM +0100, Martin Schwidefsky wrote:
> On Mon, 12 Dec 2016 16:02:30 +0100
> Frederic Weisbecker <fweisbec@gmail.com> wrote:
> 
> > On Mon, Dec 12, 2016 at 11:27:54AM +0100, Martin Schwidefsky wrote:
> > > 3) The call to vtime_flush in account_process_tick is done in irq context from
> > >    update_process_times. hardirq_offset==1 is also correct.  
> > 
> > Let's see this for example:
> > 
> > +       if ((tsk->flags & PF_VCPU) && (irq_count() - hardirq_offset == 0))
> > +               S390_lowcore.guest_timer += timer;
> > 
> > If the tick is interrupting guest, we have accounted the guest time on tick IRQ entry.
> > Now we are in the middle of the tick interrupt and since hardirq_offset is 1, we
> > are taking the above path by accounting half of the tick-IRQ time as guest, which is wrong,
> > it's actually IRQ time.
> 
> Hmm, you got me there. The system time from irq_enter until account_process_tick
> is reached is indeed IRQ time. It is not much but it is incorrect. The best fix
> would be to rip out the accounting of the system time from account_process_tick
> as irq_enter / irq_exit will do system time accounting anyway. To do that
> do_account_vtime needs to be split, because for the task switch we need to
> account the system time of the previous task.

Exactly!