Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 10 Jan 2023 13:12:09 +0100
From:   Frederic Weisbecker <frederic@kernel.org>
To:     Christoph Lameter <cl@gentwo.de>
Cc:     Marcelo Tosatti <mtosatti@redhat.com>, atomlin@atomlin.com,
        tglx@linutronix.de, mingo@kernel.org, peterz@infradead.org,
        pauld@redhat.com, neelx@redhat.com, oleksandr@natalenko.name,
        linux-kernel@vger.kernel.org, linux-mm@kvack.org
Subject: Re: [PATCH v13 1/6] mm/vmstat: Add CPU-specific variable to track a
 vmstat discrepancy
Message-ID: <Y71WGeT2+6s4ewfv@lothringen>
References: <20230105125218.031928326@redhat.com>
 <20230105125248.772766288@redhat.com>
 <e9673684-ef3-4070-18bd-2f20fbfe8d5@gentwo.de>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <e9673684-ef3-4070-18bd-2f20fbfe8d5@gentwo.de>
Precedence: bulk

On Tue, Jan 10, 2023 at 12:58:32PM +0100, Christoph Lameter wrote:
> On Thu, 5 Jan 2023, Marcelo Tosatti wrote:
> 
> > +static inline void vmstat_mark_dirty(void)
> > +{
> > +	this_cpu_write(vmstat_dirty, true);
> > +}
> 
> this_cpu_write() is intended for an per cpu atomic context. You are not
> using it in that way. The processor may have changed before or after and
> thus vmstat_dirty for another CPU may  have been marked dirty.
> 
> I guess this would have to be called __vmstat_mark_dirty() and be using
> __this_cpu_write(*) with a requirement that preemption be disabled before
> using this function.

You're right. So this patchset also arranges for these vmstat functions to be
called with preemption disabled. I'm converting the this_cpu operations
to __this_cpu versions to make sure of that. And I believe the __this_cpu
WARN if preemptible().


> 
> > +static inline void vmstat_clear_dirty(void)
> > +{
> > +	this_cpu_write(vmstat_dirty, false);
> > +}
> 
> Same
> 
> > +static inline bool is_vmstat_dirty(void)
> > +{
> > +	return this_cpu_read(vmstat_dirty);
> > +}
> 
> This function would only work correctly if preemption is disabled.
> Otherwise the processor may change.

Indeed that should apply as __this_cpu_read() as well.

Thanks!