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;
Date:   Mon, 27 Jul 2020 13:24:54 +0200
From:   Ingo Molnar <mingo@kernel.org>
To:     vincent.donnefort@arm.com
Cc:     mingo@redhat.com, peterz@infradead.org, vincent.guittot@linaro.org,
        linux-kernel@vger.kernel.org, dietmar.eggemann@arm.com,
        lukasz.luba@arm.com, valentin.schneider@arm.com
Subject: Re: [PATCH] sched/fair: provide u64 read for 32-bits arch helper
Message-ID: <20200727112454.GB55660@gmail.com>
References: <1595847564-239957-1-git-send-email-vincent.donnefort@arm.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <1595847564-239957-1-git-send-email-vincent.donnefort@arm.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk


* vincent.donnefort@arm.com <vincent.donnefort@arm.com> wrote:

> From: Vincent Donnefort <vincent.donnefort@arm.com>
> 
> Introducing two macro helpers u64_32read() and u64_32read_set_copy() to
> factorize the u64 vminruntime and last_update_time read on a 32-bits
> architecture. Those new helpers encapsulate smp_rmb() and smp_wmb()
> synchronization and therefore, have a small penalty in set_task_rq_fair()
> and init_cfs_rq().
> 
> The choice of using a macro over an inline function is driven by the
> conditional u64 variable copy declarations.

Could you please explain how "conditional u64 variable copy 
declarations" prevents us to use an inline function for this:

> +/*
> + * u64_32read() / u64_32read_set_copy()
> + *
> + * Use the copied u64 value to protect against data race. This is only
> + * applicable for 32-bits architectures.
> + */
> +#if !defined(CONFIG_64BIT) && defined(CONFIG_SMP)
> +# define u64_32read(val, copy)						\
> +({									\
> +	u64 _val;							\
> +	u64 _val_copy;							\
> +									\
> +	do {								\
> +		_val_copy = copy;					\
> +		/*							\
> +		 * paired with u64_32read_set_copy, ordering access	\
> +		 * to val and copy.					\
> +		 */							\
> +		smp_rmb();						\
> +		_val = val;						\
> +	} while (_val != _val_copy);					\
> +									\
> +	_val;								\
> +})
> +# define u64_32read_set_copy(val, copy)					\
> +do {									\
> +	/* paired with u64_32read, ordering access to val and copy */	\
> +	smp_wmb();							\
> +	copy = val;							\
> +} while (0)
> +#else
> +# define u64_32read(val, copy) (val)
> +# define u64_32read_set_copy(val, copy) do { } while (0)
> +#endif
> +

Thanks,

	Ingo