From mboxrd@z Thu Jan  1 00:00:00 1970
From: Simon Horman <horms@verge.net.au>
Date: Mon, 28 Apr 2014 00:10:59 +0000
Subject: Re: [PATCH] [RFC] ARM: shmobile: Enable shmobile Runtime PM from MSTP clock driver
Message-Id: <20140428001058.GM32087@verge.net.au>
List-Id: <linux-sh.vger.kernel.org>
References: <1398445124-16828-1-git-send-email-geert+renesas@glider.be>
In-Reply-To: <1398445124-16828-1-git-send-email-geert+renesas@glider.be>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
To: linux-sh@vger.kernel.org

On Fri, Apr 25, 2014 at 06:58:44PM +0200, Geert Uytterhoeven wrote:
> On shmobile multi-platform using the Common Clock Framework, enable
> shmobile Runtime PM from the MSTP clock driver, to make sure the bus clocks
> are enabled for all devices.
> 
> On ARCH_SHMOBILE_LEGACY, this is done unconditionally, from the shmobile
> Runtime PM code itself.
> On ARCH_SHMOBILE_MULTI, this was missing, leading to disabled clocks,
> depending on implicit reset state or on the bootloader.
> 
> The external declaration of sh_pm_runtime_enable() in clk-mstp.c is
> needed because arch/arm/mach-shmobile/include/mach/common.h can't be
> included from driver code.
> 
> More calls to sh_pm_runtime_enable() will have to be added to other gate
> clock drivers, when the corresponding SoC platform code is converted to
> multi-platform and CCF.
> 
> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> ---
> This applies on top of patches 4-7 (1-3 have been applied by Simon) of
> series "[PATCH v2 00/17] ARM: shmobile: Enable drivers/sh/pm_runtime.c on
> multi-platform]".
> 
>  drivers/clk/shmobile/clk-mstp.c |    3 +++
>  drivers/sh/pm_runtime.c         |   20 ++++++++++++--------
>  2 files changed, 15 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/clk/shmobile/clk-mstp.c b/drivers/clk/shmobile/clk-mstp.c
> index 2e3c08fff173..da2e6dcf8fed 100644
> --- a/drivers/clk/shmobile/clk-mstp.c
> +++ b/drivers/clk/shmobile/clk-mstp.c
> @@ -161,6 +161,9 @@ static void __init cpg_mstp_clocks_init(struct device_node *np)
>  	struct mstp_clock_group *group;
>  	struct clk **clks;
>  	unsigned int i;
> +	extern void sh_pm_runtime_enable(void);
> +
> +	sh_pm_runtime_enable();
>  
>  	group = kzalloc(sizeof(*group), GFP_KERNEL);
>  	clks = kmalloc(MSTP_MAX_CLOCKS * sizeof(*clks), GFP_KERNEL);
> diff --git a/drivers/sh/pm_runtime.c b/drivers/sh/pm_runtime.c
> index 3ec047ac1ff4..59d8b5fd558e 100644
> --- a/drivers/sh/pm_runtime.c
> +++ b/drivers/sh/pm_runtime.c
> @@ -75,20 +75,24 @@ static struct pm_clk_notifier_block platform_bus_notifier = {
>  	.con_ids = { NULL, },
>  };
>  
> -static bool default_pm_on;
> +#if defined(CONFIG_SUPERH) || defined(CONFIG_ARCH_SHMOBILE_LEGACY)
> +bool default_pm_on = true;
> +#else
> +bool default_pm_on;
> +#endif

I wonder if there is some way we can get the compiler
to optimise out code in the case of default_pm_on = false,
perhaps by using IS_ENABLED() in conjunction with a Kconfig variable.

> -int __init sh_pm_runtime_init(void)
> +void __init sh_pm_runtime_enable(void)
>  {
> -	if (default_pm_on)
> -		return 0;
> -
>  	default_pm_on = true;
> -	pm_clk_add_notifier(&platform_bus_type, &platform_bus_notifier);
> +}
> +
> +static int __init sh_pm_runtime_init(void)
> +{
> +	if (default_pm_on)
> +		pm_clk_add_notifier(&platform_bus_type, &platform_bus_notifier);
>  	return 0;
>  }
> -#if defined(CONFIG_SUPERH) || defined(CONFIG_ARCH_SHMOBILE_LEGACY)
>  core_initcall(sh_pm_runtime_init);
> -#endif
>  
>  static int __init sh_pm_runtime_late_init(void)
>  {
> -- 
> 1.7.9.5
>