Re: [PATCH] mailbox: imx: Support runtime PM

[Marc Kleine-Budde]

On 4/13/20 2:25 PM, Anson Huang wrote:
> Some power hungry sub-systems like VPU has its own MUs which also
> use mailbox driver, current mailbox driver uses platform driver
> model and MU's power will be ON after driver probed and left ON
> there, it may cause the whole sub-system can NOT enter lower power
> mode, take VPU driver for example, it has runtime PM support, but
> due to its MU always ON, the VPU sub-system will be always ON and
> consume many power during kernel idle.
> 
> To save power in kernel idle, mailbox driver needs to support
> runtime PM in order to power off MU when it is unused. However,
> the runtime suspend/resume can ONLY be implemented in mailbox's
> .shutdown/.startup callback, so its consumer needs to call
> mbox_request_channel()/mbox_free_channel() in consumer driver's
> runtime PM callback, then the MU's power will be ON/OFF along with
> consumer's runtime PM status.
> 
> Signed-off-by: Anson Huang <Anson.Huang@xxxxxxx>
> ---
>  drivers/mailbox/imx-mailbox.c | 27 ++++++++++++++++++++++++++-
>  1 file changed, 26 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/mailbox/imx-mailbox.c b/drivers/mailbox/imx-mailbox.c
> index 7906624..97bf0ac 100644
> --- a/drivers/mailbox/imx-mailbox.c
> +++ b/drivers/mailbox/imx-mailbox.c
> @@ -12,6 +12,7 @@
>  #include <linux/mailbox_controller.h>
>  #include <linux/module.h>
>  #include <linux/of_device.h>
> +#include <linux/pm_runtime.h>
>  #include <linux/slab.h>
>  
>  #define IMX_MU_xSR_GIPn(x)	BIT(28 + (3 - (x)))
> @@ -287,6 +288,7 @@ static int imx_mu_startup(struct mbox_chan *chan)
>  	struct imx_mu_con_priv *cp = chan->con_priv;
>  	int ret;
>  
> +	pm_runtime_get_sync(priv->dev);
>  	if (cp->type == IMX_MU_TYPE_TXDB) {
>  		/* Tx doorbell don't have ACK support */
>  		tasklet_init(&cp->txdb_tasklet, imx_mu_txdb_tasklet,
> @@ -323,6 +325,7 @@ static void imx_mu_shutdown(struct mbox_chan *chan)
>  
>  	if (cp->type == IMX_MU_TYPE_TXDB) {
>  		tasklet_kill(&cp->txdb_tasklet);
> +		pm_runtime_put_sync(priv->dev);
>  		return;
>  	}
>  
> @@ -341,6 +344,7 @@ static void imx_mu_shutdown(struct mbox_chan *chan)
>  	}
>  
>  	free_irq(priv->irq, chan);
> +	pm_runtime_put_sync(priv->dev);
>  }
>  
>  static const struct mbox_chan_ops imx_mu_ops = {
> @@ -508,7 +512,27 @@ static int imx_mu_probe(struct platform_device *pdev)
>  
>  	platform_set_drvdata(pdev, priv);
>  
> -	return devm_mbox_controller_register(dev, &priv->mbox);
> +	ret = devm_mbox_controller_register(dev, &priv->mbox);
> +	if (ret)
> +		return ret;
> +
> +	pm_runtime_enable(dev);
> +

First registering at the system and then setting up the power management
looks racy. Don't know if this is serialized by some other means.

> +	ret = pm_runtime_get_sync(dev);
> +	if (ret < 0) {
> +		pm_runtime_put_noidle(dev);
> +		goto disable_runtime_pm;
> +	}
> +
> +	ret = pm_runtime_put_sync(dev);
> +	if (ret < 0)
> +		goto disable_runtime_pm;
> +
> +	return 0;

Marc

-- 
Pengutronix e.K.                 | Marc Kleine-Budde           |
Embedded Linux                   | https://www.pengutronix.de  |
Vertretung West/Dortmund         | Phone: +49-231-2826-924     |
Amtsgericht Hildesheim, HRA 2686 | Fax:   +49-5121-206917-5555 |