Re: [PATCH 01/11] Platform: add a dev_groups pointer to struct platform_driver

[Johan Hovold]

On Thu, Jul 04, 2019 at 10:46:07AM +0200, Greg Kroah-Hartman wrote:
> Platform drivers like to add sysfs groups to their device, but right now
> they have to do it "by hand".  The driver core should handle this for
> them, but there is no way to get to the bus-default attribute groups as
> all platform devices are "special and unique" one-off drivers/devices.
> 
> To combat this, add a dev_groups pointer to platform_driver which allows
> a platform driver to set up a list of default attributes that will be
> properly created and removed by the platform driver core when a probe()
> function is successful and removed right before the device is unbound.
> 
> Cc: Richard Gong <richard.gong@xxxxxxxxxxxxxxx>
> Cc: Romain Izard <romain.izard.pro@xxxxxxxxx>
> Cc: "Rafael J. Wysocki" <rafael@xxxxxxxxxx>
> Cc: Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
> Cc: Mans Rullgard <mans@xxxxxxxxx>
> Cc: Bartosz Golaszewski <bgolaszewski@xxxxxxxxxxxx>
> Cc: Randy Dunlap <rdunlap@xxxxxxxxxxxxx>
> Cc: linux-kernel@xxxxxxxxxxxxxxx
> Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> ---
>  drivers/base/platform.c         | 40 +++++++++++++++++++++------------
>  include/linux/platform_device.h |  1 +
>  2 files changed, 27 insertions(+), 14 deletions(-)
> 
> diff --git a/drivers/base/platform.c b/drivers/base/platform.c
> index 713903290385..d81fcd435b52 100644
> --- a/drivers/base/platform.c
> +++ b/drivers/base/platform.c
> @@ -598,6 +598,21 @@ struct platform_device *platform_device_register_full(
>  }
>  EXPORT_SYMBOL_GPL(platform_device_register_full);
>  
> +static int platform_drv_remove(struct device *_dev)
> +{
> +	struct platform_driver *drv = to_platform_driver(_dev->driver);
> +	struct platform_device *dev = to_platform_device(_dev);
> +	int ret = 0;
> +
> +	if (drv->remove)
> +		ret = drv->remove(dev);
> +	if (drv->dev_groups)
> +		device_remove_groups(_dev, drv->dev_groups);

Shouldn't you remove the groups before calling driver remove(), which
could be releasing resources used by the attribute implementations?

This would also be the reverse of how what you do at probe.

> +	dev_pm_domain_detach(_dev, true);
> +
> +	return ret;
> +}
> +
>  static int platform_drv_probe(struct device *_dev)
>  {
>  	struct platform_driver *drv = to_platform_driver(_dev->driver);
> @@ -614,8 +629,18 @@ static int platform_drv_probe(struct device *_dev)
>  
>  	if (drv->probe) {
>  		ret = drv->probe(dev);
> -		if (ret)
> +		if (ret) {
>  			dev_pm_domain_detach(_dev, true);
> +			goto out;
> +		}
> +	}
> +	if (drv->dev_groups) {
> +		ret = device_add_groups(_dev, drv->dev_groups);
> +		if (ret) {
> +			platform_drv_remove(_dev);

This would lead to device_remove_groups() being called for the never
added attribute groups. Looks like that may trigger a warning in the
sysfs code judging from a quick look.

> +			return ret;
> +		}
> +		kobject_uevent(&_dev->kobj, KOBJ_CHANGE);
>  	}

Johan