Re: [PATCH 3/3] extcon: Simplify extcon_dev_register() prototype byremoving unnecessary parameter

[Greg KH]

On Wed, Sep 04, 2013 at 02:18:30PM +0900, Chanwoo Choi wrote:
> On 09/04/2013 10:16 AM, Greg KH wrote:
> > On Wed, Sep 04, 2013 at 09:17:02AM +0900, Chanwoo Choi wrote:
> >> On 09/04/2013 12:57 AM, Greg KH wrote:
> >>> On Mon, Sep 02, 2013 at 10:13:44AM +0900, Chanwoo Choi wrote:
> >>>> Hi Greg,
> >>>>
> >>>> On 09/02/2013 09:40 AM, Greg KH wrote:
> >>>>> On Mon, Sep 02, 2013 at 09:20:08AM +0900, Chanwoo Choi wrote:
> >>>>>> This patch remove extcon_dev_register()'s second parameter which means
> >>>>>> the pointer of parent device to simplify prototype of this function.
> >>>>>
> >>>>> No, please don't.  You want the parent to be passed in, as the core
> >>>>> needs it when it is registered with the system, otherwise it will not
> >>>>> show up in sysfs properly (i.e. you can't set it afterwards.)
> >>>>
> >>>> Currently, each extcon driver have allocated memory for extcon device
> >>>> by using devm_kzalloc() in each extcon device driver.
> >>>
> >>> That seems backwards, the extcon core should be the one doing the
> >>> allocation, and ownership of the device, like all other subsystem cores
> >>> do.  That makes the driver logic much simpler, and the lifetime
> >>> ownership correct (i.e. what happens when a device is unbound from a
> >>> driver by userspace?  The driver can't control the device memory
> >>> anymore...)
> >>>
> >>
> >> OK,
> >> The extcon core will control memory allocation instead of extcon device driver
> >> as following.
> >> - devm_extcon_allocate_device(struct device *dev)
> > 
> > Huh?  Why do you need a devm allocator?  This is a "real" child device,
> > just create it with a "extcon_create_device()" or some such call, like
> > all other busses do?
> 
> I refer Input/IIO subsystem to check the process of device registration.
> Input subsystem has following functions for memory allocation of input device
> and input device registration.
> 
> Input subsystem
> - struct input_dev *devm_input_allocate_device(struct device *device)
> - struct input_dev *input_allocate_device(void)
> - int input_register_device(struct input_dev *dev)
> drivers/input/input.c
> 
> devm_input_allocate_device()/input_allocate_device() can allocate
> memory for input device in input core. And then created input device
> pass input_register_device() as parameter.
> 
> So, input device driver haven't executed kmalloc() or devm_kmalloc()
> to allocate memory of input device by using input_allocate_device()/
> devm_input_allocate_device().
> 
> 
> Also,IIO subsystem has separate iio_device_alloc() function
> to allocate memory for iio device. But IIO subsystem hasn't
> "devm_" allocator. So, If iio device driver fail initialization
> in *_probe, should execute iio_device_free() to free allocated memory.
> 
> IIO Subsystem
> - iio_device_alloc(int sizeof_priv)
> - iio_device_register(struct iio_dev *indio_dev);
> drivers/iio/industrialio-core.c
> 
> 
> So, I think extcon subsystem need extcon_allocate_device()/devm_extcon_allocate_device()
> to allocate for memory extcon device. To implement devm_extcon_allocate_device() function
> , extcon subsystem need extcon_allocate_device() because devm_extcon_allocate_device()
> must call extcon_allocate_device for memory allocation.
> 
> If extcon_allocate_device()/devm_extcon_allocate_device() is implemented to extcon core,
> I think extcon core can control the memory operation of extcon device.
> 
> EXTCON Subsystem
> - devm_extcon_allocate_device(struct device *dev)
> - extcon_allocate_device(void)
> - extcon_dev_register(struct extcon_dev *edev)
> 
> If I'm wrong, please correct me. Thanks.

No, you are totally correct, I was wrong, nevermind, continue on doing a
great job, my mistake :)

greg k-h
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/