platform_device_add_data(pdev, NULL, 0)

platform_device_add_data(pdev, NULL, 0)

[Uwe Kleine-König]

Hello,

mfd_add_device() (defined in drivers/mfd/mfd-core.c) is a wrapper do
allocate platform_devices.

It contains:

	ret = platform_device_add_data(pdev,
			cell->platform_data, cell->data_size);

If cell->data_size is 0 (and so likely cell->platform_data == NULL),
still pdev->dev.platform_data get assigned ZERO_SIZE_PTR.

IMHO the result should better be that pdev->dev.platform_data ends being
NULL, too.  Agreed?

If yes, where is the right place to fix that, mfd_add_device() or
platform_device_add_data()?

Best regards
Uwe

-- 
Pengutronix e.K.                              | Uwe Kleine-König            |
Industrial Linux Solutions                    | http://www.pengutronix.de/  |

Re: platform_device_add_data(pdev, NULL, 0)

[Greg KH]

On Fri, Oct 23, 2009 at 10:26:22PM +0200, Uwe Kleine-König wrote:
> Hello,
> 
> mfd_add_device() (defined in drivers/mfd/mfd-core.c) is a wrapper do
> allocate platform_devices.
> 
> It contains:
> 
> 	ret = platform_device_add_data(pdev,
> 			cell->platform_data, cell->data_size);
> 
> If cell->data_size is 0 (and so likely cell->platform_data == NULL),
> still pdev->dev.platform_data get assigned ZERO_SIZE_PTR.
> 
> IMHO the result should better be that pdev->dev.platform_data ends being
> NULL, too.  Agreed?

Why?  Is this causing problems somewhere?

thanks,

greg k-h

Re: platform_device_add_data(pdev, NULL, 0)

[Uwe Kleine-König]

Hi Greg,

On Mon, Oct 26, 2009 at 09:48:07AM -0700, Greg KH wrote:
> On Fri, Oct 23, 2009 at 10:26:22PM +0200, Uwe Kleine-König wrote:
> > Hello,
> > 
> > mfd_add_device() (defined in drivers/mfd/mfd-core.c) is a wrapper do
> > allocate platform_devices.
> > 
> > It contains:
> > 
> > 	ret = platform_device_add_data(pdev,
> > 			cell->platform_data, cell->data_size);
> > 
> > If cell->data_size is 0 (and so likely cell->platform_data == NULL),
> > still pdev->dev.platform_data get assigned ZERO_SIZE_PTR.
> > 
> > IMHO the result should better be that pdev->dev.platform_data ends being
> > NULL, too.  Agreed?
> 
> Why?  Is this causing problems somewhere?
The "problem" I as follows:  I have a driver that doesn't use
platform_data and I want to change that.  As both the driver and the
device probably go via different trees and I want to do the following in
the driver:

	if (platform_data != NULL)
		/* use it */
	else
		/* use the old data */

OK, I could simply use:

	if (platform_data != NULL && platform_data != ZERO_SIZE_PTR)
		...

but it feels ugly.

Best regards
Uwe

-- 
Pengutronix e.K.                              | Uwe Kleine-König            |
Industrial Linux Solutions                    | http://www.pengutronix.de/  |

Re: platform_device_add_data(pdev, NULL, 0)

[Greg KH]

On Mon, Oct 26, 2009 at 10:19:03PM +0100, Uwe Kleine-König wrote:
> Hi Greg,
> 
> On Mon, Oct 26, 2009 at 09:48:07AM -0700, Greg KH wrote:
> > On Fri, Oct 23, 2009 at 10:26:22PM +0200, Uwe Kleine-König wrote:
> > > Hello,
> > > 
> > > mfd_add_device() (defined in drivers/mfd/mfd-core.c) is a wrapper do
> > > allocate platform_devices.
> > > 
> > > It contains:
> > > 
> > > 	ret = platform_device_add_data(pdev,
> > > 			cell->platform_data, cell->data_size);
> > > 
> > > If cell->data_size is 0 (and so likely cell->platform_data == NULL),
> > > still pdev->dev.platform_data get assigned ZERO_SIZE_PTR.
> > > 
> > > IMHO the result should better be that pdev->dev.platform_data ends being
> > > NULL, too.  Agreed?
> > 
> > Why?  Is this causing problems somewhere?
> The "problem" I as follows:  I have a driver that doesn't use
> platform_data and I want to change that.  As both the driver and the
> device probably go via different trees and I want to do the following in
> the driver:
> 
> 	if (platform_data != NULL)
> 		/* use it */
> 	else
> 		/* use the old data */
> 
> OK, I could simply use:
> 
> 	if (platform_data != NULL && platform_data != ZERO_SIZE_PTR)
> 		...
> 
> but it feels ugly.

Yes, that would be ugly.  How about just sending the driver and device
through the same trees to prevent this from happening?

thanks,

greg k-h