On Tue, 2007-05-29 at 11:36 -0400, Dmitry Torokhov wrote:

Hi,

On 5/29/07, Hans-Christian Egtvedt [off-list ref] wrote:

quoted

This patch adds support for simulating a mouse using GPIO lines.

The driver needs a platform_data struct to be defined and registered with the
appropriate platform_device.

The driver has been tested on AT32AP7000 microprocessor using the ATSTK1000
development board.

It looks sane although I would recommend switching to input-polldev
when implementing a polled input device.

Oh, I was not aware about this, it seems like just the thing I need.

Could it be scheduled for after the official kernel has this included?
AFAICT it will be released with 2.6.22?

I have attached a new patch with some minor modifications.

quoted

+
+       input->name = pdev->name;
+       input->cdev.dev = &pdev->dev;

Please use input->dev.parent = &pdev->dev. Input devices are being
moved from class_device to struct device.

This is not present in 2.6.21.1, could I schedule it for the next update
of the driver as above?

quoted

+       input->private = pdata;
+
+       /*
+        * Revisit: is bustype, vendor, product and version needed to
+        * input->id? And if they should be present, what values should they
+        * have?
+        */

BUS_HOST seems to be most suitable here. The rest may stay 0.

Thanks, applied.

quoted

+
+       /* private */
+       struct timer_list timer;
+};

I don't think it is a good idea to have timer structure in platform
data which should really be constant. Timer shoudl be part of the
stucture created when driver binds to a device. I can see you may not
want to introduce extra complexity in the driver; however if you use
input-polldev it will handle timer for you.

I was a bit unsure about this before submitting. I have moved the
timer_list into a struct in the driver and added the usual
kmalloc/kfree.

-- 
Best regards
Hans-Christian Egtvedt