Re: [RFC PATCH 5/17] input: rmidev character driver for RMI4 sensors

[Linus Walleij <linus.walleij@linaro.org>]

On Fri, Aug 17, 2012 at 3:17 PM, Christopher Heiny <cheiny@synaptics.com> wrote:

> Driver for Synaptics touchscreens using RMI4 protocol.

Really? This looks more like some custom char driver to get a pipe
into the device from userspace. Put in a proper description of what this
is for.

If the purpose is to read/write arbitrary addresses in the device,
you should use regmap's debugfs interface for this instead,
it is much better suited for the task.

(...)
> +#define RMI_CHAR_DEV_TMPBUF_SZ 128
> +#define RMI_REG_ADDR_PAGE_SELECT 0xFF
> +#define REG_ADDR_LIMIT 0xFFFF
> +
> +struct rmidev_data {
> +       /* mutex for file operation*/
> +       struct mutex file_mutex;
> +       /* main char dev structure */
> +       struct cdev main_dev;
> +
> +       /* pointer to the corresponding RMI4 device.  We use this to do */
> +       /* read, write, etc. */
> +       struct rmi_device *rmi_dev;
> +       /* reference count */
> +       int ref_count;

Something tells me you should atleast use <linux/kref.h> for this.
It also solves a few atomicity problems in a good standard way.
See Documentation/kref.txt

> +/*store dynamically allocated major number of char device*/
> +static int rmidev_major_num;

You need to patch your desired major number into
Documentation/devices.txt

> +static struct class *rmidev_device_class;

Last time discussed with Greg, class devices were deprecated,
and you should just use a bus instead. (But not sure.)

Yours,
Linus Walleij