Re: [PATCH 04/05] input: RMI4 F01 device control

[Dmitry Torokhov <dmitry.torokhov@gmail.com>]

Hi Chris,

On Fri, Jan 18, 2013 at 05:12:44PM -0800, Christopher Heiny wrote:
> In addition to the changes described in 0/0 of this patchset, this patch
> includes device serialization updated to conform to the latest RMI4
> specification.

I was looking at the various aspects of the RMI4 patchset, trying to
fix the issues that I see, but there is one big issue that I simply do
not have time to tackle - the driver is completely broken on big endian
architectures due to reliance on bitfileds when exchanging the data with
the device.

Consider the following structures:

>  struct f01_device_status {
> -	u8 status_code:4;
> +	enum rmi_device_status status_code:4;
>  	u8 reserved:2;
>  	u8 flash_prog:1;
>  	u8 unconfigured:1;
> @@ -159,4 +136,113 @@ struct f01_device_control_0 {
>  	u8 configured:1;
>  } __attribute__((__packed__));
>  
> +/**
> + * @reset - set this bit to force a firmware reset of the sensor.
> + */
> +struct f01_device_commands {
> +	u8 reset:1;
> +	u8 reserved:7;
> +};
> +

To make this work on BE boxes you either need to add #ifdefs to the
structures reversing the order of fields or use bit shifts and masks to
get/set specific bits in bytes.

I tried converting F01 code (you can see the [likely broken as I can't
test] result in my tree), but I really do not have time for F11.

Thanks.

-- 
Dmitry