Hello Marek,

On 05/15/2014 09:01 PM, Marek Vasut wrote:

On Friday, May 09, 2014 at 08:58:02 AM, Przemyslaw Marczak wrote:

...

Hello,

[...]

...

struct power_ops_key_power { int (*key_state) (int *state); };

This could be a key input device.

...

struct power_ops_rtc { int (*sec) (int set_get, int *val); int (*min) (int set_get, int *val); int (*hour) (int set_get, int *val); int (*day) (int set_get, int *val); int (*month) (int set_get, int *val); int (*year) (int set_get, int *val); };

RTC device.

...

struct power_ops_motor { int (*configure) (void); int (*enable) (int time, int gain); };

struct power_ops_led_flash { int (*configure) (void); int (*enable) (void); int (*disable) (void); };

LED device.

Yes, they are actually a various devices with separated options.

I looked into some Frescale and Maxim PMICs documentation. And those integrated devices usually provides various ops on one or two interfaces. So I think it would be nice to have one framework and e.g. register available devices options for each interface of each device.

It seems like you're trying to assemble a huge framework while avoiding the already-present frameworks. No?

You're right - we have some frameworks at present and this framework could be an additional abstraction level between device and uboot commands. Device could be presented as it was designed - is it bad idea?

Best regards, Marek Vasut

Thanks