Hi Jagan,

On 9 January 2015 at 03:53, Jagan Teki jagannadh.teki@gmail.com wrote:

Hi,

I'm exploring internals of dm where I'm just comparing udevice with Linux model especially on device_id scenarios.

udevice_id - u-boot

Linux:

platform_device_id of_device_id

spi_device_id input_device_id .............. ....... pci_device_id

1. Does this udevice_id is a generic model in u-boot for all variant of devices? Bcz Linux defines each devices have own device_id entries.

The device id is intended to describe a particular node compatible string which can be used with a driver. It corresponds most closely with of_device_id.

On the platform data size (not device tree) we have struct driver_info which is intended to connect a driver with its platform data, to produce a device.

2. diff of of_device_id vs udevice_id

Linux has:

struct of_device_id { char name[32]; char type[32]; char compatible[128]; const void *data; };

U-Boot doesn't have a name and type. It uses a pointer instead of 128 bytes for the compatible string (for efficiency). For U-Boot we use a ulong instead of a pointer for data.

3. udevice_id vs fdt function calls, suppose If I write a dm driver

where my board doesn't have fdt support, is this a valid scenario.

I'm not sure what you are asking. I'd encourage you to use device tree for new drivers.

4. Pls- add any info which I haven't covered.

There's quite a bit of info about platform data and device tree in the driver model README.txt.

Regards, Simon