Re: [PATCH] added kernel module for FTS sensor chip "Teutates"

[Guenter Roeck <linux@roeck-us.net>]

On 06/09/2016 02:13 AM, thilo.cestonaro@ts.fujitsu.com wrote:
> Hey Guenter,
>
>>> +static int ftsteutates_detect(struct i2c_client *client,
>>> +		struct i2c_board_info *info)
>>> +{
>>> +	int val = ftsteutates_read_byte(client, FTSTEUTATES_DEVICE_ID_REG);
>>> +
>>> +	/* Baseboard Management Controller */
>>> +	if ((val & 0xF0) == 0x10) {
>>> +		switch (val & 0x0F) {
>>> +		case 0x01:
>>> +			strlcpy(info->type, ftsteutates_id[teutates].name,
>>> +			I2C_NAME_SIZE);
>>> +			info->flags = 0;
>>> +			return 0;
>>> +		}
>>> +	}
>> This is not sufficient for a detect function; it would result in many false
>> positives. Please drop it unless a much better means to detect the chip
>> is available.
>>
> Would it be sufficient, if I use dmi_name_in_vendors and check for Fujitsu? So the
> BMC check will only be used on FJ Boards?
>
> The problem is, that the firmware itself has only this device ID register.
> For future revisions I'll ask for more possibilities to identify the chip but for now
> I need to use other methods.
>

You don't need a detect function to start with. Using DMI in an i2c driver
is not appropriate either; an i2c driver is expected to be board agnostic.

If you can determine that the chip is present in the system using DMI, you
would normally have a second (platform)) driver which detects the platform
and instantiate the device, using either i2c_new_device() or
i2c_register_board_info(). See  Documentation/i2c/instantiating-devices
for more details.

Hope this helps,
Guenter