Re: [PATCH v2 3/8] iio: light: cm32181: Handle ACPI instantiating a cm32181 client on the SMBus ARA

[Hans de Goede <hdegoede@redhat.com>]

Hi All,

On 4/27/20 5:50 PM, Hans de Goede wrote:
> Some ACPI systems list 2 I2C resources for the CM3218 sensor. On these
> systems the first I2cSerialBus ACPI-resource points to the SMBus Alert
> Response Address (ARA, 0x0c) and the second I2cSerialBus ACPI-resource
> points to the actual CM3218 sensor address.
> 
>  From the ACPI/x86 side devices with more then 1 I2cSerialBus ACPI-resource
> are handled by the drivers/platform/x86/i2c-multi-instantiate.c code.
> This code will instantiate "cm32181" i2c_client-s for both resources.
> 
> Add a check to cm32181_probe() for the client's address being the ARA
> address, and in that case fail the probe with -ENODEV.
> 
> On these ACPI systems the sensor may have a SMBus Alert asserted at boot,
> if this is the case the sensor will not respond to any i2c_transfers on
> its actual address until we read from the ARA register to clear the Alert.
> 
> Therefor we must (try to) read a byte from the client with the ARA
> register, before returning -ENODEV, so that we clear the Alert and when
> we get called again for the client instantiated for the second
> I2cSerialBus ACPI-resource the sensor will respond to our i2c-transfers.
> 
> Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
> Signed-off-by: Hans de Goede <hdegoede@redhat.com>

So this assumes that i2c-multi-inst will be used for ACPI
CPLM3218 device nodes and we get 2 separate i2c_clients for
the ARA, resp. the real address. This has been discussed on the
linux-acpi list and the conclusion is that that instanting 2
full i2c_clients is not the right solution.

Instead the cm32181 driver should create a "dummy" client
for the second address (which is part of the same chip)
itself, using an acpi version of i2c_new_dummy_device() or
i2c_new_ancillary_device()

I will prepare a v3
of this series with a better solution.

Regards,

Hans



> ---
> Changes in v2
> - s/i2c_client-s/I2C clients/ in added comment
> ---
>   drivers/iio/light/cm32181.c | 16 ++++++++++++++++
>   1 file changed, 16 insertions(+)
> 
> diff --git a/drivers/iio/light/cm32181.c b/drivers/iio/light/cm32181.c
> index fd371b36c7b3..4c26a4a8a070 100644
> --- a/drivers/iio/light/cm32181.c
> +++ b/drivers/iio/light/cm32181.c
> @@ -51,6 +51,8 @@
>   #define CM32181_CALIBSCALE_RESOLUTION	1000
>   #define MLUX_PER_LUX			1000
>   
> +#define SMBUS_ALERT_RESPONSE_ADDRESS	0x0c
> +
>   static const u8 cm32181_reg[CM32181_CONF_REG_NUM] = {
>   	CM32181_REG_ADDR_CMD,
>   };
> @@ -333,6 +335,20 @@ static int cm32181_probe(struct i2c_client *client,
>   	struct iio_dev *indio_dev;
>   	int ret;
>   
> +	/*
> +	 * Some ACPI systems list 2 I2C resources for the CM3218 sensor, the
> +	 * SMBus Alert Response Address (ARA, 0x0c) and the actual I2C address.
> +	 * drivers/platform/x86/i2c-multi-instantiate.c instantiates "cm32181"
> +	 * I2C clients for both resources, ignore the ARA client.
> +	 * On these systems the sensor may have a SMBus Alert asserted at boot,
> +	 * in that case the ARA must be read to clear the Alert otherwise the
> +	 * sensor will not respond on its actual I2C address.
> +	 */
> +	if (client->addr == SMBUS_ALERT_RESPONSE_ADDRESS) {
> +		i2c_smbus_read_byte(client);
> +		return -ENODEV;
> +	}
> +
>   	indio_dev = devm_iio_device_alloc(&client->dev, sizeof(*cm32181));
>   	if (!indio_dev) {
>   		dev_err(&client->dev, "devm_iio_device_alloc failed\n");
>