On Mon, Sep 29, 2014 at 10:50:24AM +0200, Lars-Peter Clausen wrote:
On 09/28/2014 09:57 AM, Stefan Kristiansson wrote:
Allow the ssm2602/ssm2603/ssm2604 codec driver to be instantiated from the device tree.
Also, add Kconfig prompts to allow manual selection of both the I2C and SPI configuration versions of the driver.
Signed-off-by: Stefan Kristiansson stefan.kristiansson@saunalahti.fi
Looks mostly good, but you should Cc both the devicetree and the ASoC maintainers.
Ok, I'll make sure to do that in v2.
[...]
diff --git a/sound/soc/codecs/ssm2602-i2c.c b/sound/soc/codecs/ssm2602-i2c.c index abd63d5..0d9779d 100644 --- a/sound/soc/codecs/ssm2602-i2c.c +++ b/sound/soc/codecs/ssm2602-i2c.c @@ -41,10 +41,19 @@ static const struct i2c_device_id ssm2602_i2c_id[] = { }; MODULE_DEVICE_TABLE(i2c, ssm2602_i2c_id);
+static const struct of_device_id ssm2602_of_match[] = {
- { .compatible = "adi,ssm2602", },
- { .compatible = "adi,ssm2603", },
- { .compatible = "adi,ssm2604", },
Since the driver is supposed to behave different depending on the device you should set the driver data similar to like it is done in the ssm2602_i2c_id table and add code in the probe function to extract the driver data.
As far as I understand, that is already covered for, ssm2602_i2c_probe() get called with the i2c_device_id struct set according to the compatible string. I.e. if I set it to "adi,ssm2603" in the .dts, ssm2602_i2c_probe() is called with id->name = "ssm2603" and id->driver_data = 0 and if I set it to "adi,ssm2604" I get id->name = "ssm2603" and id->driver_data = 1.
However, when looking into that, I noticed this in ssm2602_probe()
ssm2602->type = SSM2602;
so the value from id->driver_data is never used. (it should be ssm2602->type = type;)
I can send a seperate patch that fixes this, but I only have a ssm2603, so I can't test if it works.
+static const struct of_device_id ssm2602_of_match[] = {
- { .compatible = "adi,ssm2602", },
- { .compatible = "adi,ssm2603", },
- { .compatible = "adi,ssm2604", },
The ssm2603 and ssm2604 do not have a SPI interface.
Right, I'll remove those from that list.
Stefan