Re: [PATCH v6 3/8] tda9950: use cec_notifier_cec_adap_(un)register

[Hans Verkuil]

On 8/13/19 1:32 PM, Russell King - ARM Linux admin wrote:
> On Tue, Aug 13, 2019 at 01:02:35PM +0200, Dariusz Marcinkiewicz wrote:
>> Use the new cec_notifier_cec_adap_(un)register() functions to
>> (un)register the notifier for the CEC adapter.
>>
>> Signed-off-by: Dariusz Marcinkiewicz <darekm@xxxxxxxxxx>
>> Signed-off-by: Hans Verkuil <hverkuil-cisco@xxxxxxxxx>
>> Tested-by: Hans Verkuil <hverkuil-cisco@xxxxxxxxx>
>> ---
>>  drivers/gpu/drm/i2c/tda9950.c | 12 ++++++------
>>  1 file changed, 6 insertions(+), 6 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/i2c/tda9950.c b/drivers/gpu/drm/i2c/tda9950.c
>> index 8039fc0d83db4..a5a75bdeb7a5f 100644
>> --- a/drivers/gpu/drm/i2c/tda9950.c
>> +++ b/drivers/gpu/drm/i2c/tda9950.c
>> @@ -420,7 +420,8 @@ static int tda9950_probe(struct i2c_client *client,
>>  		priv->hdmi = glue->parent;
>>  
>>  	priv->adap = cec_allocate_adapter(&tda9950_cec_ops, priv, "tda9950",
>> -					  CEC_CAP_DEFAULTS,
>> +					  CEC_CAP_DEFAULTS |
>> +					  CEC_CAP_CONNECTOR_INFO,
>>  					  CEC_MAX_LOG_ADDRS);
>>  	if (IS_ERR(priv->adap))
>>  		return PTR_ERR(priv->adap);
>> @@ -457,13 +458,14 @@ static int tda9950_probe(struct i2c_client *client,
>>  	if (ret < 0)
>>  		return ret;
>>  
>> -	priv->notify = cec_notifier_get(priv->hdmi);
>> +	priv->notify = cec_notifier_cec_adap_register(priv->hdmi, NULL,
>> +						      priv->adap);
>>  	if (!priv->notify)
>>  		return -ENOMEM;
>>  
>>  	ret = cec_register_adapter(priv->adap, priv->hdmi);
>>  	if (ret < 0) {
>> -		cec_notifier_put(priv->notify);
>> +		cec_notifier_cec_adap_unregister(priv->notify);
>>  		return ret;
>>  	}
>>  
>> @@ -473,8 +475,6 @@ static int tda9950_probe(struct i2c_client *client,
>>  	 */
>>  	devm_remove_action(dev, tda9950_cec_del, priv);
>>  
>> -	cec_register_cec_notifier(priv->adap, priv->notify);
>> -
>>  	return 0;
>>  }
>>  
>> @@ -482,8 +482,8 @@ static int tda9950_remove(struct i2c_client *client)
>>  {
>>  	struct tda9950_priv *priv = i2c_get_clientdata(client);
>>  
>> +	cec_notifier_cec_adap_unregister(priv->notify);
>>  	cec_unregister_adapter(priv->adap);
>> -	cec_notifier_put(priv->notify);
> 
> It looks weird to have an unexpectedly different ordering of
> unregistration from the registration path - normally, unregistration
> is the reverse order of initialisation.
> 
> In the initialisation path, it seems that we register the notifier
> and _then_ the adapter.  Here, we unregister the notifier and then
> the adapter rather than what would normally be expected.  Why is
> this?  I suspect there will be drivers created that do this the
> "normal" way round, so if this is a requirement, it needs to be made
> plainly obvious.

It's not a requirement, it just feels better to do it in this order
since cec_unregister_adapter will in general also delete the adapter
(unless an application keeps the cec device open).

So the order is actually: allocate_adapter, then register notifier
and: unregister notifier, then unregister (and typically delete) adapter

Regards,

	Hans

> 
>>  
>>  	return 0;
>>  }
>> -- 
>> 2.23.0.rc1.153.gdeed80330f-goog
>>
>>
>