[alsa-devel] [PATCH v2 116/146] ASoC: sof: use modern dai_link style

[Pierre-Louis Bossart]

On 6/28/19 4:06 AM, Kuninori Morimoto wrote:
> 
> Hi Pierre-Louis
> 
>>> @@ -1883,7 +1883,7 @@ static int soc_tplg_fe_link_create(struct soc_tplg *tplg,
>>>           int ret;
>>>
>>>           /* link + cpu + codec */
>>> -       link = kzalloc(sizeof(*link) + (2 * sizeof(*dlc)), GFP_KERNEL);
>>> +       link = kzalloc(sizeof(*link) + (3 * sizeof(*dlc)), GFP_KERNEL);
>>>           if (link == NULL)
>>>                   return -ENOMEM;
>>>
>>> @@ -1891,9 +1891,11 @@ static int soc_tplg_fe_link_create(struct soc_tplg *tplg,
>>>
>>>           link->cpus      = &dlc[0];
>>>           link->codecs    = &dlc[1];
>>> +       link->platforms = &dlc[2];
>>>
>>>           link->num_cpus   = 1;
>>>           link->num_codecs = 1;
>>> +       link->num_platforms = 1;
>>>
>>> Can you please help me figure this out?
>>
>> Isn't this fixed by my patch "ASoC: soc-topology: fix modern dai link
>> style" applied on 6/13? Looks like the same issue to me.
> 
> This is very impertinent comment, but it is possible to allow NULL platform
> instead of dummy platform by this or similar code ?
> I guess it is nice for SOF future.
> I can't test and not familiar with SOF thought...
> 
> 	if (link->platforms)
> 		link->platforms->name = dev_name(sdev->dev);

It's a good question. To be honest I don't fully understand what this 
'platform' field is needed for... I was just trying to maintain 'as-is' 
functionality. If anyone has a good explanation on when this field might 
be required and for what purpose, and when it can be made optional, I am 
all ears.