[alsa-devel] [RFC] ASoC: Intel: skl_hda_dsp_common: set long name for skl_hda_card

[Pierre-Louis Bossart]

On 5/21/19 2:07 AM, Rander Wang wrote:
> skl_hda_generic machine driver is used by many different devices
> and userspace uses the card long name to differentiate devices.
> The card long name is figured out by DMI info and is in format
> of "vendor-product-version-board". Ucm file is searched by this
> card long name and one problem is different devices can't share
> one ucm file based on this type of long name. We have three different
> product devices with the same codecs and audio settings, and these
> devices can share the same ucm setting, but now we need to provide
> three ucm files with different long names, because we can't get a
> same long name from DMI info.
> 
> The solution is to provide card long name in machine driver like
> bytcr_rt5640 and ASoC will use this long name to generate final long
> name. The card long name is composed of codec name, input and output
> enabled by devices. The long name should be initialized after hda codec
> is initialized and before sound card checking long name which is done
> after dai link initialization, so the long name is set in codec dai
> link initialization function.
> 
> Possible card long names may be:
> skl-hda-ALC233-config135440
> skl-hda-ALC700-config69649
> skl-hda-ALC3204-config135185

Is this really what we want to do?

The configs are derived from the codec capabilities, which are typically 
a superset of what is actually supported on the production device. It's 
not usual e.g. to have a codec support an S/PDIF output but there is no 
physical connector. If you expose capabilities to the user that are not 
there, that's not so good. We have this example in one SOF device where 
the Digital output does nothing.

Also the number of combinations could really be quite large, unlike for 
bytcr_rt5640 where the options were limited to 1-2 bits here you have 5 
options with 4 bits each. How many UCM files are we going to end-up 
maintaining out of the possible 1M?

There is an alternate solution: you *could* use the DMI-base name and 
the include mechanism to reuse the same config if you wanted to. It's 
limiting in the sense that for every new platform we'll have to figure 
out which config it uses, but it'll be exactly what the user needs.

> 
> Tested on intel hda platform whiskylake with SOF driver and gemilake
> with intel SST driver
> 
> Signed-off-by: Rander Wang <rander.wang at linux.intel.com>
> ---
>   sound/soc/intel/boards/skl_hda_dsp_common.c | 54 +++++++++++++++++++++++++++++
>   1 file changed, 54 insertions(+)
> 
> diff --git a/sound/soc/intel/boards/skl_hda_dsp_common.c b/sound/soc/intel/boards/skl_hda_dsp_common.c
> index 8b68f41a5b88..0193d2138e16 100644
> --- a/sound/soc/intel/boards/skl_hda_dsp_common.c
> +++ b/sound/soc/intel/boards/skl_hda_dsp_common.c
> @@ -11,11 +11,18 @@
>   #include <sound/pcm.h>
>   #include <sound/pcm_params.h>
>   #include <sound/soc.h>
> +#include <sound/hda_codec.h>
> +#include "../../../pci/hda/hda_jack.h"
> +#include "../../../pci/hda/hda_local.h"
> +#include "../../../pci/hda/hda_auto_parser.h"
> +#include "../../../pci/hda/hda_generic.h"
>   #include "../../codecs/hdac_hdmi.h"
> +#include "../../codecs/hdac_hda.h"
>   #include "../skylake/skl.h"
>   #include "skl_hda_dsp_common.h"
>   
>   #define NAME_SIZE	32
> +static char skl_hda_long_name[NAME_SIZE];
>   
>   int skl_hda_hdmi_add_pcm(struct snd_soc_card *card, int device)
>   {
> @@ -39,6 +46,52 @@ int skl_hda_hdmi_add_pcm(struct snd_soc_card *card, int device)
>   	return 0;
>   }
>   
> +int skl_hda_long_name_init(struct snd_soc_pcm_runtime *rtd)
> +{
> +	struct snd_soc_rtdcom_list *new_rtdcom;
> +	struct snd_soc_component *component;
> +	struct hdac_hda_priv *hda_pvt;
> +	struct auto_pin_cfg *autocfg;
> +	struct hda_gen_spec *spec;
> +	struct snd_soc_card *card;
> +	const char *name;
> +	int config;
> +
> +	name = rtd->dai_link->codecs->name;
> +	card = rtd->card;
> +
> +	list_for_each_entry(new_rtdcom, &rtd->component_list, list) {
> +		component = new_rtdcom->component;
> +		if (name && strcmp(component->name, name) == 0) {
> +			hda_pvt = snd_soc_component_get_drvdata(component);
> +
> +			if (!hda_pvt)
> +				return -EINVAL;
> +
> +			spec = hda_pvt->codec.spec;
> +			autocfg = &spec->autocfg;
> +
> +			/*
> +			 * config is figured out by combining the number of
> +			 * enabled input and output.
> +			 */
> +			config = autocfg->speaker_outs | (autocfg->hp_outs << 4)
> +				| (autocfg->dig_outs << 8)
> +				| (autocfg->line_outs << 12)
> +				| (autocfg->num_inputs << 16);
> +			snprintf(skl_hda_long_name, sizeof(skl_hda_long_name),
> +				 "skl-hda-%s-config%d",
> +				 hda_pvt->codec.core.chip_name,
> +				 config);
> +
> +			card->long_name = skl_hda_long_name;
> +			break;
> +		}
> +	}
> +
> +	return 0;
> +}
> +
>   /* skl_hda_digital audio interface glue - connects codec <--> CPU */
>   struct snd_soc_dai_link skl_hda_be_dai_links[HDA_DSP_MAX_BE_DAI_LINKS] = {
>   	/* Back End DAI links */
> @@ -79,6 +132,7 @@ struct snd_soc_dai_link skl_hda_be_dai_links[HDA_DSP_MAX_BE_DAI_LINKS] = {
>   		.dpcm_playback = 1,
>   		.dpcm_capture = 1,
>   		.no_pcm = 1,
> +		.init = skl_hda_long_name_init,
>   	},
>   	{
>   		.name = "Digital Playback and Capture",
>