Re: [RFC PATCH] ASoC: Intel: sof_es8336: add Huawei BOD-WXX9 HWSP support

[Mark Brown]

On Sun, Jun 14, 2026 at 07:10:28PM +0300, Tagir Garaev wrote:
> Huawei BOD-WXX9 routes the ES8336 HPOL/HPOR outputs to both the
> headphone jack and an internal speaker amplifier. The amplifier is
> represented in ACPI as HWSP0001 and is not controlled by the ES8336
> GPIOs alone.

> +static const struct sof_es8336_hwsp_regval hwsp_speaker_restore_profile[] = {
> +	{ 0x01, 0x69 },
> +	{ 0x03, 0x16 },
> +	{ 0x04, 0x80 },
> +	{ 0x05, 0x0c },
> +	{ 0x06, 0x11 },
> +	{ 0x07, 0x93 },
> +	{ 0x09, 0x0b },
> +	{ 0x0b, 0x4b },
> +	{ 0x0c, 0x00 },
> +	{ 0x0d, 0x77 },
> +	{ 0x0f, 0x51 },
> +	{ 0x10, 0x58 },
> +	{ 0x58, 0x00 },
> +	{ 0x59, 0x80 },
> +};

Magic number write sequences in a machine integration...  this looks
awfully like it should be either user visible controls or something more
structured.  In a chip driver it's usually something the manufacturer
simply does not document but this is a third party thing.

> +	} else if (priv->gpio_hwsp) {
> +		gpiod_set_raw_value_cansleep(priv->gpio_hwsp, 0);
> +		usleep_range(SOF_ES8336_HWSP_MUTE_PULSE_MS * 1000,
> +			     SOF_ES8336_HWSP_MUTE_PULSE_MS * 1000 + 1000);
> +		gpiod_set_raw_value_cansleep(priv->gpio_hwsp, 1);

Why _raw?

> +static void sof_es8336_apply_gpio_mux(struct sof_es8336_private *priv)
> +{
> +	struct gpio_desc *speaker_gpio = priv->gpio_speakers;
> +	struct gpio_desc *headphone_gpio = priv->gpio_headphone;
> +	bool hwsp_enabled = sof_es8336_hwsp_amp_enabled() && priv->hwsp_client;
> +	bool hwsp_speaker_on = false;
> +	bool speaker_on;
> +	bool headphone_on;
> +	bool need_enable_delay;
> +
> +	if (!priv->gpio_speakers)
> +		return;
> +
> +	mutex_lock(&priv->gpio_mux_lock);
> +
> +	speaker_on = priv->output_active && !priv->hp_jack_present;
> +	headphone_on = priv->output_active && priv->hp_jack_present;

Why is this not being done by userspace, it looks like policy is being
encoded into the kernel here?  Does having the speaker and headphone
active simultaneously physically work?

Attachment: signature.asc
Description: PGP signature