2019-11-07 13:30:12

by Ravulapati Vishnu vardhan rao

Subject: [RESEND PATCH v3 5/6] ASoC: amd: handle ACP3x i2s-sp watermark interrupt.

whenever audio data equal to I2S-SP fifo watermark level is
produced/consumed, interrupt is generated.

Signed-off-by: Ravulapati Vishnu vardhan rao <[email protected]>
---
sound/soc/amd/raven/acp3x-pcm-dma.c | 14 ++++++++++++++
1 file changed, 14 insertions(+)

diff --git a/sound/soc/amd/raven/acp3x-pcm-dma.c b/sound/soc/amd/raven/acp3x-pcm-dma.c
index 8fab505..629a32f 100644
--- a/sound/soc/amd/raven/acp3x-pcm-dma.c
+++ b/sound/soc/amd/raven/acp3x-pcm-dma.c
@@ -176,6 +176,13 @@ static irqreturn_t i2s_irq_handler(int irq, void *dev_id)
snd_pcm_period_elapsed(rv_i2s_data->play_stream);
play_flag = 1;
}
+ if ((val & BIT(I2S_TX_THRESHOLD)) &&
+ rv_i2s_data->i2ssp_play_stream) {
+ rv_writel(BIT(I2S_TX_THRESHOLD),
+ rv_i2s_data->acp3x_base + mmACP_EXTERNAL_INTR_STAT);
+ snd_pcm_period_elapsed(rv_i2s_data->i2ssp_play_stream);
+ play_flag = 1;
+ }

if ((val & BIT(BT_RX_THRESHOLD)) && rv_i2s_data->capture_stream) {
rv_writel(BIT(BT_RX_THRESHOLD), rv_i2s_data->acp3x_base +
@@ -183,6 +190,13 @@ static irqreturn_t i2s_irq_handler(int irq, void *dev_id)
snd_pcm_period_elapsed(rv_i2s_data->capture_stream);
cap_flag = 1;
}
+ if ((val & BIT(I2S_RX_THRESHOLD)) &&
+ rv_i2s_data->i2ssp_capture_stream) {
+ rv_writel(BIT(I2S_RX_THRESHOLD),
+ rv_i2s_data->acp3x_base + mmACP_EXTERNAL_INTR_STAT);
+ snd_pcm_period_elapsed(rv_i2s_data->i2ssp_capture_stream);
+ cap_flag = 1;
+ }

if (play_flag | cap_flag)
return IRQ_HANDLED;
--
2.7.4

2019-11-07 17:06:45

by Pierre-Louis Bossart

Subject: Re: [alsa-devel] [RESEND PATCH v3 5/6] ASoC: amd: handle ACP3x i2s-sp watermark interrupt.

On 11/7/19 7:24 AM, Ravulapati Vishnu vardhan rao wrote:
> whenever audio data equal to I2S-SP fifo watermark level is
> produced/consumed, interrupt is generated.
>
> Signed-off-by: Ravulapati Vishnu vardhan rao <[email protected]>
> ---
> sound/soc/amd/raven/acp3x-pcm-dma.c | 14 ++++++++++++++
> 1 file changed, 14 insertions(+)
>
> diff --git a/sound/soc/amd/raven/acp3x-pcm-dma.c b/sound/soc/amd/raven/acp3x-pcm-dma.c
> index 8fab505..629a32f 100644
> --- a/sound/soc/amd/raven/acp3x-pcm-dma.c
> +++ b/sound/soc/amd/raven/acp3x-pcm-dma.c
> @@ -176,6 +176,13 @@ static irqreturn_t i2s_irq_handler(int irq, void *dev_id)
> snd_pcm_period_elapsed(rv_i2s_data->play_stream);
> play_flag = 1;
> }
> + if ((val & BIT(I2S_TX_THRESHOLD)) &&
> + rv_i2s_data->i2ssp_play_stream) {
> + rv_writel(BIT(I2S_TX_THRESHOLD),
> + rv_i2s_data->acp3x_base + mmACP_EXTERNAL_INTR_STAT);
> + snd_pcm_period_elapsed(rv_i2s_data->i2ssp_play_stream);

The commit message looks odd. If you have a DMA and signal a period
elapsed, is this really a case where the FIFO level reaches a watermark?
Usually the FIFOs are somewhat small and there are multiple DMA
transfers per period?