Re: [PATCH 07/12] ASoC: amd: add ACP5x pcm dma driver ops
From: Pierre-Louis Bossart
Date: Wed Jul 07 2021 - 12:35:53 EST
On 7/7/21 12:56 AM, Vijendar Mukunda wrote:
> This patch adds ACP5x PCM driver DMA operations.
>
> Signed-off-by: Vijendar Mukunda <Vijendar.Mukunda@xxxxxxx>
> ---
> sound/soc/amd/vangogh/acp5x-pcm-dma.c | 306 +++++++++++++++++++++++++-
> sound/soc/amd/vangogh/acp5x.h | 106 +++++++++
> 2 files changed, 410 insertions(+), 2 deletions(-)
>
> diff --git a/sound/soc/amd/vangogh/acp5x-pcm-dma.c b/sound/soc/amd/vangogh/acp5x-pcm-dma.c
> index d79712587d30..a4235cf33548 100644
> --- a/sound/soc/amd/vangogh/acp5x-pcm-dma.c
> +++ b/sound/soc/amd/vangogh/acp5x-pcm-dma.c
> @@ -17,8 +17,42 @@
>
> #define DRV_NAME "acp5x_i2s_dma"
>
> -static const struct snd_soc_component_driver acp5x_i2s_component = {
> - .name = DRV_NAME,
> +static const struct snd_pcm_hardware acp5x_pcm_hardware_playback = {
> + .info = SNDRV_PCM_INFO_INTERLEAVED |
> + SNDRV_PCM_INFO_BLOCK_TRANSFER |
> + SNDRV_PCM_INFO_MMAP | SNDRV_PCM_INFO_MMAP_VALID |
> + SNDRV_PCM_INFO_PAUSE | SNDRV_PCM_INFO_RESUME,
> + .formats = SNDRV_PCM_FMTBIT_S16_LE | SNDRV_PCM_FMTBIT_S8 |
> + SNDRV_PCM_FMTBIT_U8 | SNDRV_PCM_FMTBIT_S32_LE,
is S24_4LE supported? seems more useful than 8-bit audio these days, no?
> +static void config_acp5x_dma(struct i2s_stream_instance *rtd, int direction)
> +{
> + u16 page_idx;
> + u32 low, high, val, acp_fifo_addr, reg_fifo_addr;
> + u32 reg_dma_size, reg_fifo_size;
> + dma_addr_t addr;
> +
> + addr = rtd->dma_addr;
> + if (direction == SNDRV_PCM_STREAM_PLAYBACK) {
> + switch (rtd->i2s_instance) {
> + case I2S_HS_INSTANCE:
> + val = ACP_SRAM_HS_PB_PTE_OFFSET;
> + break;
> + case I2S_SP_INSTANCE:
> + default:
> + val = ACP_SRAM_SP_PB_PTE_OFFSET;
> + }
> + } else {
> + switch (rtd->i2s_instance) {
> + case I2S_HS_INSTANCE:
> + val = ACP_SRAM_HS_CP_PTE_OFFSET;
> + break;
> + case I2S_SP_INSTANCE:
> + default:
> + val = ACP_SRAM_SP_CP_PTE_OFFSET;
> + }
> + }
> + /* Group Enable */
> + acp_writel(ACP_SRAM_PTE_OFFSET | BIT(31), rtd->acp5x_base +
> + ACPAXI2AXI_ATU_BASE_ADDR_GRP_1);
> + acp_writel(PAGE_SIZE_4K_ENABLE, rtd->acp5x_base +
> + ACPAXI2AXI_ATU_PAGE_SIZE_GRP_1);
> +
> + for (page_idx = 0; page_idx < rtd->num_pages; page_idx++) {
> + /* Load the low address of page int ACP SRAM through SRBM */
> + low = lower_32_bits(addr);
> + high = upper_32_bits(addr);
> +
> + acp_writel(low, rtd->acp5x_base + ACP_SCRATCH_REG_0 + val);
> + high |= BIT(31);
> + acp_writel(high, rtd->acp5x_base + ACP_SCRATCH_REG_0 + val
> + + 4);
use single line? I find the indentation style quite an eyesore...
> + /* Move to next physically contiguous page */
> + val += 8;
> + addr += PAGE_SIZE;
> + }
> +
> + if (direction == SNDRV_PCM_STREAM_PLAYBACK) {
> + switch (rtd->i2s_instance) {
> + case I2S_HS_INSTANCE:
> + reg_dma_size = ACP_HS_TX_DMA_SIZE;
> + acp_fifo_addr = ACP_SRAM_PTE_OFFSET +
> + HS_PB_FIFO_ADDR_OFFSET;
> + reg_fifo_addr = ACP_HS_TX_FIFOADDR;
> + reg_fifo_size = ACP_HS_TX_FIFOSIZE;
> + acp_writel(I2S_HS_TX_MEM_WINDOW_START,
> + rtd->acp5x_base + ACP_HS_TX_RINGBUFADDR);
> + break;
> +
> + case I2S_SP_INSTANCE:
> + default:
> + reg_dma_size = ACP_I2S_TX_DMA_SIZE;
> + acp_fifo_addr = ACP_SRAM_PTE_OFFSET +
> + SP_PB_FIFO_ADDR_OFFSET;
> + reg_fifo_addr = ACP_I2S_TX_FIFOADDR;
> + reg_fifo_size = ACP_I2S_TX_FIFOSIZE;
> + acp_writel(I2S_SP_TX_MEM_WINDOW_START,
> + rtd->acp5x_base + ACP_I2S_TX_RINGBUFADDR);
> + }
> + } else {
> + switch (rtd->i2s_instance) {
> + case I2S_HS_INSTANCE:
> + reg_dma_size = ACP_HS_RX_DMA_SIZE;
> + acp_fifo_addr = ACP_SRAM_PTE_OFFSET +
> + HS_CAPT_FIFO_ADDR_OFFSET;
> + reg_fifo_addr = ACP_HS_RX_FIFOADDR;
> + reg_fifo_size = ACP_HS_RX_FIFOSIZE;
> + acp_writel(I2S_HS_RX_MEM_WINDOW_START,
> + rtd->acp5x_base + ACP_HS_RX_RINGBUFADDR);
> + break;
> +
> + case I2S_SP_INSTANCE:
> + default:
> + reg_dma_size = ACP_I2S_RX_DMA_SIZE;
> + acp_fifo_addr = ACP_SRAM_PTE_OFFSET +
> + SP_CAPT_FIFO_ADDR_OFFSET;
> + reg_fifo_addr = ACP_I2S_RX_FIFOADDR;
> + reg_fifo_size = ACP_I2S_RX_FIFOSIZE;
> + acp_writel(I2S_SP_RX_MEM_WINDOW_START,
> + rtd->acp5x_base + ACP_I2S_RX_RINGBUFADDR);
> + }
> + }
> + acp_writel(DMA_SIZE, rtd->acp5x_base + reg_dma_size);
> + acp_writel(acp_fifo_addr, rtd->acp5x_base + reg_fifo_addr);
> + acp_writel(FIFO_SIZE, rtd->acp5x_base + reg_fifo_size);
> + acp_writel(BIT(I2S_RX_THRESHOLD) | BIT(HS_RX_THRESHOLD)
> + | BIT(I2S_TX_THRESHOLD) | BIT(HS_TX_THRESHOLD),
> + rtd->acp5x_base + ACP_EXTERNAL_INTR_CNTL);
> +}
> +
> +static int acp5x_dma_hw_params(struct snd_soc_component *component,
> + struct snd_pcm_substream *substream,
> + struct snd_pcm_hw_params *params)
> +{
> + struct i2s_stream_instance *rtd;
> + struct snd_soc_pcm_runtime *prtd;
> + struct snd_soc_card *card;
> + struct acp5x_platform_info *pinfo;
> + struct i2s_dev_data *adata;
> + u64 size;
> +
> + prtd = asoc_substream_to_rtd(substream);
> + card = prtd->card;
> + pinfo = snd_soc_card_get_drvdata(card);
> + adata = dev_get_drvdata(component->dev);
> + rtd = substream->runtime->private_data;
> +
> + if (!rtd)
> + return -EINVAL;
> +
> + if (pinfo) {
> + if (substream->stream == SNDRV_PCM_STREAM_PLAYBACK) {
> + rtd->i2s_instance = pinfo->play_i2s_instance;
> + switch (rtd->i2s_instance) {
> + case I2S_HS_INSTANCE:
> + adata->play_stream = substream;
> + break;
> + case I2S_SP_INSTANCE:
> + default:
> + adata->i2ssp_play_stream = substream;
> + }
> + } else {
> + rtd->i2s_instance = pinfo->cap_i2s_instance;
> + switch (rtd->i2s_instance) {
> + case I2S_HS_INSTANCE:
> + adata->capture_stream = substream;
> + break;
> + case I2S_SP_INSTANCE:
> + default:
> + adata->i2ssp_capture_stream = substream;
> + }
> + }
> + } else {
> + pr_err("pinfo failed\n");
that seems like a rather useless message. if you want a log at least use dev_err(component->dev
> + }
> + size = params_buffer_bytes(params);
> + rtd->dma_addr = substream->dma_buffer.addr;
> + rtd->num_pages = (PAGE_ALIGN(size) >> PAGE_SHIFT);
> + config_acp5x_dma(rtd, substream->stream);
> + return 0;
> +}
> +