On 11/01/23 21:08, Pierre-Louis Bossart wrote:
@@ -167,9 +167,11 @@ struct acp63_dev_data { struct platform_device *pdev[ACP63_DEVS]; struct mutex acp_lock; /* protect shared registers */ struct fwnode_handle *sdw_fw_node;
- struct work_struct acp_sdw_dma_work; u16 pdev_mask; u16 pdev_count; u16 pdm_dev_index;
- u16 dma_intr_stat[ACP63_SDW_MAX_STREAMS];
streams and DMAs are different things in SoundWire. You can have a 1:N mapping.
u8 sdw_master_count; u16 sdw0_dev_index; u16 sdw1_dev_index; diff --git a/sound/soc/amd/ps/pci-ps.c b/sound/soc/amd/ps/pci-ps.c index 0fbe5e27f3fb..5b82ee8e3ad8 100644 --- a/sound/soc/amd/ps/pci-ps.c +++ b/sound/soc/amd/ps/pci-ps.c @@ -113,14 +113,37 @@ static int acp63_deinit(void __iomem *acp_base, struct device *dev) return 0; }
+static void acp63_sdw_dma_workthread(struct work_struct *work) +{
- struct acp63_dev_data *adata = container_of(work, struct acp63_dev_data,
acp_sdw_dma_work);- struct sdw_dma_dev_data *sdw_dma_data;
- u32 stream_index;
- u16 pdev_index;
- pdev_index = adata->sdw_dma_dev_index;
- sdw_dma_data = dev_get_drvdata(&adata->pdev[pdev_index]->dev);
- for (stream_index = 0; stream_index < ACP63_SDW_MAX_STREAMS; stream_index++) {
if (adata->dma_intr_stat[stream_index]) {if (sdw_dma_data->sdw_stream[stream_index])snd_pcm_period_elapsed(sdw_dma_data->sdw_stream[stream_index]);is there a reason why you do this in a work thread?
IIRC we did this in SOF because of an issue where during an xrun a stop IPC would be sent while we were dealing with an IPC.
I don't quite see why it's needed for a DMA?
What am I missing?
Initially, we have used in atomic context. We have seen issues during stream closure, in interrupt context , handling period_elapsed causing sleep in atomic context. To avoid that , we have declared dai_link as non-atomic and moved period_elapsed code to work queue.
adata->dma_intr_stat[stream_index] = 0;}- }
+}