Re: [alsa-devel] [PATCH v8 2/6] ASoC: amd: Refactoring of DAI from DMA driver

[vishnu]

On 19/11/19 7:23 PM, Pierre-Louis Bossart wrote:
> > +static int acp3x_dai_probe(struct platform_device *pdev)
> > +{
> > +ÂÂÂ struct resource *res;
> > +ÂÂÂ struct i2s_dev_data *adata;
> > +ÂÂÂ int status;
> > +
> > +ÂÂÂ adata = devm_kzalloc(&pdev->dev, sizeof(struct i2s_dev_data),
> > +ÂÂÂÂÂÂÂÂÂÂÂ GFP_KERNEL);
> > +ÂÂÂ if (!adata)
> > +ÂÂÂÂÂÂÂ return -ENOMEM;
> > +
> > +ÂÂÂ res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> > +ÂÂÂ if (!res) {
> > +ÂÂÂÂÂÂÂ dev_err(&pdev->dev, "IORESOURCE_MEM FAILED\n");
> > +ÂÂÂÂÂÂÂ return -ENOMEM;
> > +ÂÂÂ }
> > +
> > +ÂÂÂ adata->acp3x_base = devm_ioremap(&pdev->dev, res->start,
> > +ÂÂÂÂÂÂÂÂÂÂÂ resource_size(res));
> > +ÂÂÂ if (IS_ERR(adata->acp3x_base))
> > +ÂÂÂÂÂÂÂ return PTR_ERR(adata->acp3x_base);
> > +
> > +ÂÂÂ adata->i2s_irq = res->start;
> > +ÂÂÂ dev_set_drvdata(&pdev->dev, adata);
> > +ÂÂÂ status = devm_snd_soc_register_component(&pdev->dev,
> > +ÂÂÂÂÂÂÂÂÂÂÂ &acp3x_dai_component, &acp3x_i2s_dai, 1);
> > +ÂÂÂ if (status) {
> > +ÂÂÂÂÂÂÂ dev_err(&pdev->dev, "Fail to register acp i2s dai\n");
> > +ÂÂÂÂÂÂÂ return -ENODEV;
> > +ÂÂÂ }
> > +ÂÂÂ pm_runtime_set_autosuspend_delay(&pdev->dev, 5000);
> > +ÂÂÂ pm_runtime_use_autosuspend(&pdev->dev);
> > +ÂÂÂ pm_runtime_enable(&pdev->dev);
>
> question: here you want to use pm_runtime for this platform device...
>
> > +ÂÂÂ return 0;
> > +}
> > +
> > +static int acp3x_dai_remove(struct platform_device *pdev)
> > +{
> > +ÂÂÂ pm_runtime_disable(&pdev->dev);
> > +ÂÂÂ return 0;
> > +}
> > +static struct platform_driver acp3x_dai_driver = {
> > +ÂÂÂ .probe = acp3x_dai_probe,
> > +ÂÂÂ .remove = acp3x_dai_remove,
> > +ÂÂÂ .driver = {
> > +ÂÂÂÂÂÂÂ .name = "acp3x_i2s_playcap",
>
> ... but here there is no .pm structure and I don't see any 
> suspend/resume routines for this driver...
>
> > +ÂÂÂ },
> > +};
>
> > @@ -774,13 +586,14 @@ static struct platform_driver acp3x_dma_driver = {
> > ÂÂÂÂÂ .probe = acp3x_audio_probe,
> > ÂÂÂÂÂ .remove = acp3x_audio_remove,
> > ÂÂÂÂÂ .driver = {
> > -ÂÂÂÂÂÂÂ .name = "acp3x_rv_i2s",
> > +ÂÂÂÂÂÂÂ .name = "acp3x_rv_i2s_dma",
> > ÂÂÂÂÂÂÂÂÂ .pm = &acp3x_pm_ops,
> > ÂÂÂÂÂ },
>
> ... but for this other platform_driver you do have a .pm structure and 
> suspend-resume implementations.
>
> Wondering if this is a miss or a feature?

As per our design, ACP IP specific changes like ACP power on/off will be 
handled in ACP pci driver(parent device for DMA device and I2S 
controller(nothing but CPU DAI))

Where as In DMA driver during runtime suspend/resume interrupts will be 
disabled and enabled.

But in DAI driver there is nothing to be done in suspend and resume just 
returning zero so we have not added PM suspend/resume here in DAI.

So is it expected to add the suspend resumes with returning zero.Or if 
pm runtime is not needed in CPU DAI shall we remove the existing PM 
related calls in DAI.

Please suggest us.


Thanks,
Vishnu