On Tue, 19 Nov 2019 00:47:53 +0100, Ranjani Sridharan wrote:
On Mon, 2019-11-18 at 21:40 +0100, Takashi Iwai wrote:
On Mon, 18 Nov 2019 20:55:19 +0100, Sridharan, Ranjani wrote:
> Thanks for the clarification, Takashi. But just wondering howwould one pass > on the sync_irq when the snd_card is created? Typically in the case of the > Intel platforms, the card->dev points to the platform device for the machine > driver that registers the card and the PCI device is the parent of the machine > drv platform device.
It's completely up to the driver implementation :) You can implement the own sync_stop ops if that's easier, too.I think this would make sense in the case of the SOF driver and we'd probably need to just call synchronize_irq() in the sync_stop() operation. With this change, we can probably remove the workaround we have to address the issue we were facing during snd_pcm_period_elapsed().
I can give this a try. We might need to run some stress tests to make sure it doesn't break anything.
If this helps for SOF, it'd be great. I have converted only non-ASoC drivers regarding the sync-stop stuff, so it won't conflict my upcoming changes :)
Hi Takashi,
I just realized that In the SOF driver, we only set the component driver ops. The pcm ops are set when creating the new pcm. So, should I also add the sync_stop op in the component driver and set the pcm sync_stop op to point to the component sync_stop op? Just wanted to confirm if I am on the right track.
Yes, I didn't touch this yet, but that's the way to go I suppose. One caveat is that this ops is optional and needs NULL as default, hence you'd need to set only when defined, like copy_user, page or mmap ops, at least.
thanks,
Takashi