Re: [PATCH 1/5] ALSA: snd-usb: implement new endpoint streaming model

[Clemens Ladisch <clemens@ladisch.de>]

Daniel Mack wrote:
> +++ b/sound/usb/card.h
> +struct snd_usb_endpoint {
>  	
> +	struct snd_usb_endpoint *sync_slave;

In theory, there could be more than one slave.

> +struct snd_usb_endpoint *snd_usb_add_endpoint(struct snd_usb_audio *chip,
>  	...
> +	/* select the alt setting once so the endpoints become valid */
> +	snd_printk("%s() calling usb_set_interface()\n", __func__);
> +	usb_set_interface(ep->chip->dev, ep->iface, ep->alt_idx);

This needs error handling.

> +void snd_usb_endpoint_activate(struct snd_usb_endpoint *ep)
> +{
> +		usb_set_interface(ep->chip->dev, ep->iface, ep->alt_idx);

This too.

> +void snd_usb_handle_sync_urb(struct snd_usb_endpoint *ep,
> +{
> +		for (i = 0; i < in_ctx->packets; i++) {
> +			if (urb->iso_frame_desc[i].status == 0)
> +				out_ctx->packet_size[i] =
> +					urb->iso_frame_desc[i].actual_length / ep->stride;
> +			else
> +				out_ctx->packet_size[i] = 0;

When there is any error in any received packet (status != 0), we do not
know how many frames to send.  Sending zero frames is certain to be
wrong and will break synchronization between the two streams.

As far as I can see, the only way to handle such errors is to stop the
stream(s) and the PCM device(s).


Regards,
Clemens