Re: [RFC PATCH 00/27] Virtio sound card implementation

[Gerd Hoffmann]

  Hi,

>     - The output from the sound card is accompanied by periodic
>       white noise. I do not know why this is happening. I tried
>       debugging it by writing the buffers to a new wav file and
>       sure enough the contents of the file were different at
>       some places, but I couldn't find what must be causing it.
>       (Relevant patches: #19, #20, #21 and #25.) What steps should
>       I take for debugging this?

Hmm, I'd try to simplify the code.  The two helper functions
virtio_snd_pcm_get_buf() + virtio_snd_pcm_handle_buf_written() look
rather complex to me.  I'd suggest to make them handle a single
virtqueue element only.  That should make the logic simpler.
The loop in virtio_snd_output_cb() will need to take a few more rounds
then, and virtio_snd_pcm_get_buf() would probably have to return the
number of bytes it actually placed in the buffer so you can pass on that
value to AUD_write().

For actual debugging I typically use trace points or temporary debug
printfs or a combination of both.  I'd suggest logging the buffer
handling, filling them from virt queue, writing to AUD, also log the
offset, maybe something goes wrong with partial writes.

>     - What is the suggested way of waiting? When the driver issues
>       the VIRTIO_SND_PCM_STOP ctrl command I want to wait for the
>       buffers existing in tx vq to be consumed before closing the
>       stream.

Store a pointer to the virtqueue element, then complete it when
virtio_snd_output_cb() processed all pending buffers.

take care,
  Gerd