Re: [PATCH v9 00/12] Add VIRTIO sound card

[Stefano Garzarella]

On Wed, Sep 13, 2023 at 10:33:07AM +0300, Emmanouil Pitsidianakis wrote:
> This patch series adds an audio device implementing the recent virtio
> sound spec (1.2) and a corresponding PCI wrapper device.
>
> v9 can be found online at:
>
> https://gitlab.com/epilys/qemu/-/tree/virtio-snd-v9
>
> Ref 06e6b17186
>
> Main differences with v8 patch series [^v8]
> <cover.1693252037.git.manos.pitsidianakis@linaro.org>:
>
> - Addressed [^v8] review comments.
> - Add cpu_to_le32(_) and le32_to_cpu(_) conversions for messages from/to
>  the guest according to the virtio spec.
> - Inlined some functions and types to reduce review complexity.
> - Corrected the replies to IO messages; now both Playback and Capture
>  work correctly for me. (If you hear cracks in pulseaudio+guest, try
>  pipewire+guest).

We are seeing something strange with the virtio-sound Linux driver.
It seems that the driver modifies the buffers after exposing them to
the device via the avail ring.

It seems we have this strange behaviour with this built-in QEMU device,
but also with the vhost-device-sound, so it could be some spec
violation in the Linux driver.

Matias also reported on the v8 of this series:
https://lore.kernel.org/qemu-devel/ZPg60lzXWxHPQJEa@fedora/

Can you check if you have the same behaviour?

Nothing that blocks this series of course, but just to confirm that
there may be something to fix in the Linux driver.

Thanks,
Stefano