Re: [RFC PATCH 07/27] virtio-snd: Add properties for class init

[Shreyansh Chouhan <chouhan.shreyansh2702@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 3273 bytes --]

On Wed, 5 May 2021 at 02:00, Laurent Vivier <laurent@vivier.eu> wrote:

> Hi Shreyansh,
>
> First of all, thank you for your work, I was expecting a virtio sound
> device for some time...
>
> You're welcome :)

> Le 04/05/2021 à 21:35, Shreyansh Chouhan a écrit :
> > On Tue, 4 May 2021 at 19:02, Laurent Vivier <laurent@vivier.eu <mailto:
> laurent@vivier.eu>> wrote:
> >
> >     There is nothing specific to PCI in that code, why do you prevent
> the use of virtio-snd as a MMIO
> >     device?
> >
> > I am sorry I do not understand your question completely. If by
> preventing the use of virtio-snd, you
> > mean
> > why did I add the PCI dependencies to the Kconfig file, then I think I
> must have been a bit confused
> > while writing it. VIRTIO_PCI already includes those dependencies, I will
> change the dependency to
> > VIRTIO. (Which is what it is for other virtio devices too.)
> >
> > However if you mean why did I not add an MMIO binding for this device,
> then there is no
> > specific reason. I simply followed what QEMU had been doing for the
> other virtio devices.
> > Will there be any advantages to implementing the device as a MMIO
> device?
>
> No, the question was only about the dependencies, generally a a virtio
> device is binded to a virtio
> bus, and virtio PCI is a PCI card providing a virtio bus with the virtio
> device attached to it.
>
> For instance, for virtio-net-pci:
>
> HOST
>
>   -> PCI Host controller
>
>     -> PCI virtio net device (TYPE_VIRTIO_NET_PCI)
>
>       -> virtio Bus (TYPE_VIRTIO_BUS)
>
>         -> virtio net device (TYPE_VIRTIO_NET)
>
> TYPE_VIRTIO_NET_PCI is created by hw/virtio/virtio-net-pci.c and
> TYPE_VIRTIO_NET by hw/net/vrtio-net.c:
>
> hw/virtio/meson.build:
>
> virtio_pci_ss.add(when: 'CONFIG_VIRTIO_NET', if_true:
> files('virtio-net-pci.c'))
> virtio_ss.add_all(when: 'CONFIG_VIRTIO_PCI', if_true: virtio_pci_ss)
>
> hw/net/meson.build
> specific_ss.add(when: 'CONFIG_VIRTIO_NET', if_true: files('virtio-net.c'))
>
> hw/net/Kconfig:
>
> config VIRTIO_NET
>     bool
>     default y
>     depends on VIRTIO
>
> So:
>
> the virtio-net device is built when VIRTIO_NET is set,
> the virtio-net-pci device is build when VIRTIO_NET and VIRTIO_PCI are set.
>
> So what I expect for virtio-snd:
>
> hw/virtio/meson.build:
>
> virtio_pci_ss.add(when: 'CONFIG_VIRTIO_SND', if_true:
> files('virtio-snd-pci.c'))
>
> hw/audio/meson.build:
>
> softmmu_ss.add(when: 'CONFIG_VIRTIO_SND', if_true: files('virtio-snd.c'))
>
> hw/audio/Kconfig
>
> config VIRTIO_SND
>     bool
>     default y
>     depends on VIRTIO
>
> With that kind of config, a machine without PCI bus will be able to create
> a virtio bus to add your
> virtio device (like s390x with virtio-ccw of any other MMIO machine like
> the virt machines).
>
> Thanks a lot for the detailed explanation. It clarifies everything
regarding the separate
pci and device source files and how they are built, which was confusing me
a little bit.
I will fix the meson.build and Kconfig files in the upcoming versions of
these patches.

In short: update your hw/audio/config, and all will be fine.


>
Thanks,
> Laurent
>
--
Thanks
Shreyansh

[-- Attachment #2: Type: text/html, Size: 4552 bytes --]