Re: [PATCH 34/43] audio: rename AudioState -> AudioBE

[Paolo Bonzini <pbonzini@redhat.com>]

On 10/21/25 12:40, Marc-André Lureau wrote:
>>> BE and FE aren't used in existing code though.  Looking at current use:
> 
> There is a top-level backends/ though. I meant to say that it's
> usually clearer for everyone when we talk about "backend" vs
> "frontend" (or not?), and I wish we would follow that naming more in
> the code.

Yes, but the front-end should have a *pointer* to a backend.

>>> - BlockDriverState*/BlockBackend*
> 
> I am not very familiar with the block subsystem, but it is seems
> BlockBackend is actually the backend and BlockDriverState is
> implementation data detail. Correct me if I'm wrong

Yes, note the "*".  What I meant above is "the frontend has a 
BlockBackend*, which itself has the internal BlockDriverState*".  Same 
for character and NICs below.

>>> - Chardev*/CharBackend
>>>
> 
> Those are actually always confusing... and I remember I attempted to
> rename it in the past... CharBackend is the front-end side..

Yes, they are confusing.  I'm not sure why it failed, but 
s/CharBackend/CharFrontend/ does get my vote.

>> QEMUSoundCard's name field is basically unused, and QEMUSoundCard should
>> be just an Audiodev*.  There's no need for a separate AudioFE struct.
> 
> For audio, it's pretty obvious that we have "backend" object to
> implement the host audio subsystem details, while "QEMUSoundCard" is
> the "frontend", or "client/user" context.

I agree, but unlike character devices, QEMUSoundCard is not needed at 
all.  So you can drop it and only leave Audiodev*, or AudioBackend* (but 
I prefer Audiodev* which matches both the option name and Chardev*).

Paolo