Re: [Qemu-devel] RFC adding ioctl's to virtserial/virtconsole

[Gerd Hoffmann <kraxel@redhat.com>]

   Hi,

>> /me wonders what the point of the 'backlog' struct element is then.
>
> Because it could be used for host event but let's ignore that for now.

I doubt using the same beast for both host and guest is going to fly ...

>> Yes, we can do that. I don't think it is useful. Oh, and it also
>> changes the command line interface. Todays ...
>>
>> qemu -chardev soemthing,id=foo \
>> -device virtserport,chardev=foo
>>
>> ... would turn into something like ...
>>
>> qemu -chardev something,if=foo \
>> -charsrv simple,chardev=foo,id=bar \
>> -device virtserport,charsrv=bar
>
> If we think this is useful, then we can find a way to make the command
> line syntax work. If we don't think it's useful, then there's no point
> in doing it.

Well, when using the new model everythere this is a moot point as 
-chardev would just define a CharServerState instead of a 
CharDriverState then.

>>> Spice would then implement it's own CharServerState and would use it to
>>> understand what state the session is in.
>>
>> Spice would basically (ab-)use it as event delivery mechanism.
>
> Can you explain what spice uses these events for?

spice-vmc code registers/unregisters the interface within the spice 
server.  So the interface is only activated in case the guest uses it. 
Spice client sees the interface being active or not and can act accordingly.

http://cgit.freedesktop.org/spice/qemu/tree/hw/spice-vmc.c?h=spice.v13#n128

>> Well. I disagree. Checking the state is needed nevertheless. The
>> places where virtio-serial checks port->state today it would have to
>> check whenever port->chardev is non-NULL then. The only difference is
>> that failures to do so might become a bit more obvious as qemu will
>> segfault due to the NULL pointer dereferences then. I still think this
>> isn't worth the effort though.
>
> But I think we ultimately need to switch to having the front-ends having
> a NULL check. Even beyond front-end initiated connect/disconnect,
> front-end's need to learn to deal with back-end initiated
> disconnect/connect.

I don't think we have to go with a NULL check.  Providing chr_is_*() 
functions to query state and adding asserts() to the chr_*() function 
should provide the same level of robustness IMHO.  Also having 
CharDriverStates come and go brings its own share of problems.

> If you look at something like the serial device's chardev usage, right
> now, it writes to the stream regardless of whether the back-end is
> connected and we have different semantics in each backend about what we
> do. It would be far better to just expose the fact that the backend
> isn't connected to the device such that it can either present that to
> the guest or make it's own decision about what to do.

Indeed.

> I think the model where we always write to a chardev is fundamentally
> broken. Sending life cycle events over an always open stream is even
> more broken and I think it's a good opportunity to introduce life cycle
> awareness into the API (especially since it can be done as a pretty
> small incremental change).

The chardevs API can use a major overhaul, no question.  I doubt that 
creating CharServerState helps here much though.  Especially I don't see 
how that could be used for *both* Host and Guest side.

cheers,
   Gerd