[Qemu-devel] Re: [PATCH 2/5] virtio-console: Add support for multiple ports for generic guest-host communication

[Amit Shah]

On (Fri) Sep 11 2009 [11:38:36], Anthony Liguori wrote:
> Amit Shah wrote:
>> There wasn't any consensus; the discussion just ended abruptly.
>>   
>
> At least Gerd and I were in violent agreement.  I don't think anyone  
> disagreed.

So how's suggestion about udev making symlinks based on the ports ids?

>>> If a user does:
>>>
>>> qemu -M pc-0.11.0 -virtiocon vc -virtiocon vc
>>>
>>> This patch will break that guest.  I think the best solution to this 
>>> is      
>>
>> If there are multiple virtiocon devices specified, the first one will
>> default to port #0. The second one will error out saying port 0 is
>> taken. Isn't different from the current behaviour though.
>>   
>
> But that's a problem for -M pc-0.11.0.  We need to have a way to specify  
> a command line syntax with -M pc-0.11.0 such that the machine created  
> matches what would have been created with qemu-0.11.0.  Otherwise, we  
> have no hope of being able to do live migration or save/restore between  
> versions.

I don't actually know if I understand this completely. However, with the
current behaviour, qemu just exits saying something like too many virtio
consoles since the current MAX_VIRTIO_CONSOLES is defined to 1.

>>> We don't want to add PCI dependency to virtio console.  It isn't 
>>> always  used on platforms with PCI.
>>>     
>>
>> OK; there's a place where I need the qdev pointer from the PCIDevice.
>> Any other way of obtaining that?
>>   
>
> Why do you need the qdev pointer?

To init a chardev in the port hotplug case, qdev_init_chardev() needs
the qdev pointer.

>>> Need to integrate with QemuOpts.
>>
>> Wondering if it already entered master.. I'll sync up with kraxel if
>> not.
>
> Yup, it's been there for a while.

I remembered seeing it in your queue. Anyway it's simple enough and I'll
do it.

>>>> +    register_savevm("virtio-console", -1, 2, virtio_console_save, 
>>>> virtio_console_load, s);
>>>>         
>>> Should integrate with VMState.
>>>     
>>
>> I don't think virtio devices have been converted yet.
>>   
>
> I think that's next on Juan's list.

Right; so we'll talk about that.

                Amit