Re: [PATCH] virtio_console: Add support for multiple ports for generic guest and host communication

[Anthony Liguori <anthony@codemonkey.ws>]

Amit Shah wrote:
> On (Tue) Sep 15 2009 [07:57:10], Anthony Liguori wrote:
>   
>> Amit Shah wrote:
>>     
>>> Hey Greg,
>>>
>>> Can you tell me how this could work out -- each console port could have
>>> a "role" string associated with it (obtainable from the invoking qemu
>>> process in case of qemu/kvm). Something that I have in mind currently
>>> is:
>>>
>>> $ qemu-kvm ... -virtioconsole role=org/qemu/clipboard
>>>
>>> and then the guest kernel sees the string, and puts the
>>> "org/qemu/clipboard" in some file in sysfs. Guest userspace should then
>>> be able to open and read/write to
>>>
>>> /dev/virtio_console/org/qemu/clipboard
>>>   
>>>       
>> That's probably not what we want.  I imagine what we want is:
>>
>> /dev/ttyV0
>> /dev/ttyV1
>> /dev/ttyVN
>>
>> And then we want:
>>
>> /sys/class/virtio-console/ttyV0/name -> "org.qemu.clipboard"
>>
>> Userspace can detect when new virtio-consoles appear via udev events.   
>> When it sees a new ttyVN, it can then look in sysfs to discover it's 
>> name.
>>     
>
> OK; but that's kind of roundabout isn't it? An application, instead of
> watching for the console port it's interested in, has to instead monitor
> all the ports.
>   

If you wanted to use /dev/virtio/org/qemu/clipboard you still have the 
same problem.  You have to use udev or inotify to listen for a new file 
in a directory.

The /dev/ path may look nicer from a high level, but the code ends up 
being roughly the same for either approach.  What I propose has the 
advantage of looking like existing subsystems.  It also avoids encoding 
device information in the device name.

> So in effect there has to be one app monitoring for new ports and then
> that app exec'ing the corresponding app meant for that port.
>   

I think if you think through both models, they end up looking the same.

Regards,

Anthony Liguroi