[Spice-devel] [Spice-devil] [RFC] register vmc interface early for name != vdagent [was: Re: Read data out of the Virtqueue]

[Charles.Tsai-蔡清海-研究發展部]

Hans,

	I don't know why there is a step 5 there. But it does hit the key point of the issue that I encountered.


-----Original Message-----
From: Hans de Goede [mailto:hdegoede at redhat.com] 
Sent: Friday, April 06, 2012 4:31 PM
To: Charles.Tsai-蔡清海-研究發展部
Cc: spice-devel at lists.freedesktop.org
Subject: Re: [Spice-devil] [RFC] register vmc interface early for name != vdagent [was: Re: Read data out of the Virtqueue]

Hi,

On 04/06/2012 10:19 AM, Charles.Tsai-蔡清海-研究發展部 wrote:
> Hans,
>
> The core issue is that the "printing channel" is brought up too late and the spice client miss the timing window to initialize the printing channel.
> All of what you said are being implemented in our code. Can you scroll back to my previous message to get the full picture of the problem?

Hmm, let me see if I understand the problem correctly. The problem is:

1) You use a standard virtio_console chardev frontend with a spicevmc backend
2) You connect spice-client to the vm while the guest is booting
3) When the spice-client connects the guest has not yet opened the virtio-serial
    port
4) Because of 3. the vmc has not been registered with spice-server yet when
    the client connects, so the printer channel is not part of the channels list
    at connect time
5) In order for the client to get the printer channel you need to disconnect and
    then reconnect?

Is that about right? In that case trying Alon's latest patch as a workaround is a good idea. I still don't think we want to do things this way in the long run, but it is a good start.

Regards,

Hans