[libvirt] vTPM support in libvirt

[Andreas Sommer]

I agree on ignoring "backend" for now. The "instance" parameter defines 
a vTPM ID associated to that domain. There's a file "vtpm.db" which 
lists all mappings between domain UUID and vTPM ID, which means as long 
as you set a UUID for each of your domains, the correct vTPM is selected 
automatically (this is important for loading the last vTPM state).

Can you please give me a short introduction on how to add this feature 
to libvirt? I know how to check out the code and how to change the 
domain RelaxNG schema, but where do I need to change the source code? 
Oh, and how do I need to configure it in order to install it on a 
machine (I guess "./configure --prefix=???" is important?!).

Best regards
    Andreas

Daniel P. Berrange wrote:
> On Thu, Jun 25, 2009 at 09:16:26AM +0100, Andreas Sommer wrote:
>   
>> I'm wondering if there will be vTPM support in libvirt in the near 
>> future?! Xen does support it already with the configuration "vtpm = 
>> ['instance=1,backend=0']", for example.
>>
>> So it would be great if the libvirt XML format supported it, too... For 
>> example like this:
>>
>> <devices>
>>    <vtpm instance="1" backend="xxx" />
>> </devices>
>>
>> Both attributes are optional. The backend attribute is a VM ID (on Xen, 
>> only zero for dom0 is supported) and could be implemented as a UUID?!
>>     
>
> I'd just ignore 'backend' for now - none of the other existing devices
> suport anything other than dom0 as the backend, so its no loss to assume
> dom0 for TPM too.
>
> What is 'instance' ?
>
> For element I'd prefer to just call it '<tpm>' - the 'v' is redundant
> since every device is virtual here :-)
>
>   
>> What do you think? Are there any efforts to introducing that?
>>     
>
> No one has ever asked for it before, which is why we've not supported
> this to date. I don't have any objection to supporting it, so patches
> would be welcomed.
>
> Regards,
> Daniel
>