[GSoC Paravirtualization] Identify the sensitive register in X86 platform.

[Philipp Eppelt]

On 06/02/2014 06:58 PM, Youren Shen wrote:
> Hi, Everyone:
> 
> Since last week,I have improved the hypercall, build a new bsp named
> testHypercall to test it.Here is the commit.[1], [2], [3]. I also write
> a blog about it[4]. I have send a email to pok email-list about the change.
> 
> Then I have read the mail-list about the thread "Including
> paravirtualization headers in rtems", in my understanding, we are
> discussing about how to make call from the RTEMS to POK, and also about
> how to compile the two system into together.
> As far as I can see, it seems that we should implement some fake
> register in RTEMS and give some virtualization API in RTEMS. Is that right?
> 
> Also I have post a blog[5] about the sensitive register in X86 platform,
> and analysis the approaching about implementing virtual cpu  in XEN. In
> my opinion, If we want to implement the virtual cpu in RTEMS, we should
> not concern about sensitive register only, but should consider the
> context of cpu as an essential part.
The point is not to virtualize the privileged registers. But everything
else. We also don't care for SMP processors right now. Keep it in mind,
but SMP is beyond the scope of this project.

Also, the vCPU is for RTEMS, but has to be implemented in POK.
Keep on reading, there are a lot more other things to be aware of.

Cheers,
Philipp