Re: [patch] fix kvm_arch_vcpu_ioctl_[gs]et_regs()

[Jes Sorensen <jes@sgi.com>]

Zhang, Xiantao wrote:
> Jes
>    Good work!  I ever talked about the issue with Avi and wanted to enable the logic with the following patch, and Avi thought it may lead to security issues, so deferred to now :) 
> 
> One comment:  We still need the logic to save and restore the vcpu's stack for vcpu resuming to the guest in the same environment.  Maybe you can allocate one buf to save it in kvm_reg structure.
> Xiantao

Hi Xiantao,

I am with Avi on this one - we shouldn't be passing in pointers like
that. Either we can increase the size of kvm_regs as you mention, but it
will become *huge* since the stack is 64KB, or we introduce a new ioctl
just to handle the stack.

Do you think it would be a problem having the second ioctl for this? I
would prefer breaking it into two to avoid the kernel having to allocate
a 128KB chunk for kvm_regs.

Cheers,
Jes