Re: [PATCH v2 3/3] KVM: x86 emulator: Allow PM/VM86 switch during task switch

[Avi Kivity <avi@redhat.com>]

On 01/31/2012 11:37 AM, Gleb Natapov wrote:
> On Mon, Jan 30, 2012 at 04:26:06PM +0100, Kevin Wolf wrote:
> > >> Would that be enough or would we have to avoid clearing it in all other
> > >> places as well? Where would it be initialised if it's not enough?
> > > 
> > > Maybe vmx_vcpu_reset().
> > 
> > Do all CPL changes go through set_cr0/segment/rflags/cpl? I guess yes,
> > so initialising on reset and keeping it valid all the time should be
> > possible indeed.
> > 
> CPL can be changed while guest is running. SVM saves it for us in cpl
> field. VMX does not, so we either will have to update cpl on each exit
> (cpl = cs & 3) or somehow mark it not up-to-date and recalculate on
> access. Can VMX exit while cpl != cs & 3 or can this happen only during
> emulation? If it can we cannot know real cpl after exit.
>

Perhaps it can, with unrestricted guests, but I think we don't allow
those conditions (we trap cr0 writes).

-- 
error compiling committee.c: too many arguments to function