Re: [PATCH 2/2] KVM: VMX: Initialize vm86 TSS only once.

[Gleb Natapov <gleb@redhat.com>]

On Sun, Feb 27, 2011 at 06:04:16PM +0200, Avi Kivity wrote:
> On 02/27/2011 05:58 PM, Avi Kivity wrote:
> >
> >>The problem with using top of slot
> >>zero is that this memory is available for guest use and we do not even
> >>put it into e820 map as far as I see. Also there are patches floating
> >>around that re-arrange memslots or even put them in a tree. They will
> >>break old guests too.
> >
> >Well, slot 0 still exists even if it is moved somewhere else.
> >
> >Something we can do is put the tss slot just below the highest
> >slot that is still below 4G, and hope there is no mmio there.
> >Once the user issues KVM_SET_TSS_ADDR, use that.  We'll have to
> >keep juggling that slot as the user creates more slots, icky.
> >
> 
> Or we can keep the old behaviour.  If KVM_SET_TSS_ADDR hasn't been
> called by the time of the first entry into real mode (the first
> KVM_CREATE_VCPU?), use the top of the first slot.
> 
Do we require that KVM_SET_TSS_ADDR is called before first KVM_CREATE_VCPU?
We may still break some theoretical userspaces this way.
 
> We can avoid the SMP problem by initializing the memory in a single
> pass, writing each byte exactly once with its final value.  This way
> concurrent initialization doesn't corrupt an in-use TSS.
> 
Sounds hackish, but may work. Doing so will make entering pmode much
more slow.
 
--
			Gleb.