Re: [patch 2/3] KVM: x86: disallow KVM_{SET,GET}_LAPIC without in kernel irqchip

[Gleb Natapov <gleb@redhat.com>]

On Wed, Oct 28, 2009 at 12:32:24PM +0200, Michael S. Tsirkin wrote:
> On Wed, Oct 28, 2009 at 12:30:39PM +0200, Gleb Natapov wrote:
> > On Wed, Oct 28, 2009 at 12:20:42PM +0200, Michael S. Tsirkin wrote:
> > > On Wed, Oct 28, 2009 at 11:32:00AM +0200, Avi Kivity wrote:
> > > > On 10/27/2009 07:50 PM, Michael S. Tsirkin wrote:
> > > >> Can the value of irqchip_in_kernel be changed by userspace
> > > >> after we have checked it? If yes, this check won't help ...
> > > >>    
> > > >
> > > > A change from false to true is possible, but not the reverse.
> > > 
> > > Hmm. If we want to rely on this, we have to play with
> > > memory barriers to write/read it. Doable, but hard to get right.
> > Why? If userspace is so racy that it tries to use vpic while other
> > thread creates it let it fail and burn in hell.
> 
> Yes but reading uninitialized memory in kernel can
> lead to host kernel crashes.
> 
A you concerned that arch.vpic pointer assignment will be seen before vpic
initialization? Yes this theoretically possible.

> > > Can we always have the irqchip object exist?
> > > It doesn't use a lot of memory, does it?
> > > Maybe have it inline, save an extra indirection on
> > > fastpath ...
> > > 
> > > > -- 
> > > > error compiling committee.c: too many arguments to function
> > > --
> > > To unsubscribe from this list: send the line "unsubscribe kvm" in
> > > the body of a message to majordomo@vger.kernel.org
> > > More majordomo info at  http://vger.kernel.org/majordomo-info.html
> > 
> > --
> > 			Gleb.

--
			Gleb.