Gleb Natapov <gleb@xxxxxxxxxx> wrote on 17/04/2013 05:41:07 PM:
> On Wed, Apr 17, 2013 at 02:55:40PM +0300, Abel Gordon wrote:
> > Once L1 loads VMCS12 we enable shadow-vmcs capability and copy allthe
VMCS12
> > shadowed fields to the shadow vmcs. When we release the VMCS12, we
also
> > disable shadow-vmcs capability.
> >
> > Signed-off-by: Abel Gordon <abelg@xxxxxxxxxx>
> > ---
> > arch/x86/kvm/vmx.c | 11 +++++++++++
> > 1 file changed, 11 insertions(+)
> >
> > --- .before/arch/x86/kvm/vmx.c 2013-04-17 14:20:51.000000000 +0300
> > +++ .after/arch/x86/kvm/vmx.c 2013-04-17 14:20:51.000000000 +0300
> > @@ -5590,12 +5590,17 @@ static int nested_vmx_check_permission(s
> >
> > static inline void nested_release_vmcs12(struct vcpu_vmx *vmx)
> > {
> > + u32 exec_control;
> > if (enable_shadow_vmcs) {
> > if (vmx->nested.current_vmcs12 != NULL) {
> > /* copy to memory all shadowed fields in case
> > they were modified */
> > copy_shadow_to_vmcs12(vmx);
> > vmx->nested.sync_shadow_vmcs = false;
> > + exec_control = vmcs_read32(SECONDARY_VM_EXEC_CONTROL);
> > + exec_control &= ~SECONDARY_EXEC_SHADOW_VMCS;
> > + vmcs_write32(SECONDARY_VM_EXEC_CONTROL, exec_control);
> > + vmcs_write64(VMCS_LINK_POINTER, -1ull);
> > free_vmcs(vmx->nested.current_shadow_vmcs);
> > }
> > }
> > @@ -6084,6 +6089,7 @@ static int handle_vmptrld(struct kvm_vcp
> > gpa_t vmptr;
> > struct x86_exception e;
> > struct vmcs *shadow_vmcs;
> > + u32 exec_control;
> >
> > if (!nested_vmx_check_permission(vcpu))
> > return 1;
> > @@ -6140,6 +6146,11 @@ static int handle_vmptrld(struct kvm_vcp
> > /* init shadow vmcs */
> > vmcs_clear(shadow_vmcs);
> > vmx->nested.current_shadow_vmcs = shadow_vmcs;
> > + exec_control = vmcs_read32(SECONDARY_VM_EXEC_CONTROL);
> > + exec_control |= SECONDARY_EXEC_SHADOW_VMCS;
> > + vmcs_write32(SECONDARY_VM_EXEC_CONTROL, exec_control);
> > + vmcs_write64(VMCS_LINK_POINTER,
> > + __pa(shadow_vmcs));
> How hard would it be to disable shadowing for individual vmcs if shadow
> vmcs allocation fails? It bothers me a little that we can fail perfectly
> valid vmptrld() because of failed allocation.
That's really a corner case... IMHO, if we fail to allocate a shadow vmcs
we may experience bigger issues, like failing to allocate VMCS02.
Anyway, if we reuse the shadow vmcs as you requested, then we can allocate
the shadow vmcs once in handle_vmon. In this case, handle_vmon will fail
and
not handle_vmptrld.
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
