On 06.06.14 00:32, Alexander Graf wrote:

On 05.06.14 19:33, Aneesh Kumar K.V wrote:

quoted

Alexander Graf [off-list ref] writes:

quoted

On 05.06.14 17:50, Aneesh Kumar K.V wrote:

quoted

Alexander Graf [off-list ref] writes:

quoted

On 05.06.14 14:08, Aneesh Kumar K.V wrote:

quoted

virtual time base register is a per VM, per cpu register that needs
to be saved and restored on vm exit and entry. Writing to VTB is not
allowed in the privileged mode.

Signed-off-by: Aneesh Kumar K.V <redacted>

.......

quoted

quoted

quoted

quoted

                break;

diff --git a/arch/powerpc/kvm/book3s_emulate.c 

b/arch/powerpc/kvm/book3s_emulate.c
index 3565e775b61b..1bb16a59dcbc 100644

--- a/arch/powerpc/kvm/book3s_emulate.c
+++ b/arch/powerpc/kvm/book3s_emulate.c

@@ -577,6 +577,9 @@ int kvmppc_core_emulate_mfspr_pr(struct 

kvm_vcpu *vcpu, int sprn, ulong *spr_val
             */
            *spr_val = vcpu->arch.spurr;
            break;
+    case SPRN_VTB:
+        *spr_val = vcpu->arch.vtb;

Doesn't this mean that vtb can be the same 2 when the guest reads 
it 2
times in a row without getting preempted?

But a mfspr will result in VM exit and that would make sure we
update vcpu->arch.vtb with the correct value.

We only call kvmppc_core_vcpu_put_pr() when we context switch away from
KVM, so it won't be updated, no?

kvmppc_copy_from_svcpu is also called from VM exit path 
(book3s_interrupt.S)

... where it will run into this code path:

        /*
         * Maybe we were already preempted and synced the svcpu from
         * our preempt notifiers. Don't bother touching this svcpu then.
         */
        if (!svcpu->in_use)
                goto out;

Scratch that. We're always calling this on entry/exit, so you're right.


Alex