KVM: PPC: set IN_GUEST_MODE before checking requests

Avoid a race as described in the code comment.

Also remove a related smp_wmb() from booke's kvmppc_prepare_to_enter().
I can't see any reason for it, and the book3s_pr version doesn't have it.

Signed-off-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Alexander Graf <agraf@suse.de>

2 files changed, 9 insertions, 6 deletions

diff --git a/arch/powerpc/kvm/booke.c b/arch/powerpc/kvm/booke.c
index 69d047c22d2..3d1f35dc786 100644
--- a/arch/powerpc/kvm/booke.c
+++ b/arch/powerpc/kvm/booke.c
@@ -674,7 +674,6 @@ int kvmppc_vcpu_run(struct kvm_run *kvm_run, struct kvm_vcpu *vcpu)
 
 out:
         vcpu->mode = OUTSIDE_GUEST_MODE;
-        smp_wmb();
         return ret;
 }
 
diff --git a/arch/powerpc/kvm/powerpc.c b/arch/powerpc/kvm/powerpc.c
index 6002ea938a4..deb0d596d81 100644
--- a/arch/powerpc/kvm/powerpc.c
+++ b/arch/powerpc/kvm/powerpc.c
@@ -78,7 +78,16 @@ int kvmppc_prepare_to_enter(struct kvm_vcpu *vcpu)
                         break;
                 }
 
+                vcpu->mode = IN_GUEST_MODE;
+
+                /*
+                 * Reading vcpu->requests must happen after setting vcpu->mode,
+                 * so we don't miss a request because the requester sees
+                 * OUTSIDE_GUEST_MODE and assumes we'll be checking requests
+                 * before next entering the guest (and thus doesn't IPI).
+                 */
                 smp_mb();
+
                 if (vcpu->requests) {
                         /* Make sure we process requests preemptable */
                         local_irq_enable();
@@ -111,11 +120,6 @@ int kvmppc_prepare_to_enter(struct kvm_vcpu *vcpu)
 #endif
 
                 kvm_guest_enter();
-
-                /* Going into guest context! Yay! */
-                vcpu->mode = IN_GUEST_MODE;
-                smp_wmb();
-
                 break;
         }