ARM/ARM64: KVM: Emulate PSCI v0.2 SYSTEM_OFF and SYSTEM_RESET
authorAnup Patel <anup.patel@linaro.org>
Tue, 29 Apr 2014 05:54:20 +0000 (11:24 +0530)
committerChristoffer Dall <christoffer.dall@linaro.org>
Wed, 30 Apr 2014 11:18:58 +0000 (04:18 -0700)
The PSCI v0.2 SYSTEM_OFF and SYSTEM_RESET functions are system-level
functions hence cannot be fully emulated by in-kernel PSCI emulation code.

To tackle this, we forward PSCI v0.2 SYSTEM_OFF and SYSTEM_RESET function
calls from vcpu to user space (i.e. QEMU or KVMTOOL) via kvm_run structure
using KVM_EXIT_SYSTEM_EVENT exit reasons.

Signed-off-by: Anup Patel <anup.patel@linaro.org>
Signed-off-by: Pranavkumar Sawargaonkar <pranavkumar@linaro.org>
Reviewed-by: Christoffer Dall <christoffer.dall@linaro.org>
Acked-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <christoffer.dall@linaro.org>
arch/arm/kvm/psci.c

index 14e6fa6c8e35d8295e0b634185f32ea68f30bf91..59362131b79f015e5ac4613cf609556401e7e9fe 100644 (file)
@@ -85,6 +85,23 @@ static unsigned long kvm_psci_vcpu_on(struct kvm_vcpu *source_vcpu)
        return PSCI_RET_SUCCESS;
 }
 
+static void kvm_prepare_system_event(struct kvm_vcpu *vcpu, u32 type)
+{
+       memset(&vcpu->run->system_event, 0, sizeof(vcpu->run->system_event));
+       vcpu->run->system_event.type = type;
+       vcpu->run->exit_reason = KVM_EXIT_SYSTEM_EVENT;
+}
+
+static void kvm_psci_system_off(struct kvm_vcpu *vcpu)
+{
+       kvm_prepare_system_event(vcpu, KVM_SYSTEM_EVENT_SHUTDOWN);
+}
+
+static void kvm_psci_system_reset(struct kvm_vcpu *vcpu)
+{
+       kvm_prepare_system_event(vcpu, KVM_SYSTEM_EVENT_RESET);
+}
+
 int kvm_psci_version(struct kvm_vcpu *vcpu)
 {
        if (test_bit(KVM_ARM_VCPU_PSCI_0_2, vcpu->arch.features))
@@ -95,6 +112,7 @@ int kvm_psci_version(struct kvm_vcpu *vcpu)
 
 static int kvm_psci_0_2_call(struct kvm_vcpu *vcpu)
 {
+       int ret = 1;
        unsigned long psci_fn = *vcpu_reg(vcpu, 0) & ~((u32) 0);
        unsigned long val;
 
@@ -114,13 +132,35 @@ static int kvm_psci_0_2_call(struct kvm_vcpu *vcpu)
        case PSCI_0_2_FN64_CPU_ON:
                val = kvm_psci_vcpu_on(vcpu);
                break;
+       case PSCI_0_2_FN_SYSTEM_OFF:
+               kvm_psci_system_off(vcpu);
+               /*
+                * We should'nt be going back to guest VCPU after
+                * receiving SYSTEM_OFF request.
+                *
+                * If user space accidently/deliberately resumes
+                * guest VCPU after SYSTEM_OFF request then guest
+                * VCPU should see internal failure from PSCI return
+                * value. To achieve this, we preload r0 (or x0) with
+                * PSCI return value INTERNAL_FAILURE.
+                */
+               val = PSCI_RET_INTERNAL_FAILURE;
+               ret = 0;
+               break;
+       case PSCI_0_2_FN_SYSTEM_RESET:
+               kvm_psci_system_reset(vcpu);
+               /*
+                * Same reason as SYSTEM_OFF for preloading r0 (or x0)
+                * with PSCI return value INTERNAL_FAILURE.
+                */
+               val = PSCI_RET_INTERNAL_FAILURE;
+               ret = 0;
+               break;
        case PSCI_0_2_FN_CPU_SUSPEND:
        case PSCI_0_2_FN_AFFINITY_INFO:
        case PSCI_0_2_FN_MIGRATE:
        case PSCI_0_2_FN_MIGRATE_INFO_TYPE:
        case PSCI_0_2_FN_MIGRATE_INFO_UP_CPU:
-       case PSCI_0_2_FN_SYSTEM_OFF:
-       case PSCI_0_2_FN_SYSTEM_RESET:
        case PSCI_0_2_FN64_CPU_SUSPEND:
        case PSCI_0_2_FN64_AFFINITY_INFO:
        case PSCI_0_2_FN64_MIGRATE:
@@ -132,7 +172,7 @@ static int kvm_psci_0_2_call(struct kvm_vcpu *vcpu)
        }
 
        *vcpu_reg(vcpu, 0) = val;
-       return 1;
+       return ret;
 }
 
 static int kvm_psci_0_1_call(struct kvm_vcpu *vcpu)