Skip to contentSkip to conversation

KVM: arm/arm64: kvm_arch_destroy_vm cleanups

PATCH In review Andrew Jones 2 participants 4 messages Nov 27, 2017 – Dec 1, 2017 KVM arm64

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 KVM: arm/arm64: kvm_arch_destroy_vm cleanups

Recently commit b2c9a85dd75a ("KVM: arm/arm64: vgic: Move kvm_vgic_destroy call around") caught my eye. When I looked closer I saw that while it made the code saner, it wasn't changing anything. kvm_for_each_vcpu() checks for NULL kvm->vcpus[i], so there wasn't a NULL dereference being fixed, and because kvm_vgic_vcpu_destroy() was called by kvm_arch_vcpu_free() it was still getting called, just not by kvm_vgic_destroy() as intended. But now the call from kvm_arch_vcpu_free() is redundant, and while currently harmless, it should be removed in case kvm_vgic_vcpu_destroy() were ever to want to reference vgic state, as kvm_vgic_destroy() now comes before kvm_arch_vcpu_free(). Additionally the other architectures set kvm->online_vcpus to zero after freeing them. We might as well do that for ARM too.

---
 virt/kvm/arm/arm.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
virt/kvm/arm/arm.c +1 -1
index a6524ff27de4..c5bc79c4ccf7 100644
@@ -188,6 +188,7 @@ void kvm_arch_destroy_vm(struct kvm *kvm)
 			kvm->vcpus[i] = NULL;
 		}
 	}
+	atomic_set(&kvm->online_vcpus, 0);
 }
 
 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
@@ -296,7 +297,6 @@ void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
 {
 	kvm_mmu_free_memory_caches(vcpu);
 	kvm_timer_vcpu_terminate(vcpu);
-	kvm_vgic_vcpu_destroy(vcpu);
 	kvm_pmu_vcpu_destroy(vcpu);
 	kvm_vcpu_uninit(vcpu);
 	kmem_cache_free(kvm_vcpu_cache, vcpu);
Signature · 1 line
2.13.6
#1 lore raw

Hi Drew,

Quoted 5 lines · Andrew Jones

Recently commit b2c9a85dd75a ("KVM: arm/arm64: vgic: Move kvm_vgic_destroy call around") caught my eye. When I looked closer I saw that while it made the code saner, it wasn't changing anything. kvm_for_each_vcpu() checks for NULL kvm->vcpus[i], so there wasn't a NULL dereference being fixed, and because kvm_vgic_vcpu_destroy()

Quoted 8 lines

was called by kvm_arch_vcpu_free() it was still getting called, just not by kvm_vgic_destroy() as intended. But now the call from kvm_arch_vcpu_free() is redundant, and while currently harmless, it should be removed in case kvm_vgic_vcpu_destroy() were ever to want to reference vgic state, as kvm_vgic_destroy() now comes before kvm_arch_vcpu_free(). Additionally the other architectures set kvm->online_vcpus to zero after freeing them. We might as well do that for ARM too.

Could this commit message be rewritten to:

kvm_vgic_vcpu_destroy already gets called from kvm_vgic_destroy for each vcpu, so we don't have to call it from kvm_arch_vcpu_free.

Additionally the other architectures set kvm->online_vcpus to zero after freeing them. We might as well do that for ARM too.

?

Thanks,
-Christoffer

Quoted 29 lines
---
 virt/kvm/arm/arm.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
virt/kvm/arm/arm.c +1 -1
index a6524ff27de4..c5bc79c4ccf7 100644
@@ -188,6 +188,7 @@ void kvm_arch_destroy_vm(struct kvm *kvm)
 			kvm->vcpus[i] = NULL;
 		}
 	}
+	atomic_set(&kvm->online_vcpus, 0);
 }
 
 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
@@ -296,7 +297,6 @@ void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
 {
 	kvm_mmu_free_memory_caches(vcpu);
 	kvm_timer_vcpu_terminate(vcpu);
-	kvm_vgic_vcpu_destroy(vcpu);
 	kvm_pmu_vcpu_destroy(vcpu);
 	kvm_vcpu_uninit(vcpu);
 	kmem_cache_free(kvm_vcpu_cache, vcpu);
Signature · 1 line
2.13.6
#2 lore raw
Quoted 17 lines · Christoffer Dall

Hi Drew,

Quoted 5 lines · Andrew Jones

Recently commit b2c9a85dd75a ("KVM: arm/arm64: vgic: Move kvm_vgic_destroy call around") caught my eye. When I looked closer I saw that while it made the code saner, it wasn't changing anything. kvm_for_each_vcpu() checks for NULL kvm->vcpus[i], so there wasn't a NULL dereference being fixed, and because kvm_vgic_vcpu_destroy()

Quoted 8 lines

was called by kvm_arch_vcpu_free() it was still getting called, just not by kvm_vgic_destroy() as intended. But now the call from kvm_arch_vcpu_free() is redundant, and while currently harmless, it should be removed in case kvm_vgic_vcpu_destroy() were ever to want to reference vgic state, as kvm_vgic_destroy() now comes before kvm_arch_vcpu_free(). Additionally the other architectures set kvm->online_vcpus to zero after freeing them. We might as well do that for ARM too.

Quoted 7 lines

Could this commit message be rewritten to:

kvm_vgic_vcpu_destroy already gets called from kvm_vgic_destroy for each vcpu, so we don't have to call it from kvm_arch_vcpu_free.

Additionally the other architectures set kvm->online_vcpus to zero after freeing them. We might as well do that for ARM too.

Sure, I don't mind you removing the '-v' (verbose) from it. Should I respin? Or is that something you don't mind doing while applying?

Thanks,
drew

Quoted 37 lines

?

Thanks,
-Christoffer

Quoted 29 lines
---
 virt/kvm/arm/arm.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
virt/kvm/arm/arm.c +1 -1
index a6524ff27de4..c5bc79c4ccf7 100644
@@ -188,6 +188,7 @@ void kvm_arch_destroy_vm(struct kvm *kvm)
 			kvm->vcpus[i] = NULL;
 		}
 	}
+	atomic_set(&kvm->online_vcpus, 0);
 }
 
 int kvm_vm_ioctl_check_extension(struct kvm *kvm, long ext)
@@ -296,7 +297,6 @@ void kvm_arch_vcpu_free(struct kvm_vcpu *vcpu)
 {
 	kvm_mmu_free_memory_caches(vcpu);
 	kvm_timer_vcpu_terminate(vcpu);
-	kvm_vgic_vcpu_destroy(vcpu);
 	kvm_pmu_vcpu_destroy(vcpu);
 	kvm_vcpu_uninit(vcpu);
 	kmem_cache_free(kvm_vcpu_cache, vcpu);
Signature · 1 line
2.13.6
#3 lore raw
Quoted 20 lines · Andrew Jones
Quoted 19 lines · Christoffer Dall

Hi Drew,

Quoted 5 lines · Andrew Jones

Recently commit b2c9a85dd75a ("KVM: arm/arm64: vgic: Move kvm_vgic_destroy call around") caught my eye. When I looked closer I saw that while it made the code saner, it wasn't changing anything. kvm_for_each_vcpu() checks for NULL kvm->vcpus[i], so there wasn't a NULL dereference being fixed, and because kvm_vgic_vcpu_destroy()

Quoted 8 lines

was called by kvm_arch_vcpu_free() it was still getting called, just not by kvm_vgic_destroy() as intended. But now the call from kvm_arch_vcpu_free() is redundant, and while currently harmless, it should be removed in case kvm_vgic_vcpu_destroy() were ever to want to reference vgic state, as kvm_vgic_destroy() now comes before kvm_arch_vcpu_free(). Additionally the other architectures set kvm->online_vcpus to zero after freeing them. We might as well do that for ARM too.

Could this commit message be rewritten to:

Quoted 9 lines
Quoted 5 lines

kvm_vgic_vcpu_destroy already gets called from kvm_vgic_destroy for each vcpu, so we don't have to call it from kvm_arch_vcpu_free.

Additionally the other architectures set kvm->online_vcpus to zero after freeing them. We might as well do that for ARM too.

Sure, I don't mind you removing the '-v' (verbose) from it. Should I respin? Or is that something you don't mind doing while applying?

No need to respin, I already applied your patch with the adjusted commit message.

Thanks,
-Christoffer