KVM: Get rid of return value from kvm_arch_create_vm_debugfs()

Message ID 20240216155941.2029458-1-oliver.upton@linux.dev
State New
Headers
Series KVM: Get rid of return value from kvm_arch_create_vm_debugfs() |

Commit Message

Oliver Upton Feb. 16, 2024, 3:59 p.m. UTC
  The general expectation with debugfs is that any initialization failure
is nonfatal. Nevertheless, kvm_arch_create_vm_debugfs() allows
implementations to return an error and kvm_create_vm_debugfs() allows
that to fail VM creation.

Change to a void return to discourage architectures from making debugfs
failures fatal for the VM. Seems like everyone already had the right
idea, as all implementations already return 0 unconditionally.

Signed-off-by: Oliver Upton <oliver.upton@linux.dev>
---

Compile-tested on arm64, powerpc, and x86 since I don't trust myself to
even get this simple patch right.

 arch/powerpc/kvm/powerpc.c | 3 +--
 arch/x86/kvm/debugfs.c     | 3 +--
 include/linux/kvm_host.h   | 2 +-
 virt/kvm/kvm_main.c        | 8 ++------
 4 files changed, 5 insertions(+), 11 deletions(-)


base-commit: 6613476e225e090cc9aad49be7fa504e290dd33d
  

Comments

Marc Zyngier Feb. 16, 2024, 5:13 p.m. UTC | #1
On Fri, 16 Feb 2024 15:59:41 +0000,
Oliver Upton <oliver.upton@linux.dev> wrote:
> 
> The general expectation with debugfs is that any initialization failure
> is nonfatal. Nevertheless, kvm_arch_create_vm_debugfs() allows
> implementations to return an error and kvm_create_vm_debugfs() allows
> that to fail VM creation.
> 
> Change to a void return to discourage architectures from making debugfs
> failures fatal for the VM. Seems like everyone already had the right
> idea, as all implementations already return 0 unconditionally.
> 
> Signed-off-by: Oliver Upton <oliver.upton@linux.dev>

Acked-by: Marc Zyngier <maz@kernel.org>

	M.
  
Paolo Bonzini Feb. 23, 2024, 3:31 p.m. UTC | #2
On Fri, Feb 16, 2024 at 5:00 PM Oliver Upton <oliver.upton@linux.dev> wrote:
>
> The general expectation with debugfs is that any initialization failure
> is nonfatal. Nevertheless, kvm_arch_create_vm_debugfs() allows
> implementations to return an error and kvm_create_vm_debugfs() allows
> that to fail VM creation.
>
> Change to a void return to discourage architectures from making debugfs
> failures fatal for the VM. Seems like everyone already had the right
> idea, as all implementations already return 0 unconditionally.
>
> Signed-off-by: Oliver Upton <oliver.upton@linux.dev>

Acked-by: Paolo Bonzini <pbonzini@redhat.com>

Feel free to place it in kvm-arm.

Paolo
  
Oliver Upton Feb. 23, 2024, 9:51 p.m. UTC | #3
On Fri, 16 Feb 2024 15:59:41 +0000, Oliver Upton wrote:
> The general expectation with debugfs is that any initialization failure
> is nonfatal. Nevertheless, kvm_arch_create_vm_debugfs() allows
> implementations to return an error and kvm_create_vm_debugfs() allows
> that to fail VM creation.
> 
> Change to a void return to discourage architectures from making debugfs
> failures fatal for the VM. Seems like everyone already had the right
> idea, as all implementations already return 0 unconditionally.
> 
> [...]

Applied to kvmarm/next, thanks!

[1/1] KVM: Get rid of return value from kvm_arch_create_vm_debugfs()
      https://git.kernel.org/kvmarm/kvmarm/c/284851ee5cae

--
Best,
Oliver
  

Patch

diff --git a/arch/powerpc/kvm/powerpc.c b/arch/powerpc/kvm/powerpc.c
index 23407fbd73c9..d32abe7fe6ab 100644
--- a/arch/powerpc/kvm/powerpc.c
+++ b/arch/powerpc/kvm/powerpc.c
@@ -2538,9 +2538,8 @@  void kvm_arch_create_vcpu_debugfs(struct kvm_vcpu *vcpu, struct dentry *debugfs_
 		vcpu->kvm->arch.kvm_ops->create_vcpu_debugfs(vcpu, debugfs_dentry);
 }
 
-int kvm_arch_create_vm_debugfs(struct kvm *kvm)
+void kvm_arch_create_vm_debugfs(struct kvm *kvm)
 {
 	if (kvm->arch.kvm_ops->create_vm_debugfs)
 		kvm->arch.kvm_ops->create_vm_debugfs(kvm);
-	return 0;
 }
diff --git a/arch/x86/kvm/debugfs.c b/arch/x86/kvm/debugfs.c
index 95ea1a1f7403..999227fc7c66 100644
--- a/arch/x86/kvm/debugfs.c
+++ b/arch/x86/kvm/debugfs.c
@@ -189,9 +189,8 @@  static const struct file_operations mmu_rmaps_stat_fops = {
 	.release	= kvm_mmu_rmaps_stat_release,
 };
 
-int kvm_arch_create_vm_debugfs(struct kvm *kvm)
+void kvm_arch_create_vm_debugfs(struct kvm *kvm)
 {
 	debugfs_create_file("mmu_rmaps_stat", 0644, kvm->debugfs_dentry, kvm,
 			    &mmu_rmaps_stat_fops);
-	return 0;
 }
diff --git a/include/linux/kvm_host.h b/include/linux/kvm_host.h
index 7e7fd25b09b3..9a45f673f687 100644
--- a/include/linux/kvm_host.h
+++ b/include/linux/kvm_host.h
@@ -1507,7 +1507,7 @@  bool kvm_arch_dy_runnable(struct kvm_vcpu *vcpu);
 bool kvm_arch_dy_has_pending_interrupt(struct kvm_vcpu *vcpu);
 int kvm_arch_post_init_vm(struct kvm *kvm);
 void kvm_arch_pre_destroy_vm(struct kvm *kvm);
-int kvm_arch_create_vm_debugfs(struct kvm *kvm);
+void kvm_arch_create_vm_debugfs(struct kvm *kvm);
 
 #ifndef __KVM_HAVE_ARCH_VM_ALLOC
 /*
diff --git a/virt/kvm/kvm_main.c b/virt/kvm/kvm_main.c
index 10bfc88a69f7..c681149c382a 100644
--- a/virt/kvm/kvm_main.c
+++ b/virt/kvm/kvm_main.c
@@ -1150,10 +1150,7 @@  static int kvm_create_vm_debugfs(struct kvm *kvm, const char *fdname)
 				    &stat_fops_per_vm);
 	}
 
-	ret = kvm_arch_create_vm_debugfs(kvm);
-	if (ret)
-		goto out_err;
-
+	kvm_arch_create_vm_debugfs(kvm);
 	return 0;
 out_err:
 	kvm_destroy_vm_debugfs(kvm);
@@ -1183,9 +1180,8 @@  void __weak kvm_arch_pre_destroy_vm(struct kvm *kvm)
  * Cleanup should be automatic done in kvm_destroy_vm_debugfs() recursively, so
  * a per-arch destroy interface is not needed.
  */
-int __weak kvm_arch_create_vm_debugfs(struct kvm *kvm)
+void __weak kvm_arch_create_vm_debugfs(struct kvm *kvm)
 {
-	return 0;
 }
 
 static struct kvm *kvm_create_vm(unsigned long type, const char *fdname)