Re: [PATCH 1/8] KVM: x86: avoid loading a vCPU after .vm_destroy was called

From: Sean Christopherson <seanjc@google.com>
To: Maxim Levitsky <mlevitsk@redhat.com>
Cc: kvm@vger.kernel.org, Jim Mattson <jmattson@google.com>,
	Paolo Bonzini <pbonzini@redhat.com>,
	Wanpeng Li <wanpengli@tencent.com>,
	Borislav Petkov <bp@alien8.de>, Joerg Roedel <joro@8bytes.org>,
	Ingo Molnar <mingo@redhat.com>,
	Suravee Suthikulpanit <suravee.suthikulpanit@amd.com>,
	x86@kernel.org, "H. Peter Anvin" <hpa@zytor.com>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	linux-kernel@vger.kernel.org,
	Vitaly Kuznetsov <vkuznets@redhat.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	stable@vger.kernel.org
Subject: Re: [PATCH 1/8] KVM: x86: avoid loading a vCPU after .vm_destroy was called
Date: Wed, 30 Mar 2022 00:27:55 +0000	[thread overview]
Message-ID: <YkOkCwUgMD1SVfaD@google.com> (raw)
In-Reply-To: <20220322172449.235575-2-mlevitsk@redhat.com>

On Tue, Mar 22, 2022, Maxim Levitsky wrote:
> This can cause various unexpected issues, since VM is partially
> destroyed at that point.
> 
> For example when AVIC is enabled, this causes avic_vcpu_load to
> access physical id page entry which is already freed by .vm_destroy.

Hmm, the SEV unbinding of ASIDs should be done after MMU teardown too (which your
patch also does).

> 
> Fixes: 8221c1370056 ("svm: Manage vcpu load/unload when enable AVIC")
> Cc: stable@vger.kernel.org
> 
> Signed-off-by: Maxim Levitsky <mlevitsk@redhat.com>
> ---
>  arch/x86/kvm/x86.c | 10 +++-------
>  1 file changed, 3 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index d3a9ce07a565..ba920e537ddf 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -11759,20 +11759,15 @@ static void kvm_unload_vcpu_mmu(struct kvm_vcpu *vcpu)
>  	vcpu_put(vcpu);
>  }
>  
> -static void kvm_free_vcpus(struct kvm *kvm)
> +static void kvm_unload_vcpu_mmus(struct kvm *kvm)
>  {
>  	unsigned long i;
>  	struct kvm_vcpu *vcpu;
>  
> -	/*
> -	 * Unpin any mmu pages first.
> -	 */
>  	kvm_for_each_vcpu(i, vcpu, kvm) {
>  		kvm_clear_async_pf_completion_queue(vcpu);
>  		kvm_unload_vcpu_mmu(vcpu);
>  	}
> -
> -	kvm_destroy_vcpus(kvm);
>  }
>  
>  void kvm_arch_sync_events(struct kvm *kvm)
> @@ -11878,11 +11873,12 @@ void kvm_arch_destroy_vm(struct kvm *kvm)
>  		__x86_set_memory_region(kvm, TSS_PRIVATE_MEMSLOT, 0, 0);
>  		mutex_unlock(&kvm->slots_lock);
>  	}
> +	kvm_unload_vcpu_mmus(kvm);
>  	static_call_cond(kvm_x86_vm_destroy)(kvm);
>  	kvm_free_msr_filter(srcu_dereference_check(kvm->arch.msr_filter, &kvm->srcu, 1));
>  	kvm_pic_destroy(kvm);
>  	kvm_ioapic_destroy(kvm);
> -	kvm_free_vcpus(kvm);
> +	kvm_destroy_vcpus(kvm);

Rather than split kvm_free_vcpus(), can we instead move the call to svm_vm_destroy()
by adding a second hook, .vm_teardown(), which is needed for TDX?  I.e. keep VMX
where it is by using vm_teardown, but effectively move SVM?

https://lore.kernel.org/all/1fa2d0db387a99352d44247728c5b8ae5f5cab4d.1637799475.git.isaku.yamahata@intel.com

>  	kvfree(rcu_dereference_check(kvm->arch.apic_map, 1));
>  	kfree(srcu_dereference_check(kvm->arch.pmu_event_filter, &kvm->srcu, 1));
>  	kvm_mmu_uninit_vm(kvm);
> -- 
> 2.26.3
> 