Re: [PATCH 02/10] KVM: x86: APICv: fix race in kvm_request_apicv_update on SVM

[Paolo Bonzini]

On 23/06/21 13:29, Maxim Levitsky wrote:
> +	kvm_block_guest_entries(kvm);
> +
>   	trace_kvm_apicv_update_request(activate, bit);
>   	if (kvm_x86_ops.pre_update_apicv_exec_ctrl)
>   		static_call(kvm_x86_pre_update_apicv_exec_ctrl)(kvm, activate);
> @@ -9243,6 +9245,8 @@ void kvm_request_apicv_update(struct kvm *kvm, bool activate, ulong bit)
>   	except = kvm_get_running_vcpu();
>   	kvm_make_all_cpus_request_except(kvm, KVM_REQ_APICV_UPDATE,
>   					 except);
> +
> +	kvm_allow_guest_entries(kvm);

Doesn't this cause a busy loop during synchronize_rcu?  It should be 
possible to request the vmexit of other CPUs from 
avic_update_access_page, and do a lock/unlock of kvm->slots_lock to wait 
for the memslot to be updated.

(As an aside, I'd like to get rid of KVM_REQ_MCLOCK_IN_PROGRESS in 5.15...).

Paolo