KVM: x86: get smi pending status correctly
authorJay Zhou <jianjay.zhou@huawei.com>
Mon, 18 Jan 2021 08:47:20 +0000 (16:47 +0800)
committerPaolo Bonzini <pbonzini@redhat.com>
Mon, 25 Jan 2021 23:52:07 +0000 (18:52 -0500)
commit1f7becf1b7e21794fc9d460765fe09679bc9b9e0
treeae4b92c0db859678c257fd7f9b8a3edb6d41cd08
parent98dd2f108e448988d91e296173e773b06fb978b8
KVM: x86: get smi pending status correctly

The injection process of smi has two steps:

    Qemu                        KVM
Step1:
    cpu->interrupt_request &= \
        ~CPU_INTERRUPT_SMI;
    kvm_vcpu_ioctl(cpu, KVM_SMI)

                                call kvm_vcpu_ioctl_smi() and
                                kvm_make_request(KVM_REQ_SMI, vcpu);

Step2:
    kvm_vcpu_ioctl(cpu, KVM_RUN, 0)

                                call process_smi() if
                                kvm_check_request(KVM_REQ_SMI, vcpu) is
                                true, mark vcpu->arch.smi_pending = true;

The vcpu->arch.smi_pending will be set true in step2, unfortunately if
vcpu paused between step1 and step2, the kvm_run->immediate_exit will be
set and vcpu has to exit to Qemu immediately during step2 before mark
vcpu->arch.smi_pending true.
During VM migration, Qemu will get the smi pending status from KVM using
KVM_GET_VCPU_EVENTS ioctl at the downtime, then the smi pending status
will be lost.

Signed-off-by: Jay Zhou <jianjay.zhou@huawei.com>
Signed-off-by: Shengen Zhuang <zhuangshengen@huawei.com>
Message-Id: <20210118084720.1585-1-jianjay.zhou@huawei.com>
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
arch/x86/kvm/x86.c