Re: [PATCH v2 3/3] KVM: SVM: Add irqchip_split() checks before enabling AVIC

From	Paolo Bonzini <pbonzini@redhat.com>
Newsgroups	linux.kernel
Subject	Re: [PATCH v2 3/3] KVM: SVM: Add irqchip_split() checks before enabling AVIC
Date	2017-09-15 23:20 +0200
Message-ID	<uq6bM-4U-19@gated-at.bofh.it> (permalink)
References	<upEfw-6KR-9@gated-at.bofh.it> <upEfw-6KR-11@gated-at.bofh.it> <upEfw-6KR-7@gated-at.bofh.it> <uq6bM-4U-21@gated-at.bofh.it> <uq6bM-4U-23@gated-at.bofh.it>
Organization	linux.* mail to news gateway

Show all headers | View raw

On 15/09/2017 21:21, Suravee Suthikulpanit wrote:
>>>>
>>>>
>>>>  static bool svm_get_enable_apicv(struct kvm_vcpu *vcpu)
>>>>  {
>>> A close contender was pr_info_once() here:
>>>
>>>     if (avic && !irqchip_split(vcpu->kvm))
>>>         pr_info_once(...)
>>>
>>
>> Looks good.
> 
> Actually, thinking about it again, this would not work either since
> pr_xxx_once() would only print the message once per loading of kvm_amd
> module. However, we would prefer the message to be printed per VM
> initialization. I also tried adding the check and print this message in
> the kvm_x86_ops.vm_init(), but this also does not work since the
> vm_init() function is called before the
> kvm_vm_ioctl_enable_cap(KVM_CAP_SPLIT_IRQCHIP).
> 
> Therefore, pr_info() might be better here, even though this would get
> print per VCPU initialization. Any other suggestions?

My suggestion is to make it pr_info_once.  No one looks at dmesg anyway
unless things go very wrong.

Paolo

Back to linux.kernel | Previous | Next — Previous in thread | Find similar | Unroll thread

Thread

Re: [PATCH v2 3/3] KVM: SVM: Add irqchip_split() checks before  enabling AVIC Radim Krčmář <rkrcmar@redhat.com> - 2017-09-14 17:30 +0200
  Re: [PATCH v2 3/3] KVM: SVM: Add irqchip_split() checks before  enabling AVIC Paolo Bonzini <pbonzini@redhat.com> - 2017-09-15 23:20 +0200

csiph-web