Re: [PATCH v5 4/4] KVM: x86: Add support for local interrupt requests from userspace

[Paolo Bonzini <pbonzini@redhat.com>]

On 28/07/2015 21:06, Steve Rutherford wrote:
>>> > > +static inline int pic_in_kernel(struct kvm *kvm)
>>> > > +{
>>> > > +	int ret;
>>> > > +
>>> > > +	ret = (pic_irqchip(kvm) != NULL);
>>> > > +	smp_rmb();
>> > 
>> > What does this memory barrier pair with?  I don't think it's necessary.
> To be honest, it's probably not necessary. I couldn't find why
> irqchip_in_kernel (which this function is more or less a copy of)
> needed it's memory barrier, so I cargo culted this one in.

It's for stuff like injecting an interrupt before any CPU is created,
while another thread is doing KVM_CREATE_IRQCHIP.  In your case a VCPU
has been created so you don't need it (the synchronization point is the
mutex_lock(&kvm->lock) in kvm_vm_ioctl_create_vcpu).

Paolo