On 8/17/22 01:34, Sean Christopherson wrote:
Isn't freeing up the return from kvm_vcpu_check_block() unnecessary? Can't we
just do:
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 9f11b505cbee..ccb9f8bdeb18 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -10633,7 +10633,7 @@ static inline int vcpu_block(struct kvm_vcpu *vcpu)
if (hv_timer)
kvm_lapic_switch_to_hv_timer(vcpu);
- if (!kvm_check_request(KVM_REQ_UNHALT, vcpu))
+ if (!kvm_arch_vcpu_runnable(vcpu))
return 1;
}
which IMO is more intuitive and doesn't require reworking halt-polling (again).
This was my first idea indeed. However I didn't like calling
kvm_arch_vcpu_runnable() again and "did it schedule()" seemed to be a
less interesting result from kvm_vcpu_block() (and in fact
kvm_vcpu_halt() does not bother passing it up the return chain).
Paolo
