Fine-grained dirty region tracking optimization issue about cuju

[coldfunction]

I guess I may find a mistake in KVM module. In the old version, the kvm_guest_time_update() function gives the code as below:

kvmft_page_dirty(v->kvm, vcpu->time >> PAGE_SHIFT, shared_kaddr, 0, NULL);

memcpy(shared_kaddr + vcpu->time_offset, &vcpu->hv_clock,
	       sizeof(vcpu->hv_clock));

But in the latest version, the kvm_guest_time_update() only execute one of the statements as below:

kvmft_page_dirty(v->kvm, vcpu->time >> PAGE_SHIFT, shared_kaddr, 0, NULL);
return 0;

and return, that's all.

The old version we can find that the shared_kaddr will change after we back up the page via kvmft_page_dirty. However, the shared_kaddr doesn't assign any value to the latest version.

I think there exists other code like this smell, using kvmft_page_dirty but don't consider the address is correct or not.

[PJTPJT]

I guess I may find a mistake in KVM module. In the old version, the kvm_guest_time_update() function gives the code as below:

kvmft_page_dirty(v->kvm, vcpu->time >> PAGE_SHIFT, shared_kaddr, 0, NULL);

memcpy(shared_kaddr + vcpu->time_offset, &vcpu->hv_clock,
	       sizeof(vcpu->hv_clock));

But in the latest version, the kvm_guest_time_update() only execute one of the statements as below:

kvmft_page_dirty(v->kvm, vcpu->time >> PAGE_SHIFT, shared_kaddr, 0, NULL);
return 0;

and return, that's all.

The old version we can find that the shared_kaddr will change after we back up the page via kvmft_page_dirty. However, the shared_kaddr doesn't assign any value to the latest version.

I think there exists other code like this smell, using kvmft_page_dirty but don't consider the address is correct or not.

In latest kvm_guest_time_update()
It has no kvmft_page_dirty
Actually, the kvmft_page_dirty in KVM all insert on kvm_write_guest_cached or kvm_write_guest