Re: [PATCH v2 02/29] LoongArch: KVM: Implement kvm module related interface

[Paolo Bonzini]

On 2/20/23 07:57, Tianrui Zhao wrote:
> +	order = get_order(kvm_vector_size + kvm_enter_guest_size);
> +	addr = (void *)__get_free_pages(GFP_KERNEL, order);
> +	if (!addr) {
> +		free_percpu(vmcs);
> +		return -ENOMEM;
> +	}
> +
> +	memcpy(addr, kvm_vector_entry, kvm_vector_size);
> +	memcpy(addr + kvm_vector_size, kvm_enter_guest, kvm_enter_guest_size);
> +	flush_icache_range((unsigned long)addr, (unsigned long)addr +
> +				kvm_vector_size + kvm_enter_guest_size);
> +
> +	vpid_mask = read_csr_gstat();
> +	vpid_mask = (vpid_mask & CSR_GSTAT_GIDBIT) >> CSR_GSTAT_GIDBIT_SHIFT;
> +	if (vpid_mask)
> +		vpid_mask = GENMASK(vpid_mask - 1, 0);
> +
> +	for_each_possible_cpu(cpu) {
> +		context = per_cpu_ptr(vmcs, cpu);
> +		context->vpid_mask = vpid_mask;
> +		context->vpid_cache = context->vpid_mask + 1;
> +		context->last_vcpu = NULL;
> +		context->kvm_eentry = addr;
> +		context->kvm_enter_guest = addr + kvm_vector_size;
> +		context->page_order = order;
> +	}

A lot of these variables are constant across all pCPUs, any reason to 
have them in a per-CPU variable?  Likewise, since they are all the same 
as the constant global vmcs variable, why make them part of struct 
kvm_context instead of just making them globals?

Also, why does the world switch code need a copy?

Paolo