Re: [PATCH 3/13] KVM: kvm data structures

[Avi Kivity <avi@qumranet.com>]

Arnd Bergmann wrote:
> On Thursday 26 October 2006 19:24, Avi Kivity wrote:
>   
>> +struct kvm {
>> +       spinlock_t lock; /* protects everything except vcpus */
>> +       int nmemslots;
>> +       struct kvm_memory_slot memslots[KVM_MEMORY_SLOTS];
>> +       struct list_head active_mmu_pages;
>> +       struct kvm_vcpu vcpus[KVM_MAX_VCPUS];
>> +       int memory_config_version;
>> +       int busy;
>> +};
>>     
>
> Assuming that you move to the host-user == guest-real memory
> model, will this data structure still be needed? It would
> be really nice if a guest could simply consist of a number
> of vcpu structures that happen to be used from threads in the
> same process address space, but I find it hard to tell if
> that is realistic.
>   

We'd still need the shadow page table data structures (or the nested 
page tables pgd).

-- 
Do not meddle in the internals of kernels, for they are subtle and quick to panic.