On 4/28/2016 7:47 PM, Huang, Kai wrote:
>
>>>>
>>>> +static u64 kvm_get_max_gfn(struct kvm *kvm)
>>>> +{
>>>> + int num_gfn = -1;
>>>> + struct kvm_memslots *slots = kvm_memslots(kvm);
>>>> + struct kvm_memory_slot *memslot;
>>>> + int gfn;
>>>> +
>>>> + kvm_for_each_memslot(memslot, slots) {
>>>> + gfn = memslot->base_gfn + memslot->npages;
>>>> + if (gfn > num_gfn)
>>>> + num_gfn = gfn;
>>>> + }
>>>> + return num_gfn - 1;
>>>> +}
>>>
>>> This might be wrong if there are devices assigned to the guest, as each
>>> assigned resource also occupies a memory slot. I assume you don't want
>>> to monitor assigned resources, do you?
>>
>> max gfn is only used for sanity check. When we are told to mark a page dirty,
>> or to reset the write trap for a page, we make sure that the gfn is valid.
>>
>
> When you were saying 'sanity check', I think you were meaning 'sanity
> check' whether the gfn is guest memory, from which point your above code
> might get a much *bigger* max_gfn than real one, as I explained above,
> which basically makes your 'sanity check' meaningless, therefore you can
> probably remove this code.
>
> Btw 'gfn is valid' is very vague as gfn for assigned resource is
> certainly valid as well.
>
> Thanks,
> -Kai
>
Good point. Will fix that.
Thanks for your feedback, Kai.
--
To unsubscribe from this list: send the line "unsubscribe kvm" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
