Re: [PATCH] KVM: x86/mmu: x86: Don't overflow lpage_info when checking attributes

[Dongli Zhang]

On 3/13/24 09:25, Edgecombe, Rick P wrote:
> On Wed, 2024-03-13 at 02:49 -0700, Dongli Zhang wrote:
>> The memslot id=10 has:
>> - base_gfn=1048576
>> - npages=1024
>>
>> Therefore, "level - 1  will not contain an entry for each GFN at page
>> size
>> level". If aligned, we expect lpage_info[0] to have 512 elements.
>>
>> 1GB: lpage_info[1] has 1 element
>> 2MB: lpage_info[0] has 2 elemtnts
> 
> 1048576 GFN is 2MB aligned, 1024 pages is also 2MB aligned. There are
> 512 4k pages in a 2MB huge page, so size of 2 for npages=1024 looks
> right to me. One struct for each potential 2MB huge page in the range.
> 
> I think overall you are saying in this response that you didn't find
> any problem in the analysis or fix. Is that correct?

I do not find any problem in the analysis or fix, except the same curiosity on
KVM_LPAGE_MIXED_FLAG.

Thank you very much!

Dongli Zhang