Re: [PATCH 4/4 v3] KVM: Introduce kvm_memory_slot::arch and move lpage_info into it

[Alexander Graf <agraf@suse.de>]

On 03/07/2012 04:03 PM, Avi Kivity wrote:
> On 03/07/2012 04:57 PM, Alexander Graf wrote:
>> On 03/07/2012 03:04 PM, Avi Kivity wrote:
>>> On 03/07/2012 03:27 PM, Alexander Graf wrote:
>>>>> At the time I made this patch, it seemed that only kvm-x86 supported
>>>>> huge pages, on ppc the array should be empty:
>>>> Hrm. I suppose this refers to transparent huge pages?
>>> Just huge pages.  Whether they are static or dynamic is immaterial in
>>> this context.
>> Well, book3s_hv and e500 support hugetlbfs. I've never had to touch
>> that patches code though - so I guess I'm still not really
>> understanding what it's there for O_o.
>>
> The kvm hugepage code uses large sptes to map large pages, when
> available (either via hugetlbfs or transparent hugepages).  Since x86
> supports swapping, and needs to write-protect pages for dirty logging
> and for shadowing guest pagetables, it needs a reverse map from pages to
> sptes.  The data structure we're discussing is part of the reverse map
> for large pages.

Ah, now that makes more sense. On booke, we don't do rmap yet. On 
book3s_hv, IIRC Paul did implement something, so I'd like to hear his 
opinion on it really.

Alex