Re: [PATCH v6 5/9] KVM: MMU: introduce SPTE_MMU_WRITEABLE bit

[Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>]

On 06/13/2012 10:01 AM, Marcelo Tosatti wrote:

> On Tue, Jun 12, 2012 at 10:23:47AM +0800, Xiao Guangrong wrote:
>> On 06/12/2012 07:32 AM, Marcelo Tosatti wrote:
>>
>>> On Tue, May 29, 2012 at 02:49:14PM +0800, Xiao Guangrong wrote:
>>>> This bit indicates whether the spte can be writable on MMU, that means
>>>> the corresponding gpte is writable and the corresponding gfn is not
>>>> protected by shadow page protection
>>>
>>> Why is this still necessary, now that only sptes of direct shadow pages 
>>> are updated locklessly? 
>>>
>>
>>
>> Yes, but it is still needed, for nested npt/ept, we need protect
>> the nested page tables.
> 
> Sure, but shadowed L1 nested pagetables are not direct shadow pages.
> 
> They are shadows of L1 nested pagetables. 
> 
> Checking sp->direct should be enough (instead of the flags).
> 

Hi Marcelo,

I think it is not enough, for example:

- In host (L0), spte1 is pointing to gfn1, spte1 is a direct spte.

- in L1, L1 guest is using gfn1 in L1's ept page table for L2 guest,
  so, in host, we have a indirect spte (named spte2) whose sp->gfn = gfn1.

Since spte2 is a indirect spte, we need protect it, so, we walk all gfn1's
rmaps, spte1 will be found, then, we write-protect on spte1 to track L1
modifying gfn1.

In this case, spte1 is direct but need write-protect. :)