Re: [PATCH v6 6/9] KVM: MMU: fast path of handling guest page fault

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

On 06/14/2012 06:40 AM, Marcelo Tosatti wrote:

> On Tue, May 29, 2012 at 02:50:32PM +0800, Xiao Guangrong wrote:
>> If the the present bit of page fault error code is set, it indicates
>> the shadow page is populated on all levels, it means what we do is
>> only modify the access bit which can be done out of mmu-lock
>>
>> Currently, in order to simplify the code, we only fix the page fault
>> caused by write-protect on the fast path
>>
>> Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
>> ---
>>  arch/x86/kvm/mmu.c |  126 +++++++++++++++++++++++++++++++++++++++++++++++-----
>>  1 files changed, 114 insertions(+), 12 deletions(-)
>>
>> diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
>> index 150c5ad..d6101a8 100644
>> --- a/arch/x86/kvm/mmu.c
>> +++ b/arch/x86/kvm/mmu.c
>> @@ -445,6 +445,11 @@ static bool __check_direct_spte_mmio_pf(u64 spte)
>>  }
>>  #endif
>>
>> +static bool spte_can_be_writable(u64 spte)
>> +{
>> +	return !(~spte & (SPTE_HOST_WRITEABLE | SPTE_MMU_WRITEABLE));
>> +}
>> +
> 
> spte_is_locklessly_modifiable(). Its easy to confuse
> "spte_can_be_writable" with different things.
> 


Yes. Will update it.

>>  static bool spte_has_volatile_bits(u64 spte)
>>  {
>>  	if (!shadow_accessed_mask)
>> @@ -454,7 +459,7 @@ static bool spte_has_volatile_bits(u64 spte)
>>  		return false;
>>
>>  	if ((spte & shadow_accessed_mask) &&
>> -	      (!is_writable_pte(spte) || (spte & shadow_dirty_mask)))
>> +	      (!spte_can_be_writable(spte) || (spte & shadow_dirty_mask)))
>>  		return false;
> 
> mmu_spte_update is handling several different cases. Please rewrite
> it, add a comment on top of it (or spread comments on top of each
> significant code line) with all cases it is handling (also recheck it
> regarding new EPT accessed/dirty bits code).
> 


Okay.

> For one thing, if spte can be updated locklessly the update must be
> atomic:
> 
> if spte can be locklessly updated
> 	read-and-modify must be atomic.


Actually, i did it in the v5, Avi has some comments on that. Please
see https://lkml.org/lkml/2012/5/24/55

What the reason we should locklessly update spte here? So far i know
is for volatile bit lost and getting a stable is_writable_spte()?

But this two cases can be avoided by using spte_can_be_writable(spte)
instead of is_writable_pte(spte), right?