Re: [PATCH 0/8] KVM: Reduce mmu_lock hold time when zapping mmu pages

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

On 01/23/2013 09:28 PM, Takuya Yoshikawa wrote:
> On Wed, 23 Jan 2013 18:44:52 +0800
> Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com> wrote:
> 
>> On 01/23/2013 06:12 PM, Takuya Yoshikawa wrote:
>>> This patch set mitigates another mmu_lock hold time issue.  Although
>>> this is not enough and I'm thinking of additional work already, this
>>> alone can reduce the lock hold time to some extent.
>>>
>>
>> It is not worth doing this kind of complex thing, usually, only a few pages on
>> the invalid list. The *really* heavily case is kvm_mmu_zap_all() which can be speeded
>> up by using generation number, this is a todo work in kvm wiki:
> 
> I don't think this is so complex.  This is a basic programming technique
> using linked lists.

I just consider that only a few page exist on the invalid list, no worth introducing
this.

> 
> The current code which deletes the two link nodes in different functions
> looks unnatural to me: traversing the sp->link nodes forces us to break
> the loop and sp->hash_link nodes alone is allowed to continue ...
> 
> Making each function semantically clear should be more important than
> other things.
> 

The reason the code like this is, we have lockless shadow page walker.

> But maybe a matter of taste, so I'll wait for the maintainers' comments.
> 
>> http://www.linux-kvm.org/page/TODO: O(1) mmu invalidation using a generation number
>>
>> I am doing this work for some weeks and will post the patch out during these days.
> 
> I remember that Avi originally wrote the idea of introducing the
> generation of mmu pages in his other work.
> 

Whatever the original consideration is, the idea can speed up mmu invalidation a lot.
(Actually, i mentioned this idea to you when discussion fast write protect long time ago.)