Re: [PATCH 09/10] log-dirty: Refine common code to support PML

[Kai Huang <kai.huang@linux.intel.com>]

On 04/10/2015 05:31 PM, Tim Deegan wrote:
> At 15:38 +0800 on 10 Apr (1428680289), Kai Huang wrote:
>>
>> On 04/09/2015 08:27 PM, Tim Deegan wrote:
>>> At 10:35 +0800 on 27 Mar (1427452553), Kai Huang wrote:
>>>> --- a/xen/arch/x86/mm/paging.c
>>>> +++ b/xen/arch/x86/mm/paging.c
>>>> @@ -411,7 +411,18 @@ static int paging_log_dirty_op(struct domain *d,
>>>>        int i4, i3, i2;
>>>>    
>>>>        if ( !resuming )
>>>> +    {
>>>>            domain_pause(d);
>>>> +
>>>> +        /*
>>>> +         * Only need to flush when not resuming, as domain was paused in
>>>> +         * resuming case therefore it's not possible to have any new dirty
>>>> +         * page.
>>>> +         */
>>>> +        if ( d->arch.paging.log_dirty.flush_cached_dirty )
>>>> +            d->arch.paging.log_dirty.flush_cached_dirty(d);
>>> I think there are too many layers of indirection here. :)  How about:
>>>    - don't add a flush_cached_dirty() function to the log_dirty ops.
>>>    - just call p2m_flush_hardware_cached_dirty(d) here.
>>>
>>> Would that work OK?
>> Thanks for pointing out.
>>
>> Is it nature to call p2m layer functions in paging.c? If there's no
>> restriction on it, calling p2m_flush_hardware_cached_dirty is more
>> clear, and it should work.
> Yes, calling public p2m functions directly is OK -- it's the internal
> function pointers like ->get_entry() that aren't supposed to be called
> from outside p2m code.  I guess that's not terribly clear - maybe it
> needs some better comments.
Pretty clear to me now. Thanks. Will do.:)

Thanks,
-Kai
>
> Tim.