Re: [PATCH v2] migration: Count new_dirty instead of real_dirty

[zhukeqian <zhukeqian1@huawei.com>]

Hi Dave,

On 2020/6/16 17:58, Dr. David Alan Gilbert wrote:
> * zhukeqian (zhukeqian1@huawei.com) wrote:
>> Hi Dave,
>>
>> On 2020/6/16 17:35, Dr. David Alan Gilbert wrote:
>>> * Keqian Zhu (zhukeqian1@huawei.com) wrote:
>>>> real_dirty_pages becomes equal to total ram size after dirty log sync
>>>> in ram_init_bitmaps, the reason is that the bitmap of ramblock is
>>>> initialized to be all set, so old path counts them as "real dirty" at
>>>> beginning.
>>>>
>>>> This causes wrong dirty rate and false positive throttling at the end
>>>> of first ram save iteration.
>>>>
>>>> Signed-off-by: Keqian Zhu <zhukeqian1@huawei.com>
>>>
>>> Since this function already returns num_dirty, why not just change the
>>> caller to increment a counter based off the return value?
>> Yes, that would be better :-) .
>>
>>>
>>> Can you point to the code which is using this value that triggers the
>>> throttle?
>>>
>> In migration_trigger_throttle(), rs->num_dirty_pages_period is used.
>> And it corresponds to real_dirty_pages here.
> 
> OK; so is the problem not the same as the check that's in there for
> blk_mig_bulk_activate - don't we need to do the same trick for ram bulk
> migration (i.e. the first pass).
> 
Sorry that I do not get your idea clearly. Could you give some sample
code?

> Dave
> 
>> Thanks,
>> Keqian
>>
>>> Dave
>>>
>>>
>> [...]
>>>>
>>>>
>>> --
>>> Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK
>>>
>>> .
>>>
>>
> --
> Dr. David Alan Gilbert / dgilbert@redhat.com / Manchester, UK
> 
> .
>