On 3/2/2023 10:55 PM, David Hildenbrand wrote:
> On 02.03.23 15:33, Matthew Wilcox wrote:
>> On Thu, Mar 02, 2023 at 03:23:46PM +0100, David Hildenbrand wrote:
>>> If no workload/benchmark is affected (or simply corner cases where nobody
>>> cares about performance), I hope you understand that it's hard to argue why
>>> we should care about such an optimization then.
>>
>> In order to solve the mapcount problem, we're going to want to unmap
>> the entire folio in one call, instead of unmapping each page in it
>> individually and checking each time whether there are any remaining
>> pages from this folio still mapped.
This is a good point.
>
> Okay, thanks. That should better be added to the cover letter, ideally with more details on the mapcount goal and how this patch set will helps in that regard. So far the cover letter only talks about eventual performance gains.
This patch reconstruct the try_to_unmap_one() from:
loop:
clear and update PTE
unmap one page
goto loop
to:
loop:
clear and update PTE
goto loop
unmap the range of folio in one call
Finally, if the entire folio is always mapped, it will be
unmapped entirely also. I will add this to cover letter.
Thanks.
Regards
Yin, Fengwei
>
