> On Jul 26, 2019, at 1:44 AM, Oleg Nesterov <o...@redhat.com> wrote:
>
> On 07/25, Song Liu wrote:
>>
>> I guess I know the case now. We can probably avoid this with an simple
>> check for old_page == new_page?
>
> better yet, I think we can check PageAnon(old_page) and avoid the unnecessary
> __replace_page() in this case. See the patch below.
I added PageAnon(old_page) check in v9 of the patch.
>
> Anyway, why __replace_page() needs to lock both pages? This doesn't look nice
> even if it were correct. I think it can do lock_page(old_page) later.
>
Agreed. I have changed the v9 to only unmap old_page. So it should be cleaner.
Thanks again for these good suggestions,
Song
