Re: [PATCH] mm/khugepaged: Detecting uffd-wp vma more efficiently

[Peter Xu]

On Thu, Sep 23, 2021 at 09:47:42AM -0700, Axel Rasmussen wrote:
> My thinking there was a THP collapse bug was really just based on
> speculation, not a real reproducer, so it's very possible my
> speculation was wrong. It will take some more thinking and reading to
> convince myself one way or the other. :) Thanks to you and Hugh for
> all the details.
> 
> I'd prefer not to add this fix "just in case", if it isn't a real
> problem, as it seems like it may confuse future readers of the code.

It's not "just in case" to me - IMHO it's theoretically causing more false
positives as I used to mention, at least that's my understanding so far. So if
the theory is correct it'll 100% happen when khugepaged merged some
minor-registered regions.

Uffd-wp could have many false positives like this if we don't support swap - at
last we decided to fully support swap then we removed all the false positives
regarding swapping.  I think it's similar here, but khugepaged should trigger
much less frequently on the false positives upon uffd-minor, than swapping upon
uffd-wp.

But yes, there's definitely no rush on thinking or anything - it'll never hurt
to think more. And more importantly, verify it with some test program would be
great; after all theoretically it'll just work like a charm to me.

> 
> I'll send out a patch for it if / when I manage to build a real
> reproducer. Or, in the meantime, some of my Google colleagues are
> testing this code via their live migration implementation, so if there
> is a bug here there's a good chance we'll find it that way too.

Sounds like a good plan.

Thanks,

-- 
Peter Xu