Re: [PATCH v3] mm/gup: Allow real explicit breaking of COW

[Linus Torvalds]

On Fri, Aug 21, 2020 at 12:31 PM Peter Xu <peterx@xxxxxxxxxx> wrote:
>
> However, do we really want to revert the whole logic of
> 52d1e606ee73?  Asking because it still seems to be an improvement to me (after
> all we'll need to look after KSM pages here).  So it seems still good to keep.

Does anybody have numbers for it?

I'd rather simplify and get rid of the locking that has been
problematic, and then re-introduce limited cases with actual numbers.

Right now that commit has no real argument for it except for "do what
we do for swap cache". And since we're getting rid of the swap cache
special case, I'd say that commit 52d1e606ee73 argues for getting rid
of the KSM special case too.

Honestly, I'd expect that if KSM is effective, it's for pages that
really *are* shared. If you get a lot of "write fault on the last
copy" and that ends up being a problem, I think that says more about
the KSM issue than it says about the write fault..

               Linus