Re: [PATCH v2 1/1] fs/splice: add missing callback for inaccessible pages

[Dave Hansen]

One other thing...  The gup code will not take references on ref-frozen
pages:

> static inline __must_check bool try_get_page(struct page *page)
> {
>         page = compound_head(page);
>         if (WARN_ON_ONCE(page_ref_count(page) <= 0))
>                 return false;
>         page_ref_inc(page);
>         return true;
> }

*But*, notice that the path that skips taking a ref is also a
WARN_ON_ONCE().  Basically, if you get to try_get_page() on a ref-frozen
page, it's considered buggy.  This makes sense because you fundamentally
can't freeze refs on a page that might have more refs taken on it.

I think all the other users do this by ensuring that any PTE that could
be gup'd is set non-present before the refs are frozen and remote TLBs
are flushed which also ensures no GUPs are running.  I don't know if the
s390 code has some other way of preventing GUPs, but leaving Present=1
PTEs while you freeze refs would be quite troublesome on x86.