Re: [RFC PATCH 0/4] Restore change_pte optimization to its former glory

[Andrea Arcangeli]

On Fri, Feb 01, 2019 at 06:57:38PM -0500, Andrea Arcangeli wrote:
> If it's cleared with ptep_clear_flush_notify, change_pte still won't
> work. The above text needs updating with
> "ptep_clear_flush". set_pte_at_notify is all about having
> ptep_clear_flush only before it or it's the same as having a range
> invalidate preceding it.
> 
> With regard to the code, wp_page_copy() needs
> s/ptep_clear_flush_notify/ptep_clear_flush/ before set_pte_at_notify.

Oops, the above two statements were incorrect because
ptep_clear_flush_notify doesn't interfere with change_pte and will
only invalidate secondary MMU mappings on those secondary MMUs that
shares the same pagetables with the primary MMU and that in turn won't
ever implement a change_pte method.