On Fri, Sep 03, 2021 at 10:58:53AM +1000, Alistair Popple wrote:
> On Friday, 3 September 2021 6:18:19 AM AEST Peter Xu wrote:
> > Use the helper for the checks. Rename "check_mapping" into "zap_mapping"
> > because "check_mapping" looks like a bool but in fact it stores the mapping
> > itself. When it's set, we check the mapping (it must be non-NULL). When it's
> > cleared we skip the check, which works like the old way.
> >
> > Move the duplicated comments to the helper too.
> >
> > Signed-off-by: Peter Xu <peterx@xxxxxxxxxx>
> > ---
> > include/linux/mm.h | 15 ++++++++++++++-
> > mm/memory.c | 29 ++++++-----------------------
> > 2 files changed, 20 insertions(+), 24 deletions(-)
> >
> > diff --git a/include/linux/mm.h b/include/linux/mm.h
> > index 69259229f090..81e402a5fbc9 100644
> > --- a/include/linux/mm.h
> > +++ b/include/linux/mm.h
> > @@ -1720,10 +1720,23 @@ extern void user_shm_unlock(size_t, struct ucounts *);
> > * Parameter block passed down to zap_pte_range in exceptional cases.
> > */
> > struct zap_details {
> > - struct address_space *check_mapping; /* Check page->mapping if set */
> > + struct address_space *zap_mapping; /* Check page->mapping if set */
> > struct page *single_page; /* Locked page to be unmapped */
> > };
> >
> > +/*
> > + * We set details->zap_mappings when we want to unmap shared but keep private
> > + * pages. Return true if skip zapping this page, false otherwise.
> > + */
> > +static inline bool
> > +zap_skip_check_mapping(struct zap_details *details, struct page *page)
> > +{
> > + if (!details || !page)
> > + return false;
> > +
> > + return details->zap_mapping != page_rmapping(page);
>
> Shouldn't this check still be
> details->zap_mapping && details->zap_mapping != page_rmapping(page)?
>
> Previously we wouldn't skip zapping pages if even_cows == true (ie.
> details->check_mapping == NULL). With this change the check when
> even_cows == true becomes NULL != page_rmapping(page). Doesn't this mean we
> will now skip zapping any pages with a mapping when even_cows == true?
Yes I think so. Thanks for pointing that out, Alistair, I'll fix in v3.
But frankly I really think we should simply have that flag I used to introduce.
It'll make everything much clearer.
--
Peter Xu
