Re: [PATCH v2 02/46] mm: Add folio_rmapping()

[Matthew Wilcox]

On Wed, Jun 23, 2021 at 09:56:58AM +0200, Christoph Hellwig wrote:
> > +static inline void *folio_rmapping(struct folio *folio)
> 
> This name, just like the old one is not exaclty descriptive.  I guess the
> r stands for raw somehow?  As a casual contributor to the fringes of the
> MM I would have no idea when to use it.
> 
> All this of course also applies to the existing (__)page_rmapping, but
> maybe this is a good time to sort it out.

Yes, good point.  I don't like the name rmapping either, since
we already have rmap which has nothing to do with this.  I'll
leave page_rmapping() alone for now; no need to add that churn.
I think they all become calls to folio_raw_mapping() later.

> >  
> >  struct anon_vma *page_anon_vma(struct page *page)
> >  {
> > +	struct folio *folio = page_folio(page);
> > +	unsigned long mapping = (unsigned long)folio->mapping;
> >  
> >  	if ((mapping & PAGE_MAPPING_FLAGS) != PAGE_MAPPING_ANON)
> >  		return NULL;
> > +	return folio_rmapping(folio);
> 
> It feelds kinda silly to not just open code folio_rmapping here
> given that we alredy went half the way. 

Yeah, I thought about that too.  Done:

-       return folio_rmapping(folio);
+       return (void *)(mapping - PAGE_MAPPING_ANON);