Re: [RFC PATCH 05/12] fs/proc/task_mmu: refactor pagemap_pmd_range()

[Gregory Price]

On Mon, Oct 27, 2025 at 04:26:54PM +0000, Lorenzo Stoakes wrote:
> On Mon, Oct 27, 2025 at 01:11:46PM -0300, Jason Gunthorpe wrote:
> > On Fri, Oct 24, 2025 at 04:37:18PM -0400, Gregory Price wrote:
> > > On Fri, Oct 24, 2025 at 09:15:59PM +0100, Lorenzo Stoakes wrote:
> > > > On Fri, Oct 24, 2025 at 03:12:08PM -0400, Gregory Price wrote:
> > > >
> > > > So maybe actually that isn't too bad of an idea...
> > > >
> > > > Could also be
> > > >
> > > > nonpresent_or_swap_t but that's kinda icky...
> > >
> > > clearly we need:
> > >
> > > union {
> > > 	swp_entry_t swap;
> > > 	nonpresent_entry_t np;
> > > 	pony_entry_t pony;
> > > 	plum_emtry_t beer;
> > > } leaf_entry_t;
> 
> I think Greg meant this as a joke [correct me if wrong] :) that was my
> impression anyway (see original end of email...)
>
> I like leaf_entry_t name-wise.
> 
> I don't love the union.
>

The union was definitely a joke - see `plum_entry_t beer`

There definitely shouldn't be enough extensions to warrant a union here,
that seems like negative value.

leaf_entry_t naming replacing swp_entry_t seems reasonable since that's
basically all swp_entry_t is in its current form - even according to the
this set's cover letter:

```
  There's an established convention in the kernel that we treat leaf page
  tables (so far at the PTE, PMD level) as containing 'swap entries' should
  they be neither empty (i.e. p**_none() evaluating true) nor present
  (i.e. p**_present() evaluating true).
```

~Gregory