Received-SPF: pass (google.com: domain of linux-kernel+bounces-199652-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.80.249 as permitted sender) client-ip=147.75.80.249;
Precedence: bulk
MIME-Version: 1.0
References: <20240520194749.334896-1-vishal.moola@gmail.com>
 <ZkvErLsj8Ngj4qM3@casper.infradead.org> <CAOzc2px9sWGUPbCM+pHuqt8eAUONrDcKnEybvM6k6ot2ftpmmg@mail.gmail.com>
In-Reply-To: <CAOzc2px9sWGUPbCM+pHuqt8eAUONrDcKnEybvM6k6ot2ftpmmg@mail.gmail.com>
From: Vishal Moola <vishal.moola@gmail.com>
Date: Mon, 3 Jun 2024 12:38:47 -0700
Message-ID: <CAOzc2pzvWaqkHLCMfQ=oMggVSDU7+0do1+u0BndbS2ijVeEZ5A@mail.gmail.com>
Subject: Re: [PATCH] Convert __unmap_hugepage_range() to folios
To: Matthew Wilcox <willy@infradead.org>
Cc: linux-mm@kvack.org, akpm@linux-foundation.org, muchun.song@linux.dev, 
	linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Tue, May 21, 2024 at 2:13=E2=80=AFPM Vishal Moola <vishal.moola@gmail.co=
m> wrote:
>
> On Mon, May 20, 2024 at 2:46=E2=80=AFPM Matthew Wilcox <willy@infradead.o=
rg> wrote:
> >
> > On Mon, May 20, 2024 at 12:47:49PM -0700, Vishal Moola (Oracle) wrote:
> > > Replaces 4 calls to compound_head() with one. Also converts
> > > unmap_hugepage_range() and unmap_ref_private() to take in folios.
> >
> > This is great!
> >
> > >  void __unmap_hugepage_range(struct mmu_gather *tlb, struct vm_area_s=
truct *vma,
> > >                           unsigned long start, unsigned long end,
> > > -                         struct page *ref_page, zap_flags_t zap_flag=
s)
> > > +                         struct folio *ref_folio, zap_flags_t zap_fl=
ags)
> > >  {
> > >       struct mm_struct *mm =3D vma->vm_mm;
> > >       unsigned long address;
> > >       pte_t *ptep;
> > >       pte_t pte;
> > >       spinlock_t *ptl;
> > > -     struct page *page;
> > > +     struct folio *folio;
> > >       struct hstate *h =3D hstate_vma(vma);
> > >       unsigned long sz =3D huge_page_size(h);
> >
> > I would appreciate some further cleanup ...
> >
> >         size_t sz =3D folio_size(folio);
> >
> > I think there are further cleanups along those lines, eg
> > pages_per_huge_page(), hugetlb_mask_last_page(), huge_page_mask().
> >
>
> Gotcha, I'll look into those and change them in v2.

I took a closer look at your suggestions for cleanups here. Most callers of
unmap_hugepage_range() pass NULL as ref_folio - meaning we want to unmap
all pages in the range. This means alot of the preparatory work is
likely done without
a reference to a folio, so using folio_size() is unsafe.

For when we later have a reference to a folio, I think we should
continue to use the
hstate-defined values since using the folio in one place and hstate in
another makes
things harder to change (if we ever want to).