MIME-Version: 1.0
References: <20230306230004.1387007-1-jthoughton@google.com>
 <20230306230004.1387007-2-jthoughton@google.com> <20230307215420.GA59222@monkey>
 <CADrL8HVa3vzmrfFJD5hx_GuXVnsWhSo9hzJFb4TTzzjMhWG+sQ@mail.gmail.com> <ZAkEecMShUAGwZ62@x1n>
In-Reply-To: <ZAkEecMShUAGwZ62@x1n>
From:   James Houghton <jthoughton@google.com>
Date:   Thu, 9 Mar 2023 11:58:37 -0800
Message-ID: <CADrL8HWT8-MZeL-G+Qj4vQHVrrMxd8EWkPWO9eDfwvvZvoh69A@mail.gmail.com>
Subject: Re: [PATCH 1/2] mm: rmap: make hugetlb pages participate in _nr_pages_mapped
To:     Peter Xu <peterx@redhat.com>
Cc:     Mike Kravetz <mike.kravetz@oracle.com>,
        Hugh Dickins <hughd@google.com>,
        Muchun Song <songmuchun@bytedance.com>,
        "Matthew Wilcox (Oracle)" <willy@infradead.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>,
        David Hildenbrand <david@redhat.com>,
        David Rientjes <rientjes@google.com>,
        Axel Rasmussen <axelrasmussen@google.com>,
        Jiaqi Yan <jiaqiyan@google.com>, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Wed, Mar 8, 2023 at 1:56=E2=80=AFPM Peter Xu <peterx@redhat.com> wrote:
>
> On Tue, Mar 07, 2023 at 04:36:51PM -0800, James Houghton wrote:
> > > >       if (likely(!compound)) {
> > > > +             if (unlikely(folio_test_hugetlb(folio)))
> > > > +                     VM_BUG_ON_PAGE(HPageVmemmapOptimized(&folio->=
page),
> > > > +                                    page);
>
> How about moving folio_test_hugetlb() into the BUG_ON()?
>
>                 VM_BUG_ON_PAGE(folio_test_hugetlb(folio) &&
>                                HPageVmemmapOptimized(&folio->page),
>                                page);
>
> Note that BUG_ON() already contains an "unlikely".

Ok I can do that. It's a little cleaner.

> > > >               first =3D atomic_inc_and_test(&page->_mapcount);
> > > >               nr =3D first;
> > > >               if (first && folio_test_large(folio)) {
> > > >                       nr =3D atomic_inc_return_relaxed(mapped);
> > > >                       nr =3D (nr < COMPOUND_MAPPED);
> > > >               }
> > > > -     } else if (folio_test_pmd_mappable(folio)) {
> > > > -             /* That test is redundant: it's for safety or to opti=
mize out */
> > >
> > > I 'think' removing this check is OK.  It would seem that the caller
> > > knows if the folio is mappable.  If we want a similar test, we might =
be
> > > able to use something like:
> > >
> > >         arch_hugetlb_valid_size(folio_size(folio))
> > >
> >
> > Ack. I think leaving the check(s) removed is fine.
>
> Would it still be good to keep that as another BUG_ON()?

Sure, that sounds reasonable to me. I'll add it unless someone disagrees.

As you suggested in your other email, I'll also add a BUG_ON() if we
attempt to do a non-compound mapping of a folio that is larger than
COMPOUND_MAPPED / 2. (Maybe a BUG_ON() in alloc_hugetlb_folio() to
check that the size of the folio we're allocating is less than
COMPOUND_MAPPED / 2 makes sense instead. Just an idea.)

- James