Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20230721094043.2506691-1-fengwei.yin@intel.com> <20230721094043.2506691-2-fengwei.yin@intel.com>
In-Reply-To: <20230721094043.2506691-2-fengwei.yin@intel.com>
From:   Yu Zhao <yuzhao@google.com>
Date:   Fri, 21 Jul 2023 12:57:37 -0600
Message-ID: <CAOUHufY0AX_wyBCZ3dNMfTg7C1ZBOZHpYSgG=uOwbeeemyFX6Q@mail.gmail.com>
Subject: Re: [RFC PATCH v2 1/4] madvise: not use mapcount() against large
 folio for sharing check
To:     Yin Fengwei <fengwei.yin@intel.com>
Cc:     linux-mm@kvack.org, linux-kernel@vger.kernel.org,
        akpm@linux-foundation.org, minchan@kernel.org, willy@infradead.org,
        david@redhat.com, ryan.roberts@arm.com, shy828301@gmail.com
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Fri, Jul 21, 2023 at 3:41=E2=80=AFAM Yin Fengwei <fengwei.yin@intel.com>=
 wrote:
>
> The commit
> 07e8c82b5eff ("madvise: convert madvise_cold_or_pageout_pte_range() to
> use folios") replaced the page_mapcount() with folio_mapcount() to
> check whether the folio is shared by other mapping.
>
> But it's not correct for large folio. folio_mapcount() returns the
> total mapcount of large folio which is not suitable to detect whether
> the folio is shared.
>
> Use folio_estimated_sharers() which returns a estimated number of
> shares. That means it's not 100% correct. But it should be OK for
> madvise case here.
>
> Signed-off-by: Yin Fengwei <fengwei.yin@intel.com>

Fixes:
Cc: stable

> @@ -383,7 +383,7 @@ static int madvise_cold_or_pageout_pte_range(pmd_t *p=
md,
>                 folio =3D pfn_folio(pmd_pfn(orig_pmd));
>
>                 /* Do not interfere with other mappings of this folio */
> -               if (folio_mapcount(folio) !=3D 1)
> +               if (folio_estimated_sharers(folio) !=3D 1)

Strictly speaking, this isn't a bug. But it may be ok to include it in
the same patch.

>                         goto huge_unlock;
>
>                 if (pageout_anon_only_filter && !folio_test_anon(folio))
> @@ -459,7 +459,7 @@ static int madvise_cold_or_pageout_pte_range(pmd_t *p=
md,
>                 if (folio_test_large(folio)) {
>                         int err;
>
> -                       if (folio_mapcount(folio) !=3D 1)
> +                       if (folio_estimated_sharers(folio) !=3D 1)
>                                 break;
>                         if (pageout_anon_only_filter && !folio_test_anon(=
folio))
>                                 break;
> @@ -682,7 +682,7 @@ static int madvise_free_pte_range(pmd_t *pmd, unsigne=
d long addr,

What about madvise_free_huge_pmd()? Should it be changed as well so
that it's consistent with the first change? Either change both or neither.

>                 if (folio_test_large(folio)) {
>                         int err;
>
> -                       if (folio_mapcount(folio) !=3D 1)
> +                       if (folio_estimated_sharers(folio) !=3D 1)

This is another bug fix and should be in a separate patch.

>                                 break;
>                         if (!folio_trylock(folio))
>                                 break;

Please send two separate fixes, and then:

Reviewed-by: Yu Zhao <yuzhao@google.com>