Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
From:   Alexander Potapenko <glider@google.com>
Date:   Wed, 11 Mar 2020 18:28:42 +0100
Message-ID: <CAG_fn=WOJb7bsP2JfhH+PRmhZ=4WY8U+vUugkTpXL4F5sz752g@mail.gmail.com>
Subject: Reason to not use __GFP_ZERO in __alloc_zeroed_user_highpage() on ARM64?
To:     Andrew Morton <akpm@linux-foundation.org>,
        Mark Rutland <mark.rutland@arm.com>,
        Will Deacon <will.deacon@arm.com>,
        Catalin Marinas <catalin.marinas@arm.com>
Cc:     LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi,

I've noticed that certain arches (alpha, ia64, m68k, s390, x86) have
__alloc_zeroed_user_highpage() defined as:

  alloc_page_vma(GFP_HIGHUSER | __GFP_ZERO | movableflags, vma, vmaddr)

, whereas in other cases it is defined as
alloc_page_vma()+clear_user_page() (see
https://elixir.bootlin.com/linux/latest/source/include/linux/highmem.h#L182=
)

Is there a reason for this?

I'm asking because on ARM64 with init_on_alloc=3D1
__alloc_zeroed_user_highpage() appears to initialize the page twice.
Adding __GFP_ZERO and removing clear_user_page() seems to work (and
remove the double initialization), but I suppose this code was written
on purpose? Am I missing something?

Thanks,

--=20
Alexander Potapenko
Software Engineer

Google Germany GmbH
Erika-Mann-Stra=C3=9Fe, 33
80636 M=C3=BCnchen

Gesch=C3=A4ftsf=C3=BChrer: Paul Manicle, Halimah DeLaine Prado
Registergericht und -nummer: Hamburg, HRB 86891
Sitz der Gesellschaft: Hamburg