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;
From:   "Fabio M. De Francesco" <fmdefrancesco@gmail.com>
To:     Todd Kjos <tkjos@android.com>, linux-kernel@vger.kernel.org,
        Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Cc:     Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Arve =?ISO-8859-1?Q?Hj=F8nnev=E5g?= <arve@android.com>,
        Martijn Coenen <maco@android.com>,
        Joel Fernandes <joel@joelfernandes.org>,
        Christian Brauner <brauner@kernel.org>,
        Hridya Valsaraju <hridya@google.com>,
        Suren Baghdasaryan <surenb@google.com>,
        Ira Weiny <ira.weiny@intel.com>
Subject: Re: [PATCH 3/3] binder: Use kmap_local_page() in binder_alloc_get_page()
Date:   Sun, 24 Apr 2022 11:39:57 +0200
Message-ID: <2583087.X9hSmTKtgW@leap>
In-Reply-To: <fad918d3-6923-5bec-7830-5cddf7a725d6@wanadoo.fr>
References: <20220423102421.16869-1-fmdefrancesco@gmail.com> <20220423102421.16869-4-fmdefrancesco@gmail.com> <fad918d3-6923-5bec-7830-5cddf7a725d6@wanadoo.fr>
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="utf-8"
Precedence: bulk

On sabato 23 aprile 2022 18:02:48 CEST Christophe JAILLET wrote:
> Hi,
>=20
> Le 23/04/2022 =C3=A0 12:24, Fabio M. De Francesco a =C3=A9crit :
> > The use of kmap_atomic() is being deprecated in favor of=20
kmap_local_page()
> > where it is feasible. Each call of kmap_atomic() in the kernel creates
> > a non-preemptible section and disable pagefaults. This could be a=20
source
> > of unwanted latency, so it should be only used if it is absolutely
> > required, otherwise kmap_local_page() should be preferred.
> >=20
> > With kmap_local_page(), the mapping is per thread, CPU local and not
> > globally visible. Furthermore, the mapping can be acquired from any=20
context
> > (including interrupts).
> >=20
> > Therefore, use kmap_local_page() / kunmap_local() in place of
> > kmap_atomic() / kunmap_atomic().
> >=20
> > Signed-off-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>
> > ---
> >   drivers/android/binder_alloc.c | 6 +++---
> >   1 file changed, 3 insertions(+), 3 deletions(-)
> >=20
> > diff --git a/drivers/android/binder_alloc.c b/drivers/android/
binder_alloc.c
> > index 0875c463c002..058595cc7ff0 100644
> > --- a/drivers/android/binder_alloc.c
> > +++ b/drivers/android/binder_alloc.c
> > @@ -1250,17 +1250,17 @@ static int binder_alloc_do_buffer_copy(struct=20
binder_alloc *alloc,
> >   		page =3D binder_alloc_get_page(alloc, buffer,
> >   					     buffer_offset,=20
&pgoff);
> >   		size =3D min_t(size_t, bytes, PAGE_SIZE - pgoff);
> > -		base_ptr =3D kmap_atomic(page);
> > +		base_ptr =3D kmap_local_page(page);
> >   		tmpptr =3D base_ptr + pgoff;
> >   		if (to_buffer)
> >   			memcpy(tmpptr, ptr, size);
> >   		else
> >   			memcpy(ptr, tmpptr, size);
>=20
> in the same spirit as patch 1/3, memcpy_to_page() and memcpy_from_page()=
=20
> looks good candidate to avoid code duplication.

Hello Christophe, Todd,

I had thought to use memcpy_to_page() and memcpy_from_page() (exactly as I=
=20
did in other conversions I have been working on during the latest couple of=
=20
weeks).=20

However, I decided to avoid to use them for I should also have deleted the=
=20
comment which is before "kunmap_local(base_ptr);".

I don't know how much Maintainers think it is necessary to make readers=20
notice that "kunmap_local() takes care of flushing the cache []" (exactly=20
as kunmap_atomic() does). Actually I'd delete that comment that looks=20
redundant and unnecessary to me, but I cannot know if Todd wants it to=20
remain there.

@Todd: Can you please say what you think about this topic? Should I leave=20
the patch as-is or I should use memcpy_{to,from}_page() and delete that=20
comment?

I won't send any v2 unless I have your confirmation.

Thanks,

=46abio

>=20
> Not checked in details, but looks mostly the same.
>=20
> Just my 2c.
>=20
> CJ
>=20
> >   		/*
> > -		 * kunmap_atomic() takes care of flushing the cache
> > +		 * kunmap_local() takes care of flushing the cache
> >   		 * if this device has VIVT cache arch
> >   		 */
> > -		kunmap_atomic(base_ptr);
> > +		kunmap_local(base_ptr);
> >   		bytes -=3D size;
> >   		pgoff =3D 0;
> >   		ptr =3D ptr + size;
>=20
>=20