MIME-Version: 1.0
References: <20230301152239.531194-1-miquel.raynal@bootlin.com> <20230301152239.531194-4-miquel.raynal@bootlin.com>
In-Reply-To: <20230301152239.531194-4-miquel.raynal@bootlin.com>
From:   Rob Herring <robh+dt@kernel.org>
Date:   Thu, 2 Mar 2023 13:13:14 -0600
Message-ID: <CAL_JsqKsYeJz4M4v_NehyDqLHOHmArNgad4HDhYGN7RucHsv+A@mail.gmail.com>
Subject: Re: [PATCH 3/8] of: Create an of_device_request_module() receiving an
 OF node
To:     Miquel Raynal <miquel.raynal@bootlin.com>
Cc:     Srinivas Kandagatla <srinivas.kandagatla@linaro.org>,
        linux-kernel@vger.kernel.org,
        Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
        Michael Walle <michael@walle.cc>, devicetree@vger.kernel.org,
        Frank Rowand <frowand.list@gmail.com>,
        Robert Marko <robert.marko@sartura.hr>,
        Luka Perkov <luka.perkov@sartura.hr>,
        Thomas Petazzoni <thomas.petazzoni@bootlin.com>,
        rafal@milecki.pl
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Wed, Mar 1, 2023 at 9:22=E2=80=AFAM Miquel Raynal <miquel.raynal@bootlin=
.com> wrote:
>
> of_device_request_module() currently receives a "struct device" as main
> argument, but the only use of this pointer is to access its .of_node
> member. In practice, this function only needs a "struct
> device_node". Let's move the logic into another helper which would
> receive a "struct device_node" instead, and use that new helper from the
> ancient of_device_request_module(). Exporting this new function will be
> useful to request module loading when the "struct device" is not
> available.

of_device.h is for things that operate on struct device, so
of_device_request_module() doesn't really belong here.

I wouldn't really care, but we have this ~12 year old line in of_device.h:

#include <linux/of_platform.h> /* temporary until merge */

I started working on a very churny series to fix this only to wonder
if a header split by consumer would be better. Anyways, concrete
suggestions below...

> Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
> ---
>  drivers/of/device.c       | 17 +++++++++++++----
>  include/linux/of_device.h |  6 ++++++
>  2 files changed, 19 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/of/device.c b/drivers/of/device.c
> index 3efc17de1d57..7cdf252b9526 100644
> --- a/drivers/of/device.c
> +++ b/drivers/of/device.c
> @@ -284,16 +284,16 @@ static ssize_t of_device_get_modalias(struct device=
_node *np, char *str, ssize_t
>         return tsize;
>  }
>
> -int of_device_request_module(struct device *dev)
> +int of_device_node_request_module(struct device_node *np)'

We use 'device_node' pretty much nowhere in the DT APIs. Just
of_request_module() and define in of.h. There is only one user of
of_device_request_module, so remove it and update the user to use
of_request_module() (and hopefully drop of_device.h for it).

Rob