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: <20220613120534.36991-1-andrea.merello@iit.it> <20220613120534.36991-9-andrea.merello@iit.it>
 <CAHp75Vdh8kAH6540xCwzFh5uf=QMVTHC42a8pOgvkpObzjfD+w@mail.gmail.com>
 <164f663acbba481a8ea8f45f185aaf83@iit.it> <CAHp75VeYK=oCbbBVp01_b5LK_FqAo4F_bic9Me4Y6PpfFnDU6g@mail.gmail.com>
 <a41936e4063f4c2c9da7c7e1d915bd62@iit.it>
In-Reply-To: <a41936e4063f4c2c9da7c7e1d915bd62@iit.it>
From:   Andy Shevchenko <andy.shevchenko@gmail.com>
Date:   Tue, 14 Jun 2022 17:10:43 +0200
Message-ID: <CAHp75Vd_w1WOp9ntbNqxtuuVXi0kMGbX=OZ7cioNxoh2yUa1ag@mail.gmail.com>
Subject: Re: [v6 08/14] iio: imu: add Bosch Sensortec BNO055 core driver
To:     Andrea Merello <Andrea.Merello@iit.it>
Cc:     Jonathan Cameron <jic23@kernel.org>,
        Mauro Carvalho Chehab <mchehab+huawei@kernel.org>,
        linux-iio <linux-iio@vger.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        devicetree <devicetree@vger.kernel.org>,
        Lars-Peter Clausen <lars@metafoo.de>,
        Rob Herring <robh+dt@kernel.org>,
        Matt Ranostay <matt.ranostay@konsulko.com>,
        Alexandru Ardelean <ardeleanalex@gmail.com>,
        jmondi <jacopo@jmondi.org>,
        "andrea.merello@gmail.com" <andrea.merello@gmail.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Tue, Jun 14, 2022 at 2:15 PM Andrea Merello <Andrea.Merello@iit.it> wrot=
e:

...

> >> >> +       devm_add_action_or_reset(priv->dev, bno055_debugfs_remove, =
priv->debugfs);
> >> >
> >> >Shouldn't we report the potential error here? It's not directly
> >> >related to debugfs, but something which is not directly related.
> >>
> >> The error eventually comes out from something that has nothing to do w=
ith debugs per se (i.e. the devm stuff), but it will only affect debugfs in=
deed.
> >>
> >> Assuming that we don't want to make the whole driver fail in case debu=
gfs stuff fails (see last part of the comment above debugfs_create_file() i=
mplementation), and given that the devm_add_action_or_reset(), should indee=
d "reset" in case of failure (i.e.  we should be in a clean situation anywa=
y), I would say it should be OK not to propagate the error and let things g=
o on.
> >
> >As I said, it's not directly related to debugfs. Here is the resource
> >leak possible or bad things happen if you probe the driver, that fails
> >to add this call for removal, remove it, and try to insert again, in
> >such case the debugfs will be stale.
>
> Hum, I would say this shouldn't ever happen: AFAICS devm_add_action_or_re=
set() is a wrapper around devm_add_action() and it's purpose is exactly to =
add a check for failure; devm_add_action_or_reset() immediately invokes the=
 action handler in case devm_add_action() fails. IOW in case of failure to =
add the devm stuff, the debugfs file is removed immediately and it shouldn'=
t cause any mess with next times probe()s; just the driver will go on witho=
ut the debugfs file being here.
>
> I think this is the point of using devm_add_action_or_reset() instead of =
dev_add_action()  indeed, or am I missing something?

Reading that code again and I think you are right, so dev_warn() will
be sufficient to show that we fail. OTOH, what is the point of adding
a resource for the failed debugfs call?

--=20
With Best Regards,
Andy Shevchenko