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:   Javier Martinez Canillas <javierm@redhat.com>
To:     Maxime Ripard <maxime@cerno.tech>
Cc:     Emma Anholt <emma@anholt.net>, David Airlie <airlied@gmail.com>,
        Daniel Vetter <daniel@ffwll.ch>,
        Hans Verkuil <hverkuil@xs4all.nl>,
        linux-kernel@vger.kernel.org, dri-devel@lists.freedesktop.org,
        Dave Stevenson <dave.stevenson@raspberrypi.com>
Subject: Re: [PATCH v3 1/9] drm/vc4: Switch to container_of_const
In-Reply-To: <6ckl6zmizxsm52imeau55bicb4hmd3p46btsqih64qaoeptvww@bktsqhuq4lek>
References: <20221207-rpi-hdmi-improvements-v3-0-bdd54f66884e@cerno.tech>
 <20221207-rpi-hdmi-improvements-v3-1-bdd54f66884e@cerno.tech>
 <87354sij2i.fsf@minerva.mail-host-address-is-not-set>
 <6ckl6zmizxsm52imeau55bicb4hmd3p46btsqih64qaoeptvww@bktsqhuq4lek>
Date:   Tue, 25 Apr 2023 10:21:38 +0200
Message-ID: <87pm7sgynh.fsf@minerva.mail-host-address-is-not-set>
MIME-Version: 1.0
Content-Type: text/plain
Precedence: bulk

Maxime Ripard <maxime@cerno.tech> writes:

> Hi Javier,
>
> On Sat, Apr 22, 2023 at 07:26:13AM +0200, Javier Martinez Canillas wrote:
>> Maxime Ripard <maxime@cerno.tech> writes:
>> > container_of_const() allows to preserve the pointer constness and is
>> > thus more flexible than inline functions.
>> >
>> > Let's switch all our instances of container_of() to
>> > container_of_const().
>> >
>> > Signed-off-by: Maxime Ripard <maxime@cerno.tech>
>> > ---
>> 
>> [...]
>> 
>> > -static inline struct vc4_dpi *
>> > -to_vc4_dpi(struct drm_encoder *encoder)
>> > -{
>> > -	return container_of(encoder, struct vc4_dpi, encoder.base);
>> > -}
>> > +#define to_vc4_dpi(_encoder)						\
>> > +	container_of_const(_encoder, struct vc4_dpi, encoder.base)
>> >
>> 
>> A disadvantage of this approach though is that the type checking is lost.
>
> Not entirely, the argument is still type-checked, but yeah, it's true
> for the returned value.
>
>> Since you already had these, I would probably had changed them to return
>> a const pointer and just replace container_of() for container_of_const().
>> 
>> But I see that there are a lot of patches from Greg all over the kernel
>> that do exactly this, dropping static inline functions in favor of using
>> container_of_const() directly. So it seems the convention is what you do.
>
> More importantly, container_of_const() isn't always returning a const
> pointer or always taking a const argument, it's returning the pointer
> with the same const-ness than the argument.
>
> This is why it makes sense to remove the inline function entirely,
> because it removes the main benefit it brings.
>

Got it. Thanks for the explanations.

-- 
Best regards,

Javier Martinez Canillas
Core Platforms
Red Hat