Received-SPF: pass (google.com: domain of linux-kernel+bounces-180061-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.48.161 as permitted sender) client-ip=147.75.48.161;
Message-ID: <2c15c859-6b2b-4979-8317-698bf6cc430c@linux.dev>
Date: Wed, 15 May 2024 23:19:58 +0800
Precedence: bulk
MIME-Version: 1.0
Subject: Re: [PATCH 0/2] drm/bridge: Add 'struct device *' field to the
 drm_bridge structure
To: Maxime Ripard <mripard@kernel.org>
Cc: Neil Armstrong <neil.armstrong@linaro.org>,
 Dmitry Baryshkov <dmitry.baryshkov@linaro.org>,
 dri-devel@lists.freedesktop.org, linux-kernel@vger.kernel.org
References: <20240514154045.309925-1-sui.jingfeng@linux.dev>
 <20240514-scarlet-corgi-of-efficiency-faf2bb@penduick>
 <c44480ab-8d6b-4334-8eba-83db9b30ff1a@linux.dev>
 <20240515-fair-satisfied-myna-480dea@penduick>
 <d394ee32-4fa4-41a8-a5ca-c1c7f77f44d2@linux.dev>
 <20240515-copper-chimpanzee-of-fortitude-ff3dab@penduick>
Content-Language: en-US, en-AU
From: Sui Jingfeng <sui.jingfeng@linux.dev>
In-Reply-To: <20240515-copper-chimpanzee-of-fortitude-ff3dab@penduick>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit

Hi,


On 5/15/24 22:58, Maxime Ripard wrote:
> On Wed, May 15, 2024 at 10:53:00PM +0800, Sui Jingfeng wrote:
>> On 5/15/24 22:30, Maxime Ripard wrote:
>>> On Wed, May 15, 2024 at 12:53:33AM +0800, Sui Jingfeng wrote:
>>>> On 2024/5/15 00:22, Maxime Ripard wrote:
>>>>> Hi,
>>>>>
>>>>> On Tue, May 14, 2024 at 11:40:43PM +0800, Sui Jingfeng wrote:
>>>>>> Because a lot of implementations has already added it into their drived
>>>>>> class, promote it into drm_bridge core may benifits a lot. drm bridge is
>>>>>> a driver, it should know the underlying hardware entity.
>>>>> Is there some actual benefits, or is it theoretical at this point?
>>>>
>>>>
>>>> I think, DRM bridge drivers could remove the 'struct device *dev'
>>>> member from their derived structure. Rely on the drm bridge core
>>>> when they need the 'struct device *' pointer.
>>>
>>> Sure, but why do we need to do so?
>>>
>>> The other thread you had with Jani points out that it turns out that
>>> things are more complicated than "every bridge driver has a struct
>>> device anyway", it creates inconsistency in the API (bridges would have
>>> a struct device, but not other entities), and it looks like there's no
>>> use for it anyway.
>>>
>>> None of these things are deal-breaker by themselves, but if there's only
>>> downsides and no upside, it's not clear to me why we should do it at all.
>>
>> It can reduce boilerplate.
> 
> You're still using a conditional here.


It's for safety reason, prevent NULL pointer dereference.
drm bridge can be seen as either a software entity or a device driver.

It's fine to pass NULL if specific KMS drivers intend to see
drm bridge as a pure software entity, and for internal use only.
Both use cases are valid.

> 
> Maxime

-- 
Best regards
Sui