Re: [PATCH 1/2] drivers/clocksource/timer-of: Remove __init markings

From: Daniel Lezcano <daniel.lezcano@linaro.org>
To: Saravana Kannan <saravanak@google.com>
Cc: Baolin Wang <baolin.wang7@gmail.com>,
	Thomas Gleixner <tglx@linutronix.de>,
	Orson Zhai <orsonzhai@gmail.com>,
	Chunyan Zhang <zhang.lyra@gmail.com>,
	LKML <linux-kernel@vger.kernel.org>,
	Android Kernel Team <kernel-team@android.com>
Subject: Re: [PATCH 1/2] drivers/clocksource/timer-of: Remove __init markings
Date: Mon, 27 Apr 2020 22:50:24 +0200	[thread overview]
Message-ID: <573b7d9e-fb5d-5c81-a981-6d965db933f3@linaro.org> (raw)
In-Reply-To: <CAGETcx9sc6oWSPK=aWJ+39i9eHAyg6iOf+Xf2oVECO+g1pFjtQ@mail.gmail.com>

On 27/04/2020 22:12, Saravana Kannan wrote:
> On Mon, Apr 27, 2020 at 1:09 PM Daniel Lezcano
> <daniel.lezcano@linaro.org> wrote:
>>
>> On 27/04/2020 21:04, Saravana Kannan wrote:
>>> On Mon, Apr 27, 2020 at 10:13 AM Daniel Lezcano
>>> <daniel.lezcano@linaro.org> wrote:
>>>>
>>>> On 13/04/2020 04:55, Baolin Wang wrote:
>>>>> Hi Daniel,
>>>>>
>>>>> On Tue, Mar 24, 2020 at 1:59 PM Baolin Wang <baolin.wang7@gmail.com> wrote:
>>>>>>
>>>>>> From: Saravana Kannan <saravanak@google.com>
>>>>>>
>>>>>> This allows timer drivers to be compiled as modules.
>>>>>>
>>>>>> Signed-off-by: Saravana Kannan <saravanak@google.com>
>>>>>> Signed-off-by: Baolin Wang <baolin.wang7@gmail.com>
>>>>>
>>>>> Do you have any comments for this patch set? Thanks.
>>>>
>>>> If my understanding is correct, this patch is part of the GKI picture
>>>> where hardware drivers are converted to modules.
>>>>
>>>> But do we really want to convert timer drivers to modules ?
>>>>
>>>> Is the core time framework able to support that (eg. load + unload )
>>>
>>> So this will mainly be used for secondary timers that the system
>>> supports. Not for the main one that's set up during early boot for
>>> sched timer to work. For the primary timer during boot up, we still
>>> expect that to be the default ARM timer and don't want/expect that to
>>> be a module (it can't be).
>>
>> My question is about clockevents_config_and_register() for instance, is
>> there a function to unregister in the core framework ?
> 
> We can just have these modules be "permanent" modules that can't be
> unloaded. They just need to not implement module_exit().

You are right.

I can understand the goal of making everything as much modular as possible.

But TBH, I have a bad feeling about this. Sounds like GKI will give the
opportunity to companies to stop upstreaming their drivers and favoring
fragmentation like what we had several years ago. Not sure it is a good
thing, especially for upstream SoC support.

-- 
<http://www.linaro.org/> Linaro.org │ Open source software for ARM SoCs

Follow Linaro:  <http://www.facebook.com/pages/Linaro> Facebook |
<http://twitter.com/#!/linaroorg> Twitter |
<http://www.linaro.org/linaro-blog/> Blog