[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <8734ekwkqk.ffs@tglx>
Date: Mon, 07 Apr 2025 08:26:11 +0200
From: Thomas Gleixner <tglx@...utronix.de>
To: Maxim Kochetkov <fido_max@...ox.ru>, linux-kernel@...r.kernel.org,
linux-riscv@...ts.infradead.org, peterz@...radead.org, elver@...gle.com,
namcao@...utronix.de, samuel.holland@...ive.com,
daniel.lezcano@...aro.org, apatel@...tanamicro.com
Cc: Maxim Kochetkov <fido_max@...ox.ru>
Subject: Re: [PATCH 1/1] time/sched_clock: move sched_clock_register() out
of .init section
On Fri, Apr 04 2025 at 08:05, Maxim Kochetkov wrote:
> The sched_clock_register() is widely used by clocksource timer
> drivers. The __init prefix forces them to be initialized using
> macro TIMER_OF_DECLARE with __init prefixed function.
No, it does not. It requires that they are built in, not more.
> Clocksource devices can be consumers of some external irq, clocks,
> resets, e.t.c. Such devices can't be correctly probed if this
> dependencies are provided by platform drivers. Because of regular
> platform devices are not probed at this moment.
>
> We can convert clocksource drivers to platform device drivers to
> fix this issue, but __init prefix in sched_clock_register()
> prevents it.
Again. It does not. What the __init prefix prevents is that the driver
can be built as a module and loaded late.
> So lets drop __init prefix to allow platform device drivers to use
> sched_clock_register().
s/So let's//
"So let's" means nothing.
Also this has nothing to do with platform device drivers. It's all about
modules and nothing else.
Thanks,
tglx
Powered by blists - more mailing lists