Received-SPF: pass (google.com: domain of linux-kernel+bounces-89889-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.80.249 as permitted sender) client-ip=147.75.80.249;
Message-ID: <010ea3a0-929e-4912-ad22-9f0cf5b1a3e2@redhat.com>
Date: Sun, 3 Mar 2024 20:53:10 +0100
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Content-Language: en-US, nl
To: Yoshinori Sato <ysato@users.sourceforge.jp>, Rich Felker
 <dalias@libc.org>, John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>,
 "Rafael J. Wysocki" <rafael@kernel.org>, Pavel Machek <pavel@ucw.cz>,
 Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: linux-sh@vger.kernel.org,
 Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
 Linux PM <linux-pm@vger.kernel.org>
From: Hans de Goede <hdegoede@redhat.com>
Subject: pm_runtime_early_init() defined but not used, except on SuperH which
 has its own definition ?
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 7bit

Hi All,

I noticed that drivers/base/power/power.h defines pm_runtime_early_init()
but nothing under drivers/base uses this.

A grep over the entire tree shows that arch/sh/drivers/platform_early.c
does use pm_runtime_early_init() but rather then including
drivers/base/power/power.h it has its own definition / private copy
of both device_pm_init_common() and pm_runtime_early_init() from
drivers/base/power/power.h ???

Also the private copy of pm_runtime_early_init() in
arch/sh/drivers/platform_early.c differs from the unused one
in drivers/base/power/power.h, but only when CONFIG_PM is not set.

When CONFIG_PM is not set then the pm_runtime_early_init() in
arch/sh/drivers/platform_early.c is a no-op, where as the one in
drivers/base/power/power.h still calls device_pm_init_common()
in this case ...

I also wonder if given that pm_runtime_early_init() is not
used with the exception of arch/sh/drivers/platform_early.c
if the dev->power.early_init flag check in
device_pm_init_common() is really necessary ?

On non SuperH the only (1) caller of device_pm_init_common()
is device_pm_init(), so it seems to me that the code to
avoid doing device_pm_init_common() twice is unnecessary.

Actually it seems to me that the entire contents of
device_pm_init_common() can be moved inside device_pm_init()
and the dev->power.early_init can be completely dropped (2).

Regards,

Hans


1) Well pm_runtime_early_init() calls it too, but that itself
is unused and can be removed, removing it is even ok-ish
for SuperH since that has its own copy anyways.

2) With the exception that all of this is still necessary
for SuperH I guess.