Re: [PATCH 2/2] [RFC] rtc: pcf2127: only use watchdog when explicitly available - Uwe Kleine-König

From: "Uwe Kleine-König" <u.kleine-koenig@pengutronix.de>
To: Guenter Roeck <linux@roeck-us.net>
Cc: Bruno Thomsen <bruno.thomsen@gmail.com>,
	linux-rtc@vger.kernel.org,
	Alessandro Zummo <a.zummo@towertech.it>,
	Alexandre Belloni <alexandre.belloni@bootlin.com>,
	linux-watchdog@vger.kernel.org,
	"open list:OPEN FIRMWARE AND FLATTENED DEVICE TREE BINDINGS" 
	<devicetree@vger.kernel.org>,
	open list <linux-kernel@vger.kernel.org>,
	Rob Herring <robh+dt@kernel.org>,
	Sascha Hauer <kernel@pengutronix.de>,
	Wim Van Sebroeck <wim@linux-watchdog.org>,
	Qiang Zhao <qiang.zhao@nxp.com>
Subject: Re: [PATCH 2/2] [RFC] rtc: pcf2127: only use watchdog when explicitly available
Date: Mon, 28 Sep 2020 10:43:43 +0200	[thread overview]
Message-ID: <20200928084343.cl42lxsiionvq7tg@pengutronix.de> (raw)
In-Reply-To: <8f88f2a2-4a6d-021f-4404-f05518b0477d@roeck-us.net>

[-- Attachment #1: Type: text/plain, Size: 2502 bytes --]

On Sun, Sep 27, 2020 at 08:54:47AM -0700, Guenter Roeck wrote:
> On 9/27/20 1:09 AM, Bruno Thomsen wrote:
> > Den tor. 24. sep. 2020 kl. 12.53 skrev Uwe Kleine-König
> > <u.kleine-koenig@pengutronix.de>:
> >>
> >> Most boards using the pcf2127 chip (in my bubble) don't make use of the
> >> watchdog functionality and the respective output is not connected. The
> >> effect on such a board is that there is a watchdog device provided that
> >> doesn't work.
> >>
> >> So only register the watchdog if the device tree has a "has-watchdog"
> >> property.
> >>
> >> Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
> >> ---
> >>  drivers/rtc/rtc-pcf2127.c | 3 ++-
> >>  1 file changed, 2 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/drivers/rtc/rtc-pcf2127.c b/drivers/rtc/rtc-pcf2127.c
> >> index 5b1f1949b5e5..8bd89d641578 100644
> >> --- a/drivers/rtc/rtc-pcf2127.c
> >> +++ b/drivers/rtc/rtc-pcf2127.c
> >> @@ -340,7 +340,8 @@ static int pcf2127_watchdog_init(struct device *dev, struct pcf2127 *pcf2127)
> >>         u32 wdd_timeout;
> >>         int ret;
> >>
> >> -       if (!IS_ENABLED(CONFIG_WATCHDOG))
> >> +       if (!IS_ENABLED(CONFIG_WATCHDOG) ||
> >> +           !device_property_read_bool(dev, "has-watchdog"))
> >>                 return 0;
> > 
> > I don't think the compiler can remove the function if
> > CONFIG_WATCHDOG is disabled due to the device tree
> > value check. Maybe it can if split into 2 conditions.
> > 
> 
> If the first part of the expression is always false, the second
> part should not even be evaluated.

This is wrong. For || the second expression isn't evaluated if the first
evaluates to true (and the whole expression becomes true). This is the
intended behaviour: If CONFIG_WATCHDOG is off, we don't need to check
for the dt property and just skip the watchdog part.

> Either case, the code now hard depends on the compiler optimizing the
> code away.
> 
> It calls devm_watchdog_register_device() which doesn't exist
> if CONFIG_WATCHDOG is not enabled. I didn't know that this is safe,
> and I would personally not want to rely on it, but we live and
> learn.

AFAICT this is save and used in other places in the kernel, too.  This
is one of the reasons why you cannot compile the kernel with -O0.

Best regards
Uwe

-- 
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | https://www.pengutronix.de/ |

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]