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

From: Guenter Roeck <linux@roeck-us.net>
To: "Uwe Kleine-König" <u.kleine-koenig@pengutronix.de>
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 09:26:06 -0700	[thread overview]
Message-ID: <20200928162606.GA106276@roeck-us.net> (raw)
In-Reply-To: <20200928084343.cl42lxsiionvq7tg@pengutronix.de>

On Mon, Sep 28, 2020 at 10:43:43AM +0200, Uwe Kleine-König wrote:
> 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.
> 
Sorry, I meant to say "If the first part of the expression is always true".

Guenter

> > 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/ |