Re: [PATCH][V2][next] rtc: ds1307: check for failed memory allocation on wdt

[Alexandre Belloni <alexandre.belloni@bootlin.com>]

On 02/04/2020 07:44:44-0700, Guenter Roeck wrote:
> On 4/2/20 6:52 AM, Colin King wrote:
> > From: Colin Ian King <colin.king@canonical.com>
> > 
> > Currently a failed memory allocation will lead to a null pointer
> > dereference on point wdt.  Fix this by checking for a failed allocation
> > and adding error return handling to function ds1307_wdt_register.
> > Also move the error exit label "exit" to allow a return statement to
> > be removed.
> > 
> > Addresses-Coverity: ("Dereference null return")
> > Fixes: fd90d48db037 ("rtc: ds1307: add support for watchdog timer on ds1388")
> > Signed-off-by: Colin Ian King <colin.king@canonical.com>
> > ---
> > V2: move error exit label and remove a return statement, thanks to 
> >     Walter Harms for spotting this clean up.
> > ---
> >  drivers/rtc/rtc-ds1307.c | 16 +++++++++-------
> >  1 file changed, 9 insertions(+), 7 deletions(-)
> > 
> > diff --git a/drivers/rtc/rtc-ds1307.c b/drivers/rtc/rtc-ds1307.c
> > index fad042118862..c058b02efb4d 100644
> > --- a/drivers/rtc/rtc-ds1307.c
> > +++ b/drivers/rtc/rtc-ds1307.c
> > @@ -1665,14 +1665,16 @@ static const struct watchdog_ops ds1388_wdt_ops = {
> >  
> >  };
> >  
> > -static void ds1307_wdt_register(struct ds1307 *ds1307)
> > +static int ds1307_wdt_register(struct ds1307 *ds1307)
> >  {
> >  	struct watchdog_device	*wdt;
> >  
> >  	if (ds1307->type != ds_1388)
> > -		return;
> > +		return 0;
> >  
> >  	wdt = devm_kzalloc(ds1307->dev, sizeof(*wdt), GFP_KERNEL);
> > +	if (!wdt)
> > +		return -ENOMEM;
> >  
> >  	wdt->info = &ds1388_wdt_info;
> >  	wdt->ops = &ds1388_wdt_ops;
> > @@ -1683,10 +1685,13 @@ static void ds1307_wdt_register(struct ds1307 *ds1307)
> >  	watchdog_init_timeout(wdt, 0, ds1307->dev);
> >  	watchdog_set_drvdata(wdt, ds1307);
> >  	devm_watchdog_register_device(ds1307->dev, wdt);
> > +
> > +	return 0;
> >  }
> >  #else
> > -static void ds1307_wdt_register(struct ds1307 *ds1307)
> > +static int ds1307_wdt_register(struct ds1307 *ds1307)
> >  {
> > +	return 0;
> >  }
> >  #endif /* CONFIG_WATCHDOG_CORE */
> >  
> > @@ -1979,10 +1984,7 @@ static int ds1307_probe(struct i2c_client *client,
> >  
> >  	ds1307_hwmon_register(ds1307);
> >  	ds1307_clks_register(ds1307);
> > -	ds1307_wdt_register(ds1307);
> > -
> > -	return 0;
> > -
> > +	err = ds1307_wdt_register(ds1307);
> 
> Ah, sorry, missed this one. The original idea was to ignore errors on purpose.
> Same as with hwmon. If you want to change this, fine with me. Note though
> that rtc_nvmem_register() now leaks a sysfs file if I understand the code
> correctly.

rtc_nvmem_unregister(rtc) should be called properly by
devm_rtc_release_device when the rtc_device is destroyed so I don't
think it leaks.

As stated, I also prefer keeping the watchdog optional and ignore the
error.

-- 
Alexandre Belloni, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com