Re: [PATCH v2 06/16] watchdog: hpwdt: drop warning after calling watchdog_init_timeout

[Guenter Roeck]

On Tue, Apr 16, 2019 at 02:34:31PM -0600, Jerry Hoemann wrote:
> On Tue, Apr 16, 2019 at 12:25:05PM +0200, Wolfram Sang wrote:
> > The core will print out details now.
> > 
> > Reviewed-by: Guenter Roeck <linux@xxxxxxxxxxxx>
> > Signed-off-by: Wolfram Sang <wsa+renesas@xxxxxxxxxxxxxxxxxxxx>
> > ---
> >  drivers/watchdog/hpwdt.c | 3 +--
> >  1 file changed, 1 insertion(+), 2 deletions(-)
> > 
> > diff --git a/drivers/watchdog/hpwdt.c b/drivers/watchdog/hpwdt.c
> > index ef30c7e9728d..db1bf6f546ae 100644
> > --- a/drivers/watchdog/hpwdt.c
> > +++ b/drivers/watchdog/hpwdt.c
> > @@ -311,8 +311,7 @@ static int hpwdt_init_one(struct pci_dev *dev,
> >  		goto error_init_nmi_decoding;
> >  
> >  	watchdog_set_nowayout(&hpwdt_dev, nowayout);
> > -	if (watchdog_init_timeout(&hpwdt_dev, soft_margin, NULL))
> > -		dev_warn(&dev->dev, "Invalid soft_margin: %d.\n", soft_margin);
> > +	watchdog_init_timeout(&hpwdt_dev, soft_margin, NULL);
> 
> I applied patches 1,2 & 6 in testing.
> 
> Note, that hpwdt is passing NULL as the third parameter to watchdog_init_timeout().
> 
> The second patch in this series is using "dev" as input to dev_err and dev_warn.
> 
> This results in the following in dmesg when trying to load hpwdt w/ an invalid soft_margin:
> 
> 
> [   80.848160] (NULL device *): driver supplied timeout (4294967295) out of range
> [   80.855429] (NULL device *): falling back to default timeout (30)
> 
Good find. Thanks a lot for testing!

We'll have to address this. Wolfram - it looks like we'll need
separate error message handling for situations where dev is NULL.
We may have to leave it up to the parent after all to display
a message in that case (since we do want to see the driver name).

> 
> if the call in hpwdt driver is changed to:
> 
> 	if (watchdog_init_timeout(&hpwdt_dev, soft_margin, &dev->dev))
> 
> 
> We see the message like we'd desire:
> 
> [ 2061.167100] hpwdt 0000:01:00.0: driver supplied timeout (4294967295) out of range
> [ 2061.174633] hpwdt 0000:01:00.0: falling back to default timeout (30)
> 
> 
> 
> watchdog_init_timeout() uses dev to "try to get the timeout_sec property"
> 
> I am not familiar with this part of the core and what effect having the hpwdt
> driver pass in dev to watchdog_init_timeout would have.  (I.e. is the
> change safe?)
> 
Yes, in general it is safe. watchdog_init_timeout() only uses dev to extract
a timeout value from devicetree (and now to display error messages).

> Guenter, can you help on this question?
> 
> Note, hpwdt isn't the only watch dog device that is passing NULL to
> watchdog_init_timeout.
> 
That is indeed a problem: the pointer will be NULL if there is no parent
device (such as in softdog.c). Otherwise it should never be NULL.

Thanks,
Guenter