Re: Suspicious debounce handling code in pintctrl-baytrail

[Mika Westerberg <mika.westerberg@xxxxxxxxxxxxxxx>]

On Thu, Jan 26, 2017 at 03:00:46PM +0100, Jean Delvare wrote:
> Hi Mika, hello Andy,
> 
> On Thu, 26 Jan 2017 15:43:41 +0200, Mika Westerberg wrote:
> > On Thu, Jan 26, 2017 at 02:20:20PM +0100, Jean Delvare wrote:
> > > Hi Cristina,
> > > 
> > > In this commit:
> > > 
> > > commit 658b476c742fe379e7020309fd590a27b457a4c1
> > > Date:   Fri Apr 1 14:00:07 2016 +0300
> > > 
> > >     pinctrl: baytrail: Add debounce configuration
> > > 
> > > you added support for getting and setting debounce configuration for
> > > the Baytrail pins. Now gcc complains about the following:
> > > 
> > >   CC [M]  drivers/pinctrl/intel/pinctrl-baytrail.o
> > > drivers/pinctrl/intel/pinctrl-baytrail.c: In function ‘byt_pin_config_set’:
> > > drivers/pinctrl/intel/pinctrl-baytrail.c:1181:17: warning: variable ‘debounce’ set but not used [-Wunused-but-set-variable]
> > >   u32 conf, val, debounce;
> > >                  ^
> > > 
> > > I looked at the code, and it indeed looks wrong. You are reading the
> > > BYT_DEBOUNCE_REG register, clearing the debounce time bits from it, and
> > > then writing the debounce time bits to the *BYT_CONF0_REG* register.
> > > This is certainly corrupting the chip configuration, as the
> > > configuration register bits 2-0 have a completely different meaning
> > > (mux configuration.)
> > 
> > Yes, you are right.
> > 
> > This is fixed by following commit from Andy:
> > 
> > https://git.kernel.org/cgit/linux/kernel/git/linusw/linux-pinctrl.git/commit/?h=fixes&id=04ff5a095d662e0879f0eb04b9247e092210aeff
> 
> Thanks for the pointer, I wasn't aware of that patch. Looks overall
> good, but:
> 
> +			case 0:
> +				conf &= BYT_DEBOUNCE_EN;
> +				break;
> 
> looks suspicious to me. I think ~BYT_DEBOUNCE_EN was intended?

Good catch! Yes, definitely it should be ~BYT_DEBOUNCE_EN.

> Furthermore I would expect conf |= BYT_DEBOUNCE_EN in the other cases,
> otherwise disabling debounce once will prevent you from enabling it
> forever.
> 
> Also Andy's patch doesn't seem to address my second point:
> 
> > Additionally, I suspect that in the "get" path, the following is wrong:
> > 
> > 	case PIN_CONFIG_INPUT_DEBOUNCE:
> > 		if (!(conf & BYT_DEBOUNCE_EN))
> > 			return -EINVAL;
> > 
> > I'm not familiar with the pinctrl interface but my understanding of
> > the PIN_CONFIG_INPUT_DEBOUNCE documentation is that the driver must
> > return 0 if debouncing is disabled.

Indeed, I also agree it should return 0 here.

We will fix this, thanks!
--
To unsubscribe from this list: send the line "unsubscribe linux-gpio" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html