Re: [PATCH] pinctrl: meson: fix G12A ao pull registers base address

[Jerome Brunet]

On Mon, 2018-12-03 at 11:27 +0100, Neil Armstrong wrote:
> Hi Xingyu,
> 
> 
> On 03/12/2018 04:05, Xingyu Chen wrote:
> > Since Meson G12A SoC, Introduce new ao registers AO_RTI_PULL_UP_EN_REG
> > and AO_GPIO_O.
> > 
> > These bits of controlling output level are remapped to the new register
> > AO_GPIO_O, and the AO_GPIO_O_EN_N support only controlling output enable.
> > 
> > These bits of controlling pull enable are remapped to the new register
> > AO_RTI_PULL_UP_EN_REG, and the AO_RTI_PULL_UP_REG support only controlling
> > pull type(up/down).
> > 
> > The new layout of ao gpio/pull registers is as follows:
> > - AO_GPIO_O_EN_N        [offset: 0x9 << 2]
> > - AO_GPIO_I             [offset: 0xa << 2]
> > - AO_RTI_PULL_UP_REG    [offset: 0xb << 2]
> > - AO_RTI_PULL_UP_EN_REG [offset: 0xc << 2]
> > - AO_GPIO_O             [offset: 0xd << 2]
> > 
> > From above, we can see ao GPIO registers region has been separated by the
> > ao pull registers. In order to ensure the continuity of the region on
> > software, the ao GPIO and ao pull registers use the same base address, but
> > can be identified by the offset.
> > 
> > Fixes: 29ae0952e85f ("pinctrl: meson-g12a: add pinctrl driver support")
> > Signed-off-by: Xingyu Chen <xingyu.chen@xxxxxxxxxxx>
> > Signed-off-by: Jianxin Pan <jianxin.pan@xxxxxxxxxxx>
> > ---
> >  drivers/pinctrl/meson/pinctrl-meson.c | 22 ++++++++++++----------
> >  1 file changed, 12 insertions(+), 10 deletions(-)
> > 
> > diff --git a/drivers/pinctrl/meson/pinctrl-meson.c
> > b/drivers/pinctrl/meson/pinctrl-meson.c
> > index 53d449076dee..7ff40cd7a0cb 100644
> > --- a/drivers/pinctrl/meson/pinctrl-meson.c
> > +++ b/drivers/pinctrl/meson/pinctrl-meson.c
> > @@ -31,6 +31,9 @@
> >   * In some cases the register ranges for pull enable and pull
> >   * direction are the same and thus there are only 3 register ranges.
> >   *
> > + * Since Meson G12A SoC, the ao register ranges for gpio, pull enable
> > + * and pull direction are the same, so there are only 2 register ranges.
> > + *
> >   * For the pull and GPIO configuration every bank uses a contiguous
> >   * set of bits in the register sets described above; the same register
> >   * can be shared by more banks with different offsets.
> > @@ -487,23 +490,22 @@ static int meson_pinctrl_parse_dt(struct
> > meson_pinctrl *pc,
> >  		return PTR_ERR(pc->reg_mux);
> >  	}
> >  
> > -	pc->reg_pull = meson_map_resource(pc, gpio_np, "pull");
> > -	if (IS_ERR(pc->reg_pull)) {
> > -		dev_err(pc->dev, "pull registers not found\n");
> > -		return PTR_ERR(pc->reg_pull);
> > +	pc->reg_gpio = meson_map_resource(pc, gpio_np, "gpio");
> > +	if (IS_ERR(pc->reg_gpio)) {
> > +		dev_err(pc->dev, "gpio registers not found\n");
> > +		return PTR_ERR(pc->reg_gpio);
> >  	}
> >  
> > +	pc->reg_pull = meson_map_resource(pc, gpio_np, "pull");
> > +	/* Use gpio region if pull one is not present */
> > +	if (IS_ERR(pc->reg_pull))
> > +		pc->reg_pull = pc->reg_gpio;
> > +
> >  	pc->reg_pullen = meson_map_resource(pc, gpio_np, "pull-enable");
> >  	/* Use pull region if pull-enable one is not present */
> >  	if (IS_ERR(pc->reg_pullen))
> >  		pc->reg_pullen = pc->reg_pull;
> >  
> > -	pc->reg_gpio = meson_map_resource(pc, gpio_np, "gpio");
> > -	if (IS_ERR(pc->reg_gpio)) {
> > -		dev_err(pc->dev, "gpio registers not found\n");
> > -		return PTR_ERR(pc->reg_gpio);
> > -	}
> > -
> >  	return 0;
> >  }
> >  
> > 
> Doesn't it need an update of the bindings ?

Going even further, shouldn't we stop trying make multiple regions out of
this, and have just one ? 

On all the Amlogic SoC we have seen so far, all the regions a very (VERY)
close to each other. It seems very unlikely that there something unrelated to
GPIO in between.

It looks like everything is mostly there in the driver to deal with offset, so
change would be minimal.

Of course, for  DT stability we will need to carry the legacy, but for newer
SoC, such as the g12, does it really makes sense to have multiple regions ?

> 
> Neil
>