[PATCH 016/171] iommu: Add a Kconfig for SPL_PINCTRL_ARMADA_38X

[Pali Rohár]

On Monday 30 January 2023 13:59:21 Tom Rini wrote:
> On Mon, Jan 30, 2023 at 07:28:56PM +0100, Pali Rohár wrote:
> > On Monday 30 January 2023 13:25:25 Tom Rini wrote:
> > > On Mon, Jan 30, 2023 at 07:16:26PM +0100, Pali Rohár wrote:
> > > > On Monday 30 January 2023 11:29:27 Tom Rini wrote:
> > > > > On Mon, Jan 30, 2023 at 05:19:01PM +0100, Mark Kettenis wrote:
> > > > > > > From: Simon Glass <sjg at chromium.org>
> > > > > > > Date: Mon, 30 Jan 2023 07:40:49 -0700
> > > > > > > 
> > > > > > > This is implicitly used in the source and seems useful, so add it.
> > > > > > 
> > > > > > Not sure how this ended up with an "iommu" tag, but that seems wrong.
> > > > > > 
> > > > > > > Signed-off-by: Simon Glass <sjg at chromium.org>
> > > > > > > ---
> > > > > > > 
> > > > > > >  drivers/pinctrl/mvebu/Kconfig | 4 ++++
> > > > > > >  1 file changed, 4 insertions(+)
> > > > > > > 
> > > > > > > diff --git a/drivers/pinctrl/mvebu/Kconfig b/drivers/pinctrl/mvebu/Kconfig
> > > > > > > index 7c51d138c8b..0b2be88e3a1 100644
> > > > > > > --- a/drivers/pinctrl/mvebu/Kconfig
> > > > > > > +++ b/drivers/pinctrl/mvebu/Kconfig
> > > > > > > @@ -7,6 +7,10 @@ config PINCTRL_ARMADA_38X
> > > > > > >  	   Support pin multiplexing and pin configuration control on
> > > > > > >  	   Marvell's Armada-38x SoC.
> > > > > > >  
> > > > > > > +config SPL_PINCTRL_ARMADA_38X
> > > > > > > +	def_bool n  # Armada 38x pin control driver (SPL)
> > > > > > > +	depends on SPL && ARMADA_38X && SPL_PINCTRL_FULL
> > > > > > > +
> > > > > > >  config PINCTRL_ARMADA_37XX
> > > > > > >  	depends on ARMADA_3700 && PINCTRL_FULL
> > > > > > >  	bool "Armada 37xx pin control driver"
> > > > > 
> > > > > Looking at the code in question, it's because in
> > > > > drivers/gpio/mvebu_gpio.c we conditionally not set request / rfree
> > > > > dm_gpio_ops because turris_omnia (only platform in question) does not
> > > > > set pinctrl in SPL, but does set SPL_DM_GPIO. So the question I have, is
> > > > > all of that intentional and used today, in SPL, on the platform? I guess
> > > > > some set/get directions / values, to check board revs or something?
> > > > 
> > > > I think this is because no A38x board use DM pinctrl framework for
> > > > configuring pin muxing. All boards set pin muxing in open coded board
> > > > function which directly touch HW registers. DM pinctrl driver for A38x
> > > > is relatively new and I think nobody converted any board to use it for
> > > > initialization.
> > > 
> > > I do see that turris_omnia builds the pinctrl driver for full U-Boot
> > > today.
> > 
> > Yes, it allows u-boot pinmux command to print current pinctrl
> > configuration and also allows gpio command to work correctly for gpios
> > which are not in gpio mode by default.
> > 
> > > But, does that mean that at some point in the future it should
> > > be converted to use SPL_PINCTRL ?
> > 
> > Every a38x board could be converted. But personally I'm not planning to
> > do it.
> 
> OK, but if someone were to do it, it would then be the case that
> PINCTRL_ARMADA_38X would be the driver used, yes?

Then it would be used in SPL. In proper U-Boot it is already used by
commands like 'pinmux' or 'gpio'.

> > > As that would imply that we do not
> > > need / want SPL_PINCTRL_ARMADA_38X added (I suspect SPL_PINCTRL_STMFX
> > > and SPL_PINCTRL_ROCKCHIP are wrong) as it will be handled like the
> > > majority of other platforms in this area.
> 
> As I look at this more, I see that SPL_PINCTRL_STMFX _is_ one of the
> special cases as it seems in SPL they use the PINCTRL_STM32 driver and
> do _not_ want PINCTRL_STMFX so SPL_PINCTRL_STMFX is introduced as a way
> to keep the driver from being built in SPL.
> 
> -- 
> Tom