[RFC PATCH 2/3] pinctrl: imx: add pinmux-imx53 support

Dong Aisheng-B29396 B29396 at freescale.com
Tue Dec 6 02:21:36 EST 2011


> -----Original Message-----
> From: Guo Shawn-R65073
> Sent: Tuesday, December 06, 2011 2:58 PM
> To: Dong Aisheng-B29396
> Cc: Sascha Hauer; Linus Walleij; linux-kernel at vger.kernel.org; linux-arm-
> kernel at lists.infradead.org; linus.walleij at stericsson.com; Guo Shawn-
> R65073; kernel at pengutronix.de
> Subject: Re: [RFC PATCH 2/3] pinctrl: imx: add pinmux-imx53 support
> 
> On Tue, Dec 06, 2011 at 01:54:35PM +0800, Dong Aisheng-B29396 wrote:
> > > -----Original Message-----
> > > From: Sascha Hauer [mailto:s.hauer at pengutronix.de]
> > > Sent: Tuesday, December 06, 2011 5:19 AM
> > > To: Linus Walleij
> > > Cc: Dong Aisheng-B29396; linux-kernel at vger.kernel.org; linux-arm-
> > > kernel at lists.infradead.org; linus.walleij at stericsson.com; Guo Shawn-
> > > R65073; kernel at pengutronix.de
> > > Subject: Re: [RFC PATCH 2/3] pinctrl: imx: add pinmux-imx53 support
> > >
> > > On Mon, Dec 05, 2011 at 05:57:42PM +0100, Linus Walleij wrote:
> > > > On Sun, Dec 4, 2011 at 12:49 PM, Dong Aisheng
> > > > <b29396 at freescale.com>
> > > wrote:
> > > >
> > > > > +enum imx_mx53_pads {
> > > > > +       MX53_GPIO_19 = 0,
> > > > > +       MX53_KEY_COL0 = 1,
> > > > (...)
> > > >
> > > > First I thought it looked a bit strange since you needed enums for
> > > > all pads but then I realized that your macros use the same
> > > > enumerator name to name the pad and then it looks sort of clever.
> > > >
> > > > But maybe put in a comment about that here:
> > > >
> > > > > +/* Pad names for the pinmux subsystem */
> > > >
> > > > Like this:
> > > >
> > > > /*
> > > >  * Pad names for the pinmux subsystem.
> > > >  * These pad names are constructed from the pin enumerator names
> > > >  * in the IMX_PINCTRL_PIN() macro.
> > > >  */
> > > >
> > > > > +static const struct pinctrl_pin_desc mx53_pads[] = {
> > > > > +       IMX_PINCTRL_PIN(MX53_GPIO_19),
> > > > > +       IMX_PINCTRL_PIN(MX53_KEY_COL0),
> > > > (...)
> > > >
> > > > > +/* mx53 pin groups and mux mode */ static const unsigned
> > > > > +mx53_fec_pins[] = {
> > > > > +       MX53_FEC_MDC,
> > > > > +       MX53_FEC_MDIO,
> > > > > +       MX53_FEC_REF_CLK,
> > > > > +       MX53_FEC_RX_ER,
> > > > > +       MX53_FEC_CRS_DV,
> > > > > +       MX53_FEC_RXD1,
> > > > > +       MX53_FEC_RXD0,
> > > > > +       MX53_FEC_TX_EN,
> > > > > +       MX53_FEC_TXD1,
> > > > > +       MX53_FEC_TXD0,
> > > > > +};
> > > >
> > > > I understand this.
> > > >
> > > > > +static const unsigned mx53_fec_mux[] = { 0, 0, 0, 0, 0, 0, 0,
> > > > > +0, 0, 0 };
> > > >
> > > > But what is this? Just zeroes? Why?
> > > > Especially with a const so they really cannot be anything else.
> > > > The same pin (0) can only be enumerated once.
> > > >
> > > > > +static const unsigned mx53_sd1_pins[] = {
> > > > > +       MX53_SD1_CMD,
> > > > > +       MX53_SD1_CLK,
> > > > > +       MX53_SD1_DATA0,
> > > > > +       MX53_SD1_DATA1,
> > > > > +       MX53_SD1_DATA2,
> > > > > +       MX53_SD1_DATA3,
> > > > > +
> > > > > +};
> > > > > +static const unsigned mx53_sd1_mux[] = { 0, 0, 0, 0, 0, 0 };
> > > >
> > > > And here again.
> > > >
> > > > > +static const unsigned mx53_sd3_pins[] = {
> > > > > +       MX53_PATA_DATA8,
> > > > > +       MX53_PATA_DATA9,
> > > > > +       MX53_PATA_DATA10,
> > > > > +       MX53_PATA_DATA11,
> > > > > +       MX53_PATA_DATA0,
> > > > > +       MX53_PATA_DATA1,
> > > > > +       MX53_PATA_DATA2,
> > > > > +       MX53_PATA_DATA3,
> > > > > +       MX53_PATA_IORDY,
> > > > > +       MX53_PATA_RESET_B,
> > > > > +
> > > > > +};
> > > > > +static const unsigned mx53_sd3_mux[] = { 4, 4, 4, 4, 4, 4, 4,
> > > > > +4, 2,
> > > > > +2 };
> > > >
> > > > This also looks strange. Can you explain what these muxes are?
> > >
> > > Freescale has named the pins after their primary function which is
> > > quite confusing.
> > >
> > > The above means:
> > >
> > > MX53_PATA_DATA8 -> mux mode 4
> > > MX53_PATA_DATA9 -> mux mode 4
> > > ...
> > >
> > > This brings me to the point that currently we have the pins
> > > described as
> > >
> > > #define MX53_PAD_<name>__<function>
> > >
> > > which means that you don't have to look into the datasheet to get
> > > the different options for a pin (and don't have a chance to get it
> wrong).
> > > I don't really want to lose this.
> > >
> > Obviously current used pin defines in that way is pretty good.
> > And I also don't want to lose this.
> >
> > Actually I also have tried to see if we can reuse the current iomux-v3
> code.
> >
> > For current pinmux driver, one approach I can see is that define mux
> > in enum for each pin like:
> >
> > enum MX53_PAD_GPIO_19_MUX {
> >         MX53_PAD_GPIO_19__KPP_COL_5,
> >         MX53_PAD_GPIO_19__GPIO4_5,
> >         MX53_PAD_GPIO_19__CCM_CLKO,
> >         MX53_PAD_GPIO_19__SPDIF_OUT1,
> >         MX53_PAD_GPIO_19__RTC_CE_RTC_EXT_TRIG2,
> >         MX53_PAD_GPIO_19__ECSPI1_RDY,
> >         MX53_PAD_GPIO_19__FEC_TDATA_3,
> >         MX53_PAD_GPIO_19__SRC_INT_BOOT, };
> 
> I would say, no, do not do that, because it simply does not worth.
> Most of the definitions will probably never be used.
> 
> IMO, we can just focus on the support for device tree case (imx6) for now.
> With proper DT binding for pinctrl settled, all these data can go into DT.
> For those non-DT cases, we may want to leave them as they are for now.
> 

Can current pinctrl framework support DT well?
Linus,
Can you help answer it or if you have a plan on DT support if it's still not ready?

I was ever thought it might not support DT that why I changed to run the
driver for imx53 first.

> 
> > Then put them in a common file for each mx53 based board to use.
> >
> > Take uart1 as an example, it could be:
> > static const unsigned mx53_uart1_pins[] = {
> >         MX53_CSI0_DAT10,
> >         MX53_CSI0_DAT11,
> > };
> >
> > static const unsigned mx53_uart1_mux[] = {
> >         MX53_CSI0_DAT10__UART1_TXD_MUX,
> >         MX53_CSI0_DAT11__UART1_RXD_MUX };
> >
> > static const struct imx_pin_group mx53_pin_groups[] = {
> >         IMX_PIN_GROUP("uart1grp", mx53_uart1_pins, mx53_uart1_mux), };
> >
> > The benefit is that it's very clear and good maintainable.
> > The defect is it will increase the code size a lot by defining all
> > pin's mux enum and each pin's mux array in board file.
> >
> > Do you think if it's ok or you have any better idea?
> >
> > Regards
> > Dong Aisheng




More information about the linux-arm-kernel mailing list