Re: [PATCH v3 2/3] pinctrl: renesas: pinctrl-rzg2l: Add the missing port pins P19 to P28

From: Lad, Prabhakar
Date: Wed Jan 10 2024 - 13:31:00 EST

Next message: Steven Rostedt: "Re: [PATCH] tracefs/eventfs: Use root and instance inodes as default ownership"
Previous message: William Butler: "[PATCH] nvme: Inform controller of doorbell config before unquiescing IO queues"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

Hi Geert,

Thank you for the review.

On Wed, Dec 6, 2023 at 2:25 PM Geert Uytterhoeven <geert@linux-m68korg> wrote:
>
> Hi Prabhakar,
>
> Thanks for your patch!
>
> On Fri, Dec 1, 2023 at 2:16 PM Prabhakar <prabhakar.csengg@xxxxxxxxx> wrote:
> > From: Lad Prabhakar <prabhakar.mahadev-lad.rj@xxxxxxxxxxxxxx>
> >
> > Add the missing port pins P19 to P28 for RZ/Five SoC. These additional
> > pins provide expanded capabilities and are exclusive to the RZ/Five SoC.
> >
> > Couple of port pins have different configuration and is not identical for
>
> s/is/are/
>
OK.

> > the complete port so introduced struct rzg2l_variable_pin_cfg to handle
>
> introduce
>
OK.

> > such cases and introduced PIN_CFG_VARIABLE macro. The actual pin config is
>
> introduce the
>
OK.

> > then assigned rzg2l_pinctrl_get_variable_pin_cfg().
>
> assigned in
>
OK.

> >
> > Add an additional check in rzg2l_gpio_get_gpioint() to only allow GPIO pins
> > which support interrupt facility.
> >
> > Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@xxxxxxxxxxxxxx>
> > ---
> > drivers/pinctrl/renesas/pinctrl-rzg2l.c | 215 +++++++++++++++++++++++-
> > 1 file changed, 213 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/pinctrl/renesas/pinctrl-rzg2l.c b/drivers/pinctrl/renesas/pinctrl-rzg2l.c
> > index 94d072c8a743..083cc63c2c82 100644
> > --- a/drivers/pinctrl/renesas/pinctrl-rzg2l.c
> > +++ b/drivers/pinctrl/renesas/pinctrl-rzg2l.c
> > @@ -57,6 +57,8 @@
> > #define PIN_CFG_FILCLKSEL BIT(12)
> > #define PIN_CFG_IOLH_C BIT(13)
> > #define PIN_CFG_SOFT_PS BIT(14)
> > +#define PIN_CFG_VARIABLE BIT(15)
> > +#define PIN_CFG_NOGPIO_INT BIT(16)
>
> Note to self: this conflicts with "[PATCH 08/14] pinctrl: renesas:
> rzg2l: Add output enable support", so the numbers need to be adapted.
>
> https://lore.kernel.org/all/20231120070024.4079344-9-claudiu.beznea.uj@xxxxxxxxxxxxx
>
I'll rebase the changes, as the patches have now merged.

> >
> > #define RZG2L_MPXED_COMMON_PIN_FUNCS(group) \
> > (PIN_CFG_IOLH_##group | \
> > @@ -82,6 +84,11 @@
> > */
> > #define RZG2L_GPIO_PORT_PACK(n, a, f) (((n) > 0 ? ((u64)(GENMASK_ULL(((n) - 1 + 28), 28))) : 0) | \
> > ((a) << 20) | (f))
>
> I'd rather define RZG2L_GPIO_PORT_PACK() using RZG2L_GPIO_PORT_SPARSE_PACK():
>
> #define RZG2L_GPIO_PORT_PACK(n, a, f) \
> RZG2L_GPIO_PORT_SPARSE_PACK((1U << (n)) -1, (a), (f))
>
OK, i'll update it.

>
> > +/*
> > + * m indicates the bitmap of supported pins, a is the register index
> > + * and f is pin configuration capabilities supported.
> > + */
> > +#define RZG2L_GPIO_PORT_SPARSE_PACK(m, a, f) (((u64)(m) << 28) | ((a) << 20) | (f))
> > #define RZG2L_GPIO_PORT_GET_PINMAP(x) (((x) & GENMASK_ULL(35, 28)) >> 28)
> > #define RZG2L_GPIO_PORT_GET_PINCNT(x) (hweight8(RZG2L_GPIO_PORT_GET_PINMAP((x))))
> >
> > @@ -185,6 +192,18 @@ struct rzg2l_dedicated_configs {
> > u64 config;
> > };
> >
> > +/**
> > + * struct rzg2l_variable_pin_cfg - pin data cfg
> > + * @cfg: port pin configuration
> > + * @port: port number
> > + * @pin: port pin
> > + */
> > +struct rzg2l_variable_pin_cfg {
> > + u32 cfg;
> > + u8 port;
> > + u8 pin;
>
> As cfg only contains the lower bits (PIN_CFG_*), I think you can fit
> everything in a u32:
>
> u32 cfg: 20;
> u32 port: 5;
> u32 pin: 3;
>
Agreed.

> > +};
> > +
> > struct rzg2l_pinctrl_data {
> > const char * const *port_pins;
> > const u64 *port_pin_configs;
> > @@ -193,6 +212,8 @@ struct rzg2l_pinctrl_data {
> > unsigned int n_port_pins;
> > unsigned int n_dedicated_pins;
> > const struct rzg2l_hwcfg *hwcfg;
> > + const struct rzg2l_variable_pin_cfg *variable_pin_cfg;
> > + unsigned int n_variable_pin_cfg;
> > };
> >
> > /**
> > @@ -228,6 +249,158 @@ struct rzg2l_pinctrl {
> >
> > static const u16 available_ps[] = { 1800, 2500, 3300 };
> >
> > +#ifdef CONFIG_RISCV
> > +static u64 rzg2l_pinctrl_get_variable_pin_cfg(struct rzg2l_pinctrl *pctrl,
> > + u64 pincfg,
> > + unsigned int port,
> > + u8 pin)
> > +{
> > + unsigned int i;
> > + u8 pincount;
> > + u8 pinmap;
> > + u32 off;
> > +
> > + if (!pctrl->data->n_variable_pin_cfg)
> > + return pincfg;
>
> This cannot happen (but implies a driver table bug).
>
agreed, I will drop it.

> > +
> > + for (i = 0; i < pctrl->data->n_variable_pin_cfg; i++) {
> > + if (pctrl->data->variable_pin_cfg[i].port == port &&
> > + pctrl->data->variable_pin_cfg[i].pin == pin)
> > + break;
> > + }
> > + if (i == pctrl->data->n_variable_pin_cfg)
> > + return pincfg;
>
> My first thought was that this cannot happen either, but this function
> is called for non-existent pins on sparse ports?
>
I will drop that.

> > +
> > + pinmap = RZG2L_GPIO_PORT_GET_PINMAP(pincfg);
> > + pincount = RZG2L_GPIO_PORT_GET_PINCNT(pincfg);
> > + off = RZG2L_PIN_CFG_TO_PORT_OFFSET(pincfg);
> > +
> > + if (pinmap == pincount)
>
> Huh?
>
Oops.

> > + return RZG2L_GPIO_PORT_PACK(pincount, off, pctrl->data->variable_pin_cfg[i].cfg);
> > +
> > + return RZG2L_GPIO_PORT_SPARSE_PACK(pinmap, off, pctrl->data->variable_pin_cfg[i].cfg);
>
> Can't you just replace the lower bits of pincfg by
> pctrl->data->variable_pin_cfg[i].cfg?
>
> return (pincfg & ~PIN_CFG_...) | pctrl->data->variable_pin_cfg[i].cfg;
>
> And just move this single statement into if-condition in the for-loop
> above?
>
Agreed.

> > +}
> > +
> > +static const struct rzg2l_variable_pin_cfg r9a07g043f_variable_pin_cfg[] = {
> > + {
> > + .port = 20,
> > + .pin = 0,
> > + .cfg = PIN_CFG_IOLH_B | PIN_CFG_SR | PIN_CFG_PUPD |
> > + PIN_CFG_FILONOFF | PIN_CFG_FILNUM | PIN_CFG_FILCLKSEL |
> > + PIN_CFG_IEN | PIN_CFG_NOGPIO_INT,
>
> Why do all new pins have PIN_CFG_NOGPIO_INT set?
> P19_1, P20_0-2, P24_5, P25_1, P28_0-4 do have bits defined in an
> Interrupt Enable Control Register (ISEL)?
>
I have got clarification from the HW team that the ISEL bits will be
dropped for P19-P28 (and also we dont corresponding GPIOINTx for
P19-P28).

> > + },
>
> > @@ -1320,6 +1493,27 @@ static const u64 r9a07g043_gpio_configs[] = {
> > RZG2L_GPIO_PORT_PACK(2, 0x20, RZG2L_MPXED_PIN_FUNCS),
> > RZG2L_GPIO_PORT_PACK(4, 0x21, RZG2L_MPXED_PIN_FUNCS),
> > RZG2L_GPIO_PORT_PACK(6, 0x22, RZG2L_MPXED_PIN_FUNCS),
> > +#ifdef CONFIG_RISCV
> > + /* Below additional port pins (P19 - P28) are exclusively available on RZ/Five SoC only */
> > + RZG2L_GPIO_PORT_SPARSE_PACK(0x2, 0x06, PIN_CFG_IOLH_B | PIN_CFG_SR | PIN_CFG_PUPD |
> > + PIN_CFG_FILONOFF | PIN_CFG_FILNUM | PIN_CFG_FILCLKSEL |
> > + PIN_CFG_IEN | PIN_CFG_NOGPIO_INT), /* P19 */
> > + RZG2L_GPIO_PORT_PACK(8, 0x07, PIN_CFG_VARIABLE), /* P20 */
> > + RZG2L_GPIO_PORT_SPARSE_PACK(0x2, 0x08, PIN_CFG_IOLH_B | PIN_CFG_SR | PIN_CFG_PUPD |
> > + PIN_CFG_IEN | PIN_CFG_NOGPIO_INT), /* P21 */
> > + RZG2L_GPIO_PORT_PACK(4, 0x09, PIN_CFG_IOLH_B | PIN_CFG_SR | PIN_CFG_PUPD |
> > + PIN_CFG_IEN | PIN_CFG_NOGPIO_INT), /* P22 */
> > + RZG2L_GPIO_PORT_SPARSE_PACK(0x3e, 0x0a, PIN_CFG_VARIABLE), /* P23 */
> > + RZG2L_GPIO_PORT_PACK(6, 0x0b, PIN_CFG_VARIABLE), /* P24 */
> > + RZG2L_GPIO_PORT_SPARSE_PACK(0x2, 0x0c, PIN_CFG_IOLH_B | PIN_CFG_SR | PIN_CFG_FILONOFF |
> > + PIN_CFG_FILNUM | PIN_CFG_FILCLKSEL |
> > + PIN_CFG_NOGPIO_INT), /* P25 */
> > + 0x0, /* Dummy port P26 */
> > + 0x0, /* Dummy port P27 */
> > + RZG2L_GPIO_PORT_PACK(6, 0x0f, PIN_CFG_IOLH_A | PIN_CFG_SR | PIN_CFG_PUPD |
> > + PIN_CFG_FILONOFF | PIN_CFG_FILNUM | PIN_CFG_FILCLKSEL |
> > + PIN_CFG_NOGPIO_INT), /* P28 */
>
> The P28 config can be simplified to "RZG2L_MPXED_PIN_FUNCS |
> PIN_CFG_NOGPIO_INT".
>
Agreed.

> > +#endif
> > };
> >
> > static const u64 r9a08g045_gpio_configs[] = {
> > @@ -1478,12 +1672,18 @@ static const struct rzg2l_dedicated_configs rzg3s_dedicated_pins[] = {
> > PIN_CFG_IO_VMC_SD1)) },
> > };
> >
> > -static int rzg2l_gpio_get_gpioint(unsigned int virq, const struct rzg2l_pinctrl_data *data)
> > +static int rzg2l_gpio_get_gpioint(unsigned int virq, struct rzg2l_pinctrl *pctrl)
> > {
> > + const struct pinctrl_pin_desc *pin_desc = &pctrl->desc.pins[virq];
> > + const struct rzg2l_pinctrl_data *data = pctrl->data;
> > + u64 *pin_data = pin_desc->drv_data;
> > unsigned int gpioint;
> > unsigned int i;
> > u32 port, bit;
> >
> > + if (*pin_data & PIN_CFG_NOGPIO_INT)
> > + return -EINVAL;
> > +
> > port = virq / 8;
> > bit = virq % 8;
>
> Out-of-context, you have:
>
> gpioint = bit;
> for (i = 0; i < port; i++)
> gpioint +=
> RZG2L_GPIO_PORT_GET_PINCNT(data->port_pin_configs[i]);
>
> return gpioint;
>
> Shouldn't the for-loop skip pins with PIN_CFG_NOGPIO_INT set?
>
As of now P19-P28 cannot be used as interrupt pins and there is no SoC
where we can test this case (as this has to match GPIOINTx for a given
port pin). So we can ignore it for now.

> Gr{oetje,eeting}s,
>
> Geert
>
> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx
>
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
> -- Linus Torvalds

Next message: Steven Rostedt: "Re: [PATCH] tracefs/eventfs: Use root and instance inodes as default ownership"
Previous message: William Butler: "[PATCH] nvme: Inform controller of doorbell config before unquiescing IO queues"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]