| Message ID | 1485367787-8109-2-git-send-email-jacopo+renesas@jmondi.org |
|---|---|
| State | New |
| Headers | show |
Hi Jacopo, On Wednesday, January 25, 2017, Jacopo Mondi wrote: > drivers/pinctrl/Kconfig | 1 + > drivers/pinctrl/Makefile | 1 + > drivers/pinctrl/rz-pfc/Kconfig | 18 ++ > drivers/pinctrl/rz-pfc/Makefile | 1 + > drivers/pinctrl/rz-pfc/pinctrl-rz.c | 447 > ++++++++++++++++++++++++++++++++++++ > drivers/pinctrl/rz-pfc/pinctrl-rz.h | 114 +++++++++ > 6 files changed, 582 insertions(+) > create mode 100644 drivers/pinctrl/rz-pfc/Kconfig create mode 100644 > drivers/pinctrl/rz-pfc/Makefile create mode 100644 drivers/pinctrl/rz- > pfc/pinctrl-rz.c > create mode 100644 drivers/pinctrl/rz-pfc/pinctrl-rz.h I think we should try to avoid the rz naming as much as possible since this driver will hopefully be useful for other future Renesas devices if they move to a similar pin-control type method. Maybe future "R-car" SoCs? Or, maybe Renesas Marketing decides to come up with a new name for SoCs. Otherwise, you could end up with a directly like "mach-shmobile" filled with a bunch of...well...not SH-Mobile parts. Maybe just drivers/pinctrl/renesas/pinctrl-renesas.c Chris -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Chris, On 26/01/2017 03:58, Chris Brandt wrote: > Hi Jacopo, > > On Wednesday, January 25, 2017, Jacopo Mondi wrote: >> drivers/pinctrl/Kconfig | 1 + >> drivers/pinctrl/Makefile | 1 + >> drivers/pinctrl/rz-pfc/Kconfig | 18 ++ >> drivers/pinctrl/rz-pfc/Makefile | 1 + >> drivers/pinctrl/rz-pfc/pinctrl-rz.c | 447 >> ++++++++++++++++++++++++++++++++++++ >> drivers/pinctrl/rz-pfc/pinctrl-rz.h | 114 +++++++++ >> 6 files changed, 582 insertions(+) >> create mode 100644 drivers/pinctrl/rz-pfc/Kconfig create mode 100644 >> drivers/pinctrl/rz-pfc/Makefile create mode 100644 drivers/pinctrl/rz- >> pfc/pinctrl-rz.c >> create mode 100644 drivers/pinctrl/rz-pfc/pinctrl-rz.h > > > > I think we should try to avoid the rz naming as much as possible since > this driver will hopefully be useful for other future Renesas devices if > they move to a similar pin-control type method. Maybe future "R-car" SoCs? > Or, maybe Renesas Marketing decides to come up with a new name for SoCs. > > Otherwise, you could end up with a directly like "mach-shmobile" filled > with a bunch of...well...not SH-Mobile parts. > > Maybe just > drivers/pinctrl/renesas/pinctrl-renesas.c > Wouldn't this be confusing, as most of renesas SoC are supported through drivers/pinctrl/sh-pfc instead? I agree on dropping the rz name, if more SoC will come and join the pin-based PFC realm... Thanks j > > > Chris > > -- > To unsubscribe from this list: send the line "unsubscribe linux-gpio" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Jacopo, On Thursday, January 26, 2017, jacopo mondi wrote: > > I think we should try to avoid the rz naming as much as possible since > > this driver will hopefully be useful for other future Renesas devices > > if they move to a similar pin-control type method. Maybe future "R-car" > SoCs? > > Or, maybe Renesas Marketing decides to come up with a new name for SoCs. > > > > Otherwise, you could end up with a directly like "mach-shmobile" > > filled with a bunch of...well...not SH-Mobile parts. > > > > Maybe just > > drivers/pinctrl/renesas/pinctrl-renesas.c > > > > Wouldn't this be confusing, as most of renesas SoC are supported through > drivers/pinctrl/sh-pfc instead? It's already confusing that a bunch of ARM parts use a "sh" pin controller. In reality, any Renesas device could be a peripheral mix of ex-Mitsubishi, ex-Hitachi, ex-NEC or Renesas original. Or, some IP was licensed from a 3rd party (ie, the SDHI/TMIO controller). The only thing consistent is "Renesas". > I agree on dropping the rz name, if more SoC will come and join the pin- > based PFC realm... I will suggest it for the next generation of SoCs, so we'll see what happens... Thanks, Chris -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Jacopo, On Wed, Jan 25, 2017 at 7:09 PM, Jacopo Mondi <jacopo+renesas@jmondi.org> wrote: > Add core module for per-pin Renesas RZ series pin controller. > The core module allows SoC driver to register their pins and SoC > specific operations and interfaces with pinctrl and pinmux core on their > behalf. > > Signed-off-by: Jacopo Mondi <jacopo+renesas@jmondi.org> I believe nothing besides the DT property "renesas-rz,pins" and the value of the RZ_PIN_ARGS_COUNT macro is really specific to Renesas or RZ(A)? So this could become something really generic, and part of core pinctrl? > --- /dev/null > +++ b/drivers/pinctrl/rz-pfc/pinctrl-rz.c > @@ -0,0 +1,447 @@ > +#define RZ_PIN_ARGS_COUNT 3 This should be a parameter in the SoC-specific subdriver. > +/** > + * rz_pinctrl_pos_to_index() - Retrieve the index of pin at position [bank:pin] > + * > + * This can be improved, as it walks all the pins reported by the SoC driver > + * > + * @return: pin number between [0 - npins]; -1 if not found > + */ > +static int rz_pinctrl_pos_to_index(struct rz_pinctrl_dev *rz_pinctrl, > + unsigned int bank, unsigned int pin) > +{ > + struct rz_pinctrl_info *info; > + struct rz_pin_desc *rz_pin; const > + int i, npins; unsigned int > +/* ---------------------------------------------------------------------------- > + * pinctrl operations > + */ > +static void rz_pin_dbg_show(struct pinctrl_dev *pctldev, struct seq_file *s, const ... pctldev > + unsigned int pin) > +{ > + struct rz_pinctrl_dev *rz_pinctrl; > + struct rz_pinctrl_info *info; const > +static int rz_dt_node_to_map(struct pinctrl_dev *pctldev, > + struct device_node *np_config, > + struct pinctrl_map **map, unsigned int *num_maps) > +{ > + fngrps = devm_kzalloc(rz_pinctrl->dev, sizeof(*fngrps), GFP_KERNEL); > + if (unlikely(!fngrps)) { Please don't use unlikely. Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org 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 -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Jacopo, On Wednesday, January 25, 2017, Jacopo Mondi wrote: > + > + return 0; > + > +free_map: > + devm_kfree(rz_pinctrl->dev, *map); > +free_fngrps: > + devm_kfree(rz_pinctrl->dev, fngrps); > +free_pins: > + devm_kfree(rz_pinctrl->dev, mux_modes); > + devm_kfree(rz_pinctrl->dev, grpins); > + return ret; > +} Since one of the benefits of using devm_kzalloc is that if the probe fails and returns an error, all the memory associated with that device will automatically get freed, you 'might' not need this code to free memory. I say might because I'm not sure if returning an error here will kill the driver or not. But, might be interesting to look into. > +#define RZ_PIN_NAME(bank, pin) \ > + PIN_##bank##_##pin > + > +#define RZ_PIN_DESC(b, p) \ > + { .number = RZ_PIN_NAME(b, p), \ > + .name = __stringify(RZ_PIN_NAME(b, p)), \ The hardware manual uses the names "P1_0" for ports, so it might be better to match that format for consistency. Chris -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Chris, On 30/01/2017 20:19, Chris Brandt wrote: > Hi Jacopo, > > On Wednesday, January 25, 2017, Jacopo Mondi wrote: >> + >> + return 0; >> + >> +free_map: >> + devm_kfree(rz_pinctrl->dev, *map); >> +free_fngrps: >> + devm_kfree(rz_pinctrl->dev, fngrps); >> +free_pins: >> + devm_kfree(rz_pinctrl->dev, mux_modes); >> + devm_kfree(rz_pinctrl->dev, grpins); >> + return ret; >> +} > > Since one of the benefits of using devm_kzalloc is that if > the probe fails and returns an error, all the memory associated with > that device will automatically get freed, you 'might' not > need this code to free memory. > > I say might because I'm not sure if returning an error here will > kill the driver or not. But, might be interesting to look into. > No, returning an error here would not kill the driver BUT if dt_node_to_map fails, dt_free_map is called immediately later [1] So here we maybe should not use device managed memory as it does not bring anything, do not free *map as it is freed later in dt_free_map, but release fngrps mux_modes and grpins, as we lose reference to them outside the scope of this function. Do you agree? > >> +#define RZ_PIN_NAME(bank, pin) \ >> + PIN_##bank##_##pin >> + >> +#define RZ_PIN_DESC(b, p) \ >> + { .number = RZ_PIN_NAME(b, p), \ >> + .name = __stringify(RZ_PIN_NAME(b, p)), \ > > The hardware manual uses the names "P1_0" for ports, so it might > be better to match that format for consistency. > > Noted Thanks j [1] http://lxr.free-electrons.com/source/drivers/pinctrl/devicetree.c#L236 -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Jacopo, On Tuesday, January 31, 2017, Jacopo Mondi wrote: > > Since one of the benefits of using devm_kzalloc is that if the probe > > fails and returns an error, all the memory associated with that device > > will automatically get freed, you 'might' not need this code to free > > memory. > > > > I say might because I'm not sure if returning an error here will kill > > the driver or not. But, might be interesting to look into. > > > > No, returning an error here would not kill the driver BUT if > dt_node_to_map fails, dt_free_map is called immediately later [1] > > So here we maybe should not use device managed memory as it does not bring > anything, do not free *map as it is freed later in dt_free_map, but > release fngrps mux_modes and grpins, as we lose reference to them outside > the scope of this function. > > Do you agree? Like you said, there is no benefit one way of the other. But, doing a grep of the pinctrl directory, devm_kzalloc is used more by the other drivers than kzalloc, so maybe we just stick with devm_kzalloc (baah goes the sheep) Chris > -----Original Message----- > From: jacopo mondi [mailto:jacopo@jmondi.org] > Sent: Tuesday, January 31, 2017 4:01 AM > To: Chris Brandt <Chris.Brandt@renesas.com>; Jacopo Mondi > <jacopo+renesas@jmondi.org>; laurent.pinchart@ideasonboard.com; > geert+renesas@glider.be; linus.walleij@linaro.org > Cc: linux-renesas-soc@vger.kernel.org; linux-gpio@vger.kernel.org > Subject: Re: [RFC 1/5] pinctrl: rz-pfc: Add Renesas RZ pinctrl core module > > Hi Chris, > > On 30/01/2017 20:19, Chris Brandt wrote: > > Hi Jacopo, > > > > On Wednesday, January 25, 2017, Jacopo Mondi wrote: > >> + > >> + return 0; > >> + > >> +free_map: > >> + devm_kfree(rz_pinctrl->dev, *map); > >> +free_fngrps: > >> + devm_kfree(rz_pinctrl->dev, fngrps); > >> +free_pins: > >> + devm_kfree(rz_pinctrl->dev, mux_modes); > >> + devm_kfree(rz_pinctrl->dev, grpins); > >> + return ret; > >> +} > > > > Since one of the benefits of using devm_kzalloc is that if the probe > > fails and returns an error, all the memory associated with that device > > will automatically get freed, you 'might' not need this code to free > > memory. > > > > I say might because I'm not sure if returning an error here will kill > > the driver or not. But, might be interesting to look into. > > > > No, returning an error here would not kill the driver BUT if > dt_node_to_map fails, dt_free_map is called immediately later [1] > > So here we maybe should not use device managed memory as it does not bring > anything, do not free *map as it is freed later in dt_free_map, but > release fngrps mux_modes and grpins, as we lose reference to them outside > the scope of this function. > > Do you agree? > > > > > >> +#define RZ_PIN_NAME(bank, pin) \ > >> + PIN_##bank##_##pin > >> + > >> +#define RZ_PIN_DESC(b, p) \ > >> + { .number = RZ_PIN_NAME(b, p), \ > >> + .name = __stringify(RZ_PIN_NAME(b, p)), \ > > > > The hardware manual uses the names "P1_0" for ports, so it might be > > better to match that format for consistency. > > > > > > Noted > > Thanks > j > > > [1] http://lxr.free-electrons.com/source/drivers/pinctrl/devicetree.c#L236 > -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Chris, On Tuesday 31 Jan 2017 13:48:57 Chris Brandt wrote: > On Tuesday, January 31, 2017, Jacopo Mondi wrote: > >> Since one of the benefits of using devm_kzalloc is that if the probe > >> fails and returns an error, all the memory associated with that device > >> will automatically get freed, you 'might' not need this code to free > >> memory. > >> > >> I say might because I'm not sure if returning an error here will kill > >> the driver or not. But, might be interesting to look into. > > > > No, returning an error here would not kill the driver BUT if > > dt_node_to_map fails, dt_free_map is called immediately later [1] > > > > So here we maybe should not use device managed memory as it does not bring > > anything, do not free *map as it is freed later in dt_free_map, but > > release fngrps mux_modes and grpins, as we lose reference to them outside > > the scope of this function. > > > > Do you agree? > > Like you said, there is no benefit one way of the other. > > But, doing a grep of the pinctrl directory, devm_kzalloc is used more by > the other drivers than kzalloc, so maybe we just stick with devm_kzalloc > > (baah goes the sheep) Given that the devm_*() functions incur an overhead, let's use kzalloc/kfree directly.
Hi Jacopo, Thank you for the patch. On Wednesday 25 Jan 2017 19:09:43 Jacopo Mondi wrote: > Add core module for per-pin Renesas RZ series pin controller. > The core module allows SoC driver to register their pins and SoC > specific operations and interfaces with pinctrl and pinmux core on their > behalf. > > Signed-off-by: Jacopo Mondi <jacopo+renesas@jmondi.org> > --- > drivers/pinctrl/Kconfig | 1 + > drivers/pinctrl/Makefile | 1 + > drivers/pinctrl/rz-pfc/Kconfig | 18 ++ > drivers/pinctrl/rz-pfc/Makefile | 1 + > drivers/pinctrl/rz-pfc/pinctrl-rz.c | 447 ++++++++++++++++++++++++++++++++ > drivers/pinctrl/rz-pfc/pinctrl-rz.h | 114 +++++++++ > 6 files changed, 582 insertions(+) > create mode 100644 drivers/pinctrl/rz-pfc/Kconfig > create mode 100644 drivers/pinctrl/rz-pfc/Makefile > create mode 100644 drivers/pinctrl/rz-pfc/pinctrl-rz.c > create mode 100644 drivers/pinctrl/rz-pfc/pinctrl-rz.h As Chris pointed out, s/rz-pfc/renesas/ would be a good idea. This code isn't specific to RZ chips (even if it's only used by them at the moment), so the rz prefix and suffix should probably replaced with something else (in file names and in the code too). [snip] > diff --git a/drivers/pinctrl/rz-pfc/Kconfig b/drivers/pinctrl/rz-pfc/Kconfig > new file mode 100644 > index 0000000..3714c10 > --- /dev/null > +++ b/drivers/pinctrl/rz-pfc/Kconfig > @@ -0,0 +1,18 @@ > +# > +# Renesas RZ pinctrl drivers > +# > + > +if ARCH_RENESAS > + > +config PINCTRL_RZ_PINCTRL You can add a depends on ARCH_RENESAS to replace the above if. It should actually be depends on ARCH_RENESAS || COMPILE_TEST to extend compile-test coverage. An even better option could be to drop the depends line completely, and replace def_bool with bool. That way the option will only be enabled if explicitly selected by another option (as done in patch 2/5). > + select PINMUX > + select PINCONF > + select GPIOLIB > + select GENERIC_PINCONF > + select GENERIC_PINMUX_FUNCTIONS > + select GENERIC_PINCTRL_GROUPS > + def_bool y Won't this enable the driver by default on all platforms, while only RZ needs it ? It's also customary to have the bool/tristate line as the very first line in the config option. > + help > + This enables pin control drivers for Renesas RZ platforms > + > +endif [snip] > diff --git a/drivers/pinctrl/rz-pfc/pinctrl-rz.c > b/drivers/pinctrl/rz-pfc/pinctrl-rz.c new file mode 100644 > index 0000000..3efbf03 > --- /dev/null > +++ b/drivers/pinctrl/rz-pfc/pinctrl-rz.c > @@ -0,0 +1,447 @@ > +/* > + * Pinctrl support for Renesas RZ Series > + * > + * Copyright (C) 2017 Jacopo Mondi > + * Copyright (C) 2017 Renesas Electronics Corporation > + * > + * This file is licensed under the terms of the GNU General Public > + * License version 2. This program is licensed "as is" without any > + * warranty of any kind, whether express or implied. > + */ > + > +#include <linux/err.h> > +#include <linux/init.h> > +#include <linux/ioport.h> > +#include <linux/module.h> > + No need for a blank line here. > +#include <linux/of.h> > +#include <linux/of_address.h> > +#include <linux/of_device.h> > + Or here. > +#include <linux/pinctrl/pinconf-generic.h> > +#include <linux/pinctrl/pinctrl.h> > +#include <linux/pinctrl/pinmux.h> > + > +#include "../core.h" > +#include "../devicetree.h" > +#include "../pinmux.h" > + > +#include "pinctrl-rz.h" > + > +#define DRIVER_NAME "rz-pinctrl" > +#define RZ_PIN_ARGS_COUNT 3 > + > +/** > + * rz_pinctrl_pos_to_index() - Retrieve the index of pin at position > [bank:pin] > + * > + * This can be improved, as it walks all the pins reported by the SoC > driver > + * > + * @return: pin number between [0 - npins]; -1 if not found > + */ > +static int rz_pinctrl_pos_to_index(struct rz_pinctrl_dev *rz_pinctrl, > + unsigned int bank, unsigned int pin) > +{ > + struct rz_pinctrl_info *info; > + struct rz_pin_desc *rz_pin; > + int i, npins; i and npins are never negative, you can make them unsigned int. One variable declaration per line is also preferred. > + > + info = rz_pinctrl->info; You can move this to the variable declaration line. > + npins = info->npins; > + for (i = 0; i < npins; ++i) { You can drop the npins variable and use info->npins directly. > + rz_pin = &info->pins[i]; > + > + /* > + * return the pin index in the array, not the pin number, > + * so that we can access it easily when muxing group's pins Please start sentences with a capital letter and end them with a period. > + */ > + if (rz_pin->bank == bank && rz_pin->pin == pin) > + return i; > + } > + > + return -1; How about -EINVAL ? -1 is -EPERM, which would not be an appropriate error code if you end up leaking it to upper layers. > +} > + > +/* ------------------------------------------------------------------------ > + * pinctrl operations > + */ > +static void rz_pin_dbg_show(struct pinctrl_dev *pctldev, struct seq_file > *s, > + unsigned int pin) > +{ > + struct rz_pinctrl_dev *rz_pinctrl; > + struct rz_pinctrl_info *info; > + > + rz_pinctrl = pinctrl_dev_get_drvdata(pctldev); > + info = rz_pinctrl->info; You can also assign at declaration time here. > + seq_printf(s, "%s %s\n", info->pins[pin].name, DRIVER_NAME); The pinctrl core already prints the pin name: seq_printf(s, "pin %d (%s) ", pin, desc->name); /* Driver-specific info per pin */ if (ops->pin_dbg_show) ops->pin_dbg_show(pctldev, s, pin); You can just print the driver name here. I'm actually wondering whether printing the driver name is useful, or if we could simply drop this function. > +} > + > +/** > + * rz_dt_node_to_map() - Parse device tree nodes and collect pins, groups > and > + * functions I don't think we have groups and functions, do we ? > + * > + * Pins for RZ series pin controller described by "renesas-rz,pins" > property > + * are arrays of u32 values in the form: As mentioned elsewhere, "renesas,pins" will do. We could even use "pins", but then we'd have to comply with the pinctrl generic bindings (with a bin number in the first call instead of separate bank and pin for instance). > + * > + * "renesas-rz,pins" = <BANK PIN MUX>, ... ; > + * > + * Parse the list arguments and identify pins through their position > + * (bank and pin offset) and save the provided mux mode for later use. > + * > + * TODO: the array can be expended to support additional parameters for > + * pin configurations values (IO direction etc) > + * > + * @pctldev: pin controller device > + * @np_config: device tree node to parse > + * @map: pointer to pin map (output) > + * @num_maps: number of collected maps (output) > + * > + * @return: 0 for success; != 0 otherwise "; a negative error code otherwise" ? The function never returns a positive value. I don't think kerneldoc uses @return. Please try to compile the documentation and fix errors and warnings. > + */ > +static int rz_dt_node_to_map(struct pinctrl_dev *pctldev, > + struct device_node *np_config, > + struct pinctrl_map **map, unsigned int *num_maps) > +{ > + struct rz_pinctrl_dev *rz_pinctrl; > + const char *prop_name = "renesas-rz,pins", > + *grpname, **fngrps; > + unsigned int *grpins, > + *mux_modes; One variable declaration per line please. > + int i, npins, ret; i is never negative, you can make it an unsigned int. > + > + rz_pinctrl = pinctrl_dev_get_drvdata(pctldev); You can initialize the variable at declaration time. > + npins = pinctrl_count_index_with_args(np_config, prop_name); > + if (npins <= 0) { > + dev_err(rz_pinctrl->dev, "Missing pins configuration prop\n"); Maybe (... "Missing %s property\n", prop_name); to be more explicit ? > + return -EINVAL; > + } > + > + mux_modes = devm_kzalloc(rz_pinctrl->dev, sizeof(*mux_modes) * npins, > + GFP_KERNEL); > + grpins = devm_kzalloc(rz_pinctrl->dev, sizeof(*grpins) * npins, > + GFP_KERNEL); As explained in another e-mail, you can use kzalloc() here instead of devm_kzalloc(). Or, even better, kcalloc(). > + if (unlikely(!grpins || !mux_modes)) > + return -ENOMEM; > + > + /* > + * functions are made of 1 group only; > + * in facts, functions and groups are identical for RZ pin controller, > + * except that functions carry an array of mux modes (aka alternate > + * functions IDs) > + */ > + fngrps = devm_kzalloc(rz_pinctrl->dev, sizeof(*fngrps), GFP_KERNEL); > + if (unlikely(!fngrps)) { > + ret = -ENOMEM; > + goto free_pins; > + } > + > + *num_maps = 0; > + (*map) = devm_kzalloc(rz_pinctrl->dev, sizeof(**map), GFP_KERNEL); No need for parentheses around *map. > + if (unlikely(!*map)) { > + ret = -ENOMEM; > + goto free_fngrps; > + } > + > + /* collect pin numbers and mux confs to create groups and functions */ > + for (i = 0; i < npins; ++i) { > + struct of_phandle_args of_pins_args; > + unsigned int bank, offs, mode, pin_idx; You can spell offset fully. > + > + /* DTS identifies pin by position: bank and pin offset */ > + ret = pinctrl_parse_index_with_args(np_config, prop_name, i, > + &of_pins_args); > + if (ret) > + goto free_map; > + > + if (of_pins_args.args_count < RZ_PIN_ARGS_COUNT) { > + dev_err(rz_pinctrl->dev, > + "Wrong arguments number for renesas-rz,pins"); > + ret = -EINVAL; > + goto free_map; > + } > + > + bank = of_pins_args.args[0]; > + offs = of_pins_args.args[1]; > + mode = of_pins_args.args[2]; > + > + /* pin_idx is the index on the static pin array */ > + pin_idx = rz_pinctrl_pos_to_index(rz_pinctrl, bank, offs); > + if (pin_idx < 0) { > + dev_err(rz_pinctrl->dev, > + "Invalid pin position %d-%d\n", bank, offs); bank and offset are unsigned, you should use %u. > + ret = -EINVAL; > + goto free_map; > + } > + > + grpins[i] = pin_idx; > + mux_modes[i] = mode; > + } > + > + grpname = np_config->name; > + fngrps[0] = grpname; > + > + mutex_lock(&rz_pinctrl->mutex); This seems weird. The pinctrl generic functions indeed state that the caller must take care of locking, but there doesn't seem to be any lock protecting races between .dt_node_to_map() calling the generic functions below, and the generic .get_group_*() handlers. > + ret = pinctrl_generic_add_group(pctldev, grpname, grpins, npins, NULL); > + if (ret) { > + mutex_unlock(&rz_pinctrl->mutex); > + goto free_map; > + } > + > + ret = pinmux_generic_add_function(pctldev, grpname, fngrps, > + 1, mux_modes); > + if (ret) { > + mutex_unlock(&rz_pinctrl->mutex); > + goto free_map; > + } > + mutex_unlock(&rz_pinctrl->mutex); > + > + (*map)->type = PIN_MAP_TYPE_MUX_GROUP; > + (*map)->data.mux.group = grpname; > + (*map)->data.mux.function = grpname; > + *num_maps = 1; > + > + return 0; > + > +free_map: > + devm_kfree(rz_pinctrl->dev, *map); > +free_fngrps: > + devm_kfree(rz_pinctrl->dev, fngrps); > +free_pins: > + devm_kfree(rz_pinctrl->dev, mux_modes); > + devm_kfree(rz_pinctrl->dev, grpins); > + return ret; > +} > + > +/** > + * rz_dt_free_map() > + */ > +static void rz_dt_free_map(struct pinctrl_dev *pctldev, > + struct pinctrl_map *map, unsigned int num_maps) > +{ > + struct rz_pinctrl_dev *rz_pinctrl = pinctrl_dev_get_drvdata(pctldev); > + > + devm_kfree(rz_pinctrl->dev, map); Shouldn't you also free the other allocated pieces of memory (fngrps, mux_modes and grpins) ? > +} > + > +static const struct pinctrl_ops rz_pinctrl_ops = { > + .get_groups_count = pinctrl_generic_get_group_count, > + .get_group_name = pinctrl_generic_get_group_name, > + .get_group_pins = pinctrl_generic_get_group_pins, > + .pin_dbg_show = rz_pin_dbg_show, > + .dt_node_to_map = rz_dt_node_to_map, > + .dt_free_map = rz_dt_free_map, > +}; > + > +/* ------------------------------------------------------------------------ > + * pinconfig operations > + */ > + > +/* TODO */ Yes ? :-) > +/* ------------------------------------------------------------------------ > + * pinmux operations > + */ > + > +/** > + * rz_pinmux_set() - Retrieve pins from a group and apply mux settings > + * > + * @pctldev: pin controller device > + * @selector: Function selector > + * @group: Group selector > + * @return: 0 for success; != otherwise Same comment as above. > + */ > +static int rz_pinmux_set(struct pinctrl_dev *pctldev, unsigned int > selector, > + unsigned int group) > +{ > + struct rz_pinctrl_dev *rz_pinctrl; > + struct rz_pinctrl_info *info; > + struct rz_pinctrl_ops *ops; > + struct group_desc *grp; > + struct function_desc *func; > + unsigned int npins, *mux_modes; > + int i; i is never negative. > + > + rz_pinctrl = pinctrl_dev_get_drvdata(pctldev); > + info = rz_pinctrl->info; > + ops = info->ops; You can do this at variable declaration time. > + grp = pinctrl_generic_get_group(pctldev, group); > + if (!grp) > + return -EINVAL; > + > + func = pinmux_generic_get_function(pctldev, selector); > + if (!func) > + return -EINVAL; > + > + npins = grp->num_pins; No need for an intermediate variable, use grp->num_pins directly. > + mux_modes = (unsigned int *)func->data; > + > + for (i = 0; i < npins; ++i) { > + unsigned int pin_idx, mux_mode; > + struct rz_pin_desc *pin; > + int ret; > + > + /* > + * use pin index to retrieve pin descriptor; > + * then provide it to the set_mux SoC operation > + */ > + pin_idx = grp->pins[i]; > + pin = &info->pins[pin_idx]; > + mux_mode = mux_modes[i]; > + > + if (!ops || !ops->set_mux) { > + dev_err(rz_pinctrl->dev, "Pin muxing not supported\n"); > + return -EINVAL; > + } > + > + ret = ops->set_mux(rz_pinctrl, pin, mux_mode); > + if (ret) > + return ret; > + } > + > + return 0; > +} > + > +struct pinmux_ops rz_pinmux_ops = { > + .get_functions_count = pinmux_generic_get_function_count, > + .get_function_name = pinmux_generic_get_function_name, > + .get_function_groups = pinmux_generic_get_function_groups, > + .set_mux = rz_pinmux_set, > +}; > + > +/* ------------------------------------------------------------------------ > + * rz pinctrl operations > + */ > + > +/** > + * rz_add_pins() - Enumerate pins reported by SoC driver in pin desc array Does this function really enumerate pins ? > + * > + * @rz_pinctrl: RZ pincontroller > + * > + * @return: 0 for success; != 0 otherwise Same as above, and some comment for below. > + */ > +static int rz_add_pins(struct rz_pinctrl_dev *rz_pinctrl) > +{ > + struct rz_pinctrl_info *info; > + struct rz_pin_desc *pin; > + unsigned int npins, i; > + > + info = rz_pinctrl->info; > + npins = info->npins; Fold this with variable declaration. I'd get rid of the npins variable, you can use info->npins. > + rz_pinctrl->pins = devm_kzalloc(rz_pinctrl->dev, > + npins * sizeof(*rz_pinctrl->pins), > + GFP_KERNEL); kcalloc. > + if (unlikely(!rz_pinctrl->pins)) > + return -ENOMEM; > + > + rz_pinctrl->desc.pins = rz_pinctrl->pins; > + rz_pinctrl->desc.npins = npins; > + > + for (i = 0; i < npins; ++i) { > + pin = &info->pins[i]; > + > + rz_pinctrl->pins[i].number = pin->number; > + rz_pinctrl->pins[i].name = pin->name; > + } > + > + return 0; > +} > + > +/** > + * rz_pinctrl_map_res() - Map memory resources for pincontroller You can call the function ..._map_resources. > + * @pdev: platform device > + * @rz_pincrl: RZ pincontroller > + * > + * @return: 0 for success; != 0 otherwise > + */ > +static int rz_pinctrl_map_res(struct platform_device *pdev, > + struct rz_pinctrl_dev *rz_pinctrl) > +{ > + struct resource *res; > + struct rz_pinctrl_res *rz_res; > + unsigned int i, nres; > + > + nres = pdev->num_resources; > + rz_res = devm_kzalloc(&pdev->dev, nres * sizeof(*rz_res), > + GFP_KERNEL); kcalloc() here too. > + if (unlikely(!rz_res)) > + return -ENOMEM; > + > + rz_pinctrl->nres = nres; > + rz_pinctrl->res = rz_res; > + > + for (i = 0; i < nres; i++) { > + res = platform_get_resource(pdev, IORESOURCE_MEM, i); > + if (!res) > + return -ENODEV; > + > + rz_res = &rz_pinctrl->res[i]; > + > + rz_res->start = res->start; > + rz_res->size = resource_size(res); If you remove the debugging statement from patch 2/5 you can drop the start and size fields. > + rz_res->base = devm_ioremap_resource(&pdev->dev, res); > + if (IS_ERR(rz_res->base)) > + return PTR_ERR(rz_res->base); > + } > + > + return 0; > +} > + > +/** > + * rz_pinctrl_probe() - Register pincontroller driver operations > + * > + * @pdev: platform device > + * @info: SoC device info structure > + */ > +int rz_pinctrl_probe(struct platform_device *pdev, struct rz_pinctrl_info > *info) > +{ > + int ret; > + struct rz_pinctrl_dev *rz_pinctrl; > + > + rz_pinctrl = devm_kzalloc(&pdev->dev, sizeof(*rz_pinctrl), GFP_KERNEL); kzalloc() > + if (!rz_pinctrl) > + return -ENOMEM; > + > + rz_pinctrl->dev = &pdev->dev; > + rz_pinctrl->info = info; > + > + ret = rz_pinctrl_map_res(pdev, rz_pinctrl); > + if (ret) > + return ret; > + > + platform_set_drvdata(pdev, rz_pinctrl); > + > + rz_pinctrl->desc.name = DRIVER_NAME; > + rz_pinctrl->desc.pctlops = &rz_pinctrl_ops; > + rz_pinctrl->desc.pmxops = &rz_pinmux_ops; > + rz_pinctrl->desc.owner = THIS_MODULE; > + > + ret = rz_add_pins(rz_pinctrl); > + if (ret) > + return ret; > + > + mutex_init(&rz_pinctrl->mutex); I'd move this earlier, to make sure the mutex is initialized before it gets used in case we start using it in rz_pinctrl_map_res() or rz_add_pins() later. > + ret = pinctrl_register_and_init(&rz_pinctrl->desc, rz_pinctrl->dev, > + rz_pinctrl, &rz_pinctrl->pctl); > + if (ret) { > + dev_err(&pdev->dev, "could not register rz pinctrl driver\n"); > + return ret; > + } > + > + return 0; > +} > + > +void rz_pinctrl_remove(struct platform_device *pdev) > +{ > + struct rz_pinctrl_dev *rz_pinctrl = platform_get_drvdata(pdev); > + > + pinctrl_unregister(rz_pinctrl->pctl); > +} The code is always compiled in, and the device never goes away. I'd drop remove support, as it's completely useless. You then need to set the device_driver structure suppress_bind_attrs field. > +MODULE_AUTHOR("Jacopo Mondi <jacopo+renesas@jmondi.org"); > +MODULE_DESCRIPTION("Pinctrl driver for Reneas RZ series"); > +MODULE_LICENSE("GPL v2"); > diff --git a/drivers/pinctrl/rz-pfc/pinctrl-rz.h > b/drivers/pinctrl/rz-pfc/pinctrl-rz.h new file mode 100644 > index 0000000..b873c89 > --- /dev/null > +++ b/drivers/pinctrl/rz-pfc/pinctrl-rz.h > @@ -0,0 +1,114 @@ > +/* > + * Pinctrl support for Renesas RZ Series > + * > + * Copyright (C) 2017 Jacopo Mondi > + * Copyright (C) 2017 Renesas Electronics Corporation > + * > + * This file is licensed under the terms of the GNU General Public > + * License version 2. This program is licensed "as is" without any > + * warranty of any kind, whether express or implied. > + */ > +#ifndef __RZ_PINCTRL_H__ > +#define __RZ_PINCTRL_H__ > + > +#include <linux/pinctrl/pinctrl.h> > +#include <linux/types.h> > + > +/* > --------------------------------------------------------------------------- > - + * pinctrl-rz data types > + */ > +#define RZ_PIN_NAME(bank, pin) \ > + PIN_##bank##_##pin > + > +#define RZ_PIN_DESC(b, p) \ > + { .number = RZ_PIN_NAME(b, p), \ > + .name = __stringify(RZ_PIN_NAME(b, p)), \ > + .bank = b, .pin = p } > + > +/** > + * rz_pin_desc - Single RZ pin descriptor > + * > + * @number: pin number as enumerated by SoC driver > + * @name: pin name as reported by SoC driver > + * @bank: pin bank location > + * @pin: pin register offset > + */ > +struct rz_pin_desc { > + unsigned int bank, pin; > + unsigned int number; > + const char *name; > +}; > + > +/** > + * rz_pinctrl_info - SoC info data > + * > + * @ops: SoC specific operations > + * @npins: number of pins > + * @pins: pin array > + */ > +struct rz_pinctrl_ops; > +struct rz_pinctrl_info { > + struct rz_pinctrl_ops *ops; > + > + unsigned int npins; > + struct rz_pin_desc *pins; > +}; > + > +/** > + * rz_pinctrl_res - Memory resource for RZ SoC > + * > + * @start: physical address base > + * @size: memory region size > + * @base: logical address base > + */ > +struct rz_pinctrl_res { > + resource_size_t start; > + resource_size_t size; > + void __iomem *base; > +}; > + > +/** > + * rz_pinctrl_dev - RZ pincontroller device > + * > + * @dev: device > + * @info: SoC data > + * @nres: number of memory regions > + * @res: memory regions > + * @mutex: protect pin*_generic functions > + * @pins: pin array for pinctrl core > + * @desc: pincontroller desc for pinctrl core > + * @pctl: pinctrl device > + */ > +struct rz_pinctrl_dev { > + struct device *dev; > + > + struct rz_pinctrl_info *info; > + > + unsigned int nres; > + struct rz_pinctrl_res *res; > + > + struct mutex mutex; > + > + struct pinctrl_pin_desc *pins; > + struct pinctrl_desc desc; > + struct pinctrl_dev *pctl; > +}; > + > +/** > + * rz_pinctrl_ops - SoC operations > + * > + * @set_mux: perform pin muxing on SoC registers > + */ > +struct rz_pinctrl_ops { > + int (*set_mux)(struct rz_pinctrl_dev *, struct rz_pin_desc *, > + unsigned int mux_mode); > +}; > + > +/* ------------------------------------------------------------------------ > + * pinctrl-rz prototypes > + */ > +int rz_pinctrl_probe(struct platform_device *pdev, > + struct rz_pinctrl_info *info); > +void rz_pinctrl_remove(struct platform_device *pdev); > + > +#endif
Hi Laurent, On 01/02/2017 16:21, Laurent Pinchart wrote: > Hi Jacopo, > [snip] > >> +} >> + >> +/** >> + * rz_dt_node_to_map() - Parse device tree nodes and collect pins, groups >> and >> + * functions > > I don't think we have groups and functions, do we ? > Sort of.. The hardware does not have that, but we create them from pin configuration sub-nodes, as the pinmux core API is sort of group/function centric (see the pinmux_ops.set_mux function prototype) [snip] > + mutex_lock(&rz_pinctrl->mutex); This seems weird. The pinctrl generic functions indeed state that the caller must take care of locking, but there doesn't seem to be any lock protecting races between .dt_node_to_map() calling the generic functions below, and the generic .get_group_*() handlers. > + ret = pinctrl_generic_add_group(pctldev, grpname, grpins, npins, NULL); > + if (ret) { > + mutex_unlock(&rz_pinctrl->mutex); > + goto free_map; > + } > + > + ret = pinmux_generic_add_function(pctldev, grpname, fngrps, > + 1, mux_modes); > + if (ret) { > + mutex_unlock(&rz_pinctrl->mutex); > + goto free_map; > + } > + mutex_unlock(&rz_pinctrl->mutex); radix trees are RCU protected structures, so no locking should be required when accessing them from different places, so I guess lock here is just to protect the num_groups and num_functions variables in core pin controller driver... [snip] >> +/** >> + * rz_dt_free_map() >> + */ >> +static void rz_dt_free_map(struct pinctrl_dev *pctldev, >> + struct pinctrl_map *map, unsigned int num_maps) >> +{ >> + struct rz_pinctrl_dev *rz_pinctrl = pinctrl_dev_get_drvdata(pctldev); >> + >> + devm_kfree(rz_pinctrl->dev, map); > > Shouldn't you also free the other allocated pieces of memory (fngrps, > mux_modes and grpins) ? > That's an interesting one. If I have to free all data associated with that map (and I assume so) I shall be able to retrieve groups and functions that map represents and delete them and Currently there is no generic pinctrl and pinmux function to retrieve a group or function by name, but only by their id (selector). It would take 10minutes to add them, but I wonder if that's desirable or there are other ways to do so I haven't found out yet. Linus, Tony and gpio people: do you have opinions on this? I can add those functions to core/pinmux in my next patch series if I get your ack here. Thanks j -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
* jacopo mondi <jacopo@jmondi.org> [170206 10:16]: > Currently there is no generic pinctrl and pinmux function to retrieve a > group or function by name, but only by their id (selector). > It would take 10minutes to add them, but I wonder if that's desirable or > there are other ways to do so I haven't found out yet. > > Linus, Tony and gpio people: do you have opinions on this? I can add those > functions to core/pinmux in my next patch series if I get your ack here. I would just get rid of the names and allow optionally configuring them from the consumer. Just like we do for interrupts and GPIOs. Regards, Tony -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Hi Tony, On 06/02/2017 19:28, Tony Lindgren wrote: > * jacopo mondi <jacopo@jmondi.org> [170206 10:16]: >> Currently there is no generic pinctrl and pinmux function to retrieve a >> group or function by name, but only by their id (selector). >> It would take 10minutes to add them, but I wonder if that's desirable or >> there are other ways to do so I haven't found out yet. >> >> Linus, Tony and gpio people: do you have opinions on this? I can add those >> functions to core/pinmux in my next patch series if I get your ack here. > > I would just get rid of the names and allow optionally configuring them > from the consumer. Just like we do for interrupts and GPIOs. > Mostly, I wonder why so far nobody had to remove function and groups and release their associated data when deleting a map. It makes me think this is not required or I didn't fully get dt_free_map() purpose :/ Or, put it in another way, I see lot of code releasing memory associated with PIN_MAP_TYPE_CONFIGS_GROUP map types (as map.data.configs.configs it's easy accessible), but none releasing data associated to PIN_MAP_TYPE_MUX_GROUP map types... Thanks j > Regards, > > Tony > -- To unsubscribe from this list: send the line "unsubscribe linux-gpio" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/pinctrl/Kconfig b/drivers/pinctrl/Kconfig index 8f8c2af..6d72e58 100644 --- a/drivers/pinctrl/Kconfig +++ b/drivers/pinctrl/Kconfig @@ -294,6 +294,7 @@ source "drivers/pinctrl/mvebu/Kconfig" source "drivers/pinctrl/nomadik/Kconfig" source "drivers/pinctrl/pxa/Kconfig" source "drivers/pinctrl/qcom/Kconfig" +source "drivers/pinctrl/rz-pfc/Kconfig" source "drivers/pinctrl/samsung/Kconfig" source "drivers/pinctrl/sh-pfc/Kconfig" source "drivers/pinctrl/spear/Kconfig" diff --git a/drivers/pinctrl/Makefile b/drivers/pinctrl/Makefile index a251f43..96e7ece 100644 --- a/drivers/pinctrl/Makefile +++ b/drivers/pinctrl/Makefile @@ -48,6 +48,7 @@ obj-$(CONFIG_PINCTRL_MVEBU) += mvebu/ obj-y += nomadik/ obj-$(CONFIG_PINCTRL_PXA) += pxa/ obj-$(CONFIG_ARCH_QCOM) += qcom/ +obj-$(CONFIG_PINCTRL_RZ_PINCTRL) += rz-pfc/ obj-$(CONFIG_PINCTRL_SAMSUNG) += samsung/ obj-$(CONFIG_PINCTRL_SH_PFC) += sh-pfc/ obj-$(CONFIG_PINCTRL_SPEAR) += spear/ diff --git a/drivers/pinctrl/rz-pfc/Kconfig b/drivers/pinctrl/rz-pfc/Kconfig new file mode 100644 index 0000000..3714c10 --- /dev/null +++ b/drivers/pinctrl/rz-pfc/Kconfig @@ -0,0 +1,18 @@ +# +# Renesas RZ pinctrl drivers +# + +if ARCH_RENESAS + +config PINCTRL_RZ_PINCTRL + select PINMUX + select PINCONF + select GPIOLIB + select GENERIC_PINCONF + select GENERIC_PINMUX_FUNCTIONS + select GENERIC_PINCTRL_GROUPS + def_bool y + help + This enables pin control drivers for Renesas RZ platforms + +endif diff --git a/drivers/pinctrl/rz-pfc/Makefile b/drivers/pinctrl/rz-pfc/Makefile new file mode 100644 index 0000000..cba8283 --- /dev/null +++ b/drivers/pinctrl/rz-pfc/Makefile @@ -0,0 +1 @@ +obj-$(CONFIG_PINCTRL_RZ_PINCTRL) += pinctrl-rz.o diff --git a/drivers/pinctrl/rz-pfc/pinctrl-rz.c b/drivers/pinctrl/rz-pfc/pinctrl-rz.c new file mode 100644 index 0000000..3efbf03 --- /dev/null +++ b/drivers/pinctrl/rz-pfc/pinctrl-rz.c @@ -0,0 +1,447 @@ +/* + * Pinctrl support for Renesas RZ Series + * + * Copyright (C) 2017 Jacopo Mondi + * Copyright (C) 2017 Renesas Electronics Corporation + * + * This file is licensed under the terms of the GNU General Public + * License version 2. This program is licensed "as is" without any + * warranty of any kind, whether express or implied. + */ + +#include <linux/err.h> +#include <linux/init.h> +#include <linux/ioport.h> +#include <linux/module.h> + +#include <linux/of.h> +#include <linux/of_address.h> +#include <linux/of_device.h> + +#include <linux/pinctrl/pinconf-generic.h> +#include <linux/pinctrl/pinctrl.h> +#include <linux/pinctrl/pinmux.h> + +#include "../core.h" +#include "../devicetree.h" +#include "../pinmux.h" + +#include "pinctrl-rz.h" + +#define DRIVER_NAME "rz-pinctrl" +#define RZ_PIN_ARGS_COUNT 3 + +/** + * rz_pinctrl_pos_to_index() - Retrieve the index of pin at position [bank:pin] + * + * This can be improved, as it walks all the pins reported by the SoC driver + * + * @return: pin number between [0 - npins]; -1 if not found + */ +static int rz_pinctrl_pos_to_index(struct rz_pinctrl_dev *rz_pinctrl, + unsigned int bank, unsigned int pin) +{ + struct rz_pinctrl_info *info; + struct rz_pin_desc *rz_pin; + int i, npins; + + info = rz_pinctrl->info; + npins = info->npins; + for (i = 0; i < npins; ++i) { + rz_pin = &info->pins[i]; + + /* + * return the pin index in the array, not the pin number, + * so that we can access it easily when muxing group's pins + */ + if (rz_pin->bank == bank && rz_pin->pin == pin) + return i; + } + + return -1; +} + +/* ---------------------------------------------------------------------------- + * pinctrl operations + */ +static void rz_pin_dbg_show(struct pinctrl_dev *pctldev, struct seq_file *s, + unsigned int pin) +{ + struct rz_pinctrl_dev *rz_pinctrl; + struct rz_pinctrl_info *info; + + rz_pinctrl = pinctrl_dev_get_drvdata(pctldev); + info = rz_pinctrl->info; + + seq_printf(s, "%s %s\n", info->pins[pin].name, DRIVER_NAME); +} + +/** + * rz_dt_node_to_map() - Parse device tree nodes and collect pins, groups and + * functions + * + * Pins for RZ series pin controller described by "renesas-rz,pins" property + * are arrays of u32 values in the form: + * + * "renesas-rz,pins" = <BANK PIN MUX>, ... ; + * + * Parse the list arguments and identify pins through their position + * (bank and pin offset) and save the provided mux mode for later use. + * + * TODO: the array can be expended to support additional parameters for + * pin configurations values (IO direction etc) + * + * @pctldev: pin controller device + * @np_config: device tree node to parse + * @map: pointer to pin map (output) + * @num_maps: number of collected maps (output) + * + * @return: 0 for success; != 0 otherwise + */ +static int rz_dt_node_to_map(struct pinctrl_dev *pctldev, + struct device_node *np_config, + struct pinctrl_map **map, unsigned int *num_maps) +{ + struct rz_pinctrl_dev *rz_pinctrl; + const char *prop_name = "renesas-rz,pins", + *grpname, **fngrps; + unsigned int *grpins, + *mux_modes; + int i, npins, ret; + + rz_pinctrl = pinctrl_dev_get_drvdata(pctldev); + + npins = pinctrl_count_index_with_args(np_config, prop_name); + if (npins <= 0) { + dev_err(rz_pinctrl->dev, "Missing pins configuration prop\n"); + return -EINVAL; + } + + mux_modes = devm_kzalloc(rz_pinctrl->dev, sizeof(*mux_modes) * npins, + GFP_KERNEL); + grpins = devm_kzalloc(rz_pinctrl->dev, sizeof(*grpins) * npins, + GFP_KERNEL); + if (unlikely(!grpins || !mux_modes)) + return -ENOMEM; + + /* + * functions are made of 1 group only; + * in facts, functions and groups are identical for RZ pin controller, + * except that functions carry an array of mux modes (aka alternate + * functions IDs) + */ + fngrps = devm_kzalloc(rz_pinctrl->dev, sizeof(*fngrps), GFP_KERNEL); + if (unlikely(!fngrps)) { + ret = -ENOMEM; + goto free_pins; + } + + *num_maps = 0; + (*map) = devm_kzalloc(rz_pinctrl->dev, sizeof(**map), GFP_KERNEL); + if (unlikely(!*map)) { + ret = -ENOMEM; + goto free_fngrps; + } + + /* collect pin numbers and mux confs to create groups and functions */ + for (i = 0; i < npins; ++i) { + struct of_phandle_args of_pins_args; + unsigned int bank, offs, mode, pin_idx; + + /* DTS identifies pin by position: bank and pin offset */ + ret = pinctrl_parse_index_with_args(np_config, prop_name, i, + &of_pins_args); + if (ret) + goto free_map; + + if (of_pins_args.args_count < RZ_PIN_ARGS_COUNT) { + dev_err(rz_pinctrl->dev, + "Wrong arguments number for renesas-rz,pins"); + ret = -EINVAL; + goto free_map; + } + + bank = of_pins_args.args[0]; + offs = of_pins_args.args[1]; + mode = of_pins_args.args[2]; + + /* pin_idx is the index on the static pin array */ + pin_idx = rz_pinctrl_pos_to_index(rz_pinctrl, bank, offs); + if (pin_idx < 0) { + dev_err(rz_pinctrl->dev, + "Invalid pin position %d-%d\n", bank, offs); + ret = -EINVAL; + goto free_map; + } + + grpins[i] = pin_idx; + mux_modes[i] = mode; + } + + grpname = np_config->name; + fngrps[0] = grpname; + + mutex_lock(&rz_pinctrl->mutex); + ret = pinctrl_generic_add_group(pctldev, grpname, grpins, npins, NULL); + if (ret) { + mutex_unlock(&rz_pinctrl->mutex); + goto free_map; + } + + ret = pinmux_generic_add_function(pctldev, grpname, fngrps, + 1, mux_modes); + if (ret) { + mutex_unlock(&rz_pinctrl->mutex); + goto free_map; + } + mutex_unlock(&rz_pinctrl->mutex); + + (*map)->type = PIN_MAP_TYPE_MUX_GROUP; + (*map)->data.mux.group = grpname; + (*map)->data.mux.function = grpname; + *num_maps = 1; + + return 0; + +free_map: + devm_kfree(rz_pinctrl->dev, *map); +free_fngrps: + devm_kfree(rz_pinctrl->dev, fngrps); +free_pins: + devm_kfree(rz_pinctrl->dev, mux_modes); + devm_kfree(rz_pinctrl->dev, grpins); + return ret; +} + +/** + * rz_dt_free_map() + */ +static void rz_dt_free_map(struct pinctrl_dev *pctldev, + struct pinctrl_map *map, unsigned int num_maps) +{ + struct rz_pinctrl_dev *rz_pinctrl = pinctrl_dev_get_drvdata(pctldev); + + devm_kfree(rz_pinctrl->dev, map); +} + +static const struct pinctrl_ops rz_pinctrl_ops = { + .get_groups_count = pinctrl_generic_get_group_count, + .get_group_name = pinctrl_generic_get_group_name, + .get_group_pins = pinctrl_generic_get_group_pins, + .pin_dbg_show = rz_pin_dbg_show, + .dt_node_to_map = rz_dt_node_to_map, + .dt_free_map = rz_dt_free_map, +}; + +/* ---------------------------------------------------------------------------- + * pinconfig operations + */ + +/* TODO */ + +/* ---------------------------------------------------------------------------- + * pinmux operations + */ + +/** + * rz_pinmux_set() - Retrieve pins from a group and apply mux settings + * + * @pctldev: pin controller device + * @selector: Function selector + * @group: Group selector + * @return: 0 for success; != otherwise + */ +static int rz_pinmux_set(struct pinctrl_dev *pctldev, unsigned int selector, + unsigned int group) +{ + struct rz_pinctrl_dev *rz_pinctrl; + struct rz_pinctrl_info *info; + struct rz_pinctrl_ops *ops; + struct group_desc *grp; + struct function_desc *func; + unsigned int npins, *mux_modes; + int i; + + rz_pinctrl = pinctrl_dev_get_drvdata(pctldev); + info = rz_pinctrl->info; + ops = info->ops; + + grp = pinctrl_generic_get_group(pctldev, group); + if (!grp) + return -EINVAL; + + func = pinmux_generic_get_function(pctldev, selector); + if (!func) + return -EINVAL; + + npins = grp->num_pins; + mux_modes = (unsigned int *)func->data; + + for (i = 0; i < npins; ++i) { + unsigned int pin_idx, mux_mode; + struct rz_pin_desc *pin; + int ret; + + /* + * use pin index to retrieve pin descriptor; + * then provide it to the set_mux SoC operation + */ + pin_idx = grp->pins[i]; + pin = &info->pins[pin_idx]; + mux_mode = mux_modes[i]; + + if (!ops || !ops->set_mux) { + dev_err(rz_pinctrl->dev, "Pin muxing not supported\n"); + return -EINVAL; + } + + ret = ops->set_mux(rz_pinctrl, pin, mux_mode); + if (ret) + return ret; + } + + return 0; +} + +struct pinmux_ops rz_pinmux_ops = { + .get_functions_count = pinmux_generic_get_function_count, + .get_function_name = pinmux_generic_get_function_name, + .get_function_groups = pinmux_generic_get_function_groups, + .set_mux = rz_pinmux_set, +}; + +/* ---------------------------------------------------------------------------- + * rz pinctrl operations + */ + +/** + * rz_add_pins() - Enumerate pins reported by SoC driver in pin desc array + * + * @rz_pinctrl: RZ pincontroller + * + * @return: 0 for success; != 0 otherwise + */ +static int rz_add_pins(struct rz_pinctrl_dev *rz_pinctrl) +{ + struct rz_pinctrl_info *info; + struct rz_pin_desc *pin; + unsigned int npins, i; + + info = rz_pinctrl->info; + npins = info->npins; + + rz_pinctrl->pins = devm_kzalloc(rz_pinctrl->dev, + npins * sizeof(*rz_pinctrl->pins), + GFP_KERNEL); + if (unlikely(!rz_pinctrl->pins)) + return -ENOMEM; + + rz_pinctrl->desc.pins = rz_pinctrl->pins; + rz_pinctrl->desc.npins = npins; + + for (i = 0; i < npins; ++i) { + pin = &info->pins[i]; + + rz_pinctrl->pins[i].number = pin->number; + rz_pinctrl->pins[i].name = pin->name; + } + + return 0; +} + +/** + * rz_pinctrl_map_res() - Map memory resources for pincontroller + * + * @pdev: platform device + * @rz_pincrl: RZ pincontroller + * + * @return: 0 for success; != 0 otherwise + */ +static int rz_pinctrl_map_res(struct platform_device *pdev, + struct rz_pinctrl_dev *rz_pinctrl) +{ + struct resource *res; + struct rz_pinctrl_res *rz_res; + unsigned int i, nres; + + nres = pdev->num_resources; + rz_res = devm_kzalloc(&pdev->dev, nres * sizeof(*rz_res), + GFP_KERNEL); + if (unlikely(!rz_res)) + return -ENOMEM; + + rz_pinctrl->nres = nres; + rz_pinctrl->res = rz_res; + + for (i = 0; i < nres; i++) { + res = platform_get_resource(pdev, IORESOURCE_MEM, i); + if (!res) + return -ENODEV; + + rz_res = &rz_pinctrl->res[i]; + + rz_res->start = res->start; + rz_res->size = resource_size(res); + rz_res->base = devm_ioremap_resource(&pdev->dev, res); + if (IS_ERR(rz_res->base)) + return PTR_ERR(rz_res->base); + } + + return 0; +} + +/** + * rz_pinctrl_probe() - Register pincontroller driver operations + * + * @pdev: platform device + * @info: SoC device info structure + */ +int rz_pinctrl_probe(struct platform_device *pdev, struct rz_pinctrl_info *info) +{ + int ret; + struct rz_pinctrl_dev *rz_pinctrl; + + rz_pinctrl = devm_kzalloc(&pdev->dev, sizeof(*rz_pinctrl), GFP_KERNEL); + if (!rz_pinctrl) + return -ENOMEM; + + rz_pinctrl->dev = &pdev->dev; + rz_pinctrl->info = info; + + ret = rz_pinctrl_map_res(pdev, rz_pinctrl); + if (ret) + return ret; + + platform_set_drvdata(pdev, rz_pinctrl); + + rz_pinctrl->desc.name = DRIVER_NAME; + rz_pinctrl->desc.pctlops = &rz_pinctrl_ops; + rz_pinctrl->desc.pmxops = &rz_pinmux_ops; + rz_pinctrl->desc.owner = THIS_MODULE; + + ret = rz_add_pins(rz_pinctrl); + if (ret) + return ret; + + mutex_init(&rz_pinctrl->mutex); + + ret = pinctrl_register_and_init(&rz_pinctrl->desc, rz_pinctrl->dev, + rz_pinctrl, &rz_pinctrl->pctl); + if (ret) { + dev_err(&pdev->dev, "could not register rz pinctrl driver\n"); + return ret; + } + + return 0; +} + +void rz_pinctrl_remove(struct platform_device *pdev) +{ + struct rz_pinctrl_dev *rz_pinctrl = platform_get_drvdata(pdev); + + pinctrl_unregister(rz_pinctrl->pctl); +} + +MODULE_AUTHOR("Jacopo Mondi <jacopo+renesas@jmondi.org"); +MODULE_DESCRIPTION("Pinctrl driver for Reneas RZ series"); +MODULE_LICENSE("GPL v2"); diff --git a/drivers/pinctrl/rz-pfc/pinctrl-rz.h b/drivers/pinctrl/rz-pfc/pinctrl-rz.h new file mode 100644 index 0000000..b873c89 --- /dev/null +++ b/drivers/pinctrl/rz-pfc/pinctrl-rz.h @@ -0,0 +1,114 @@ +/* + * Pinctrl support for Renesas RZ Series + * + * Copyright (C) 2017 Jacopo Mondi + * Copyright (C) 2017 Renesas Electronics Corporation + * + * This file is licensed under the terms of the GNU General Public + * License version 2. This program is licensed "as is" without any + * warranty of any kind, whether express or implied. + */ +#ifndef __RZ_PINCTRL_H__ +#define __RZ_PINCTRL_H__ + +#include <linux/pinctrl/pinctrl.h> +#include <linux/types.h> + +/* ---------------------------------------------------------------------------- + * pinctrl-rz data types + */ +#define RZ_PIN_NAME(bank, pin) \ + PIN_##bank##_##pin + +#define RZ_PIN_DESC(b, p) \ + { .number = RZ_PIN_NAME(b, p), \ + .name = __stringify(RZ_PIN_NAME(b, p)), \ + .bank = b, .pin = p } + +/** + * rz_pin_desc - Single RZ pin descriptor + * + * @number: pin number as enumerated by SoC driver + * @name: pin name as reported by SoC driver + * @bank: pin bank location + * @pin: pin register offset + */ +struct rz_pin_desc { + unsigned int bank, pin; + unsigned int number; + const char *name; +}; + +/** + * rz_pinctrl_info - SoC info data + * + * @ops: SoC specific operations + * @npins: number of pins + * @pins: pin array + */ +struct rz_pinctrl_ops; +struct rz_pinctrl_info { + struct rz_pinctrl_ops *ops; + + unsigned int npins; + struct rz_pin_desc *pins; +}; + +/** + * rz_pinctrl_res - Memory resource for RZ SoC + * + * @start: physical address base + * @size: memory region size + * @base: logical address base + */ +struct rz_pinctrl_res { + resource_size_t start; + resource_size_t size; + void __iomem *base; +}; + +/** + * rz_pinctrl_dev - RZ pincontroller device + * + * @dev: device + * @info: SoC data + * @nres: number of memory regions + * @res: memory regions + * @mutex: protect pin*_generic functions + * @pins: pin array for pinctrl core + * @desc: pincontroller desc for pinctrl core + * @pctl: pinctrl device + */ +struct rz_pinctrl_dev { + struct device *dev; + + struct rz_pinctrl_info *info; + + unsigned int nres; + struct rz_pinctrl_res *res; + + struct mutex mutex; + + struct pinctrl_pin_desc *pins; + struct pinctrl_desc desc; + struct pinctrl_dev *pctl; +}; + +/** + * rz_pinctrl_ops - SoC operations + * + * @set_mux: perform pin muxing on SoC registers + */ +struct rz_pinctrl_ops { + int (*set_mux)(struct rz_pinctrl_dev *, struct rz_pin_desc *, + unsigned int mux_mode); +}; + +/* ---------------------------------------------------------------------------- + * pinctrl-rz prototypes + */ +int rz_pinctrl_probe(struct platform_device *pdev, + struct rz_pinctrl_info *info); +void rz_pinctrl_remove(struct platform_device *pdev); + +#endif
Add core module for per-pin Renesas RZ series pin controller. The core module allows SoC driver to register their pins and SoC specific operations and interfaces with pinctrl and pinmux core on their behalf. Signed-off-by: Jacopo Mondi <jacopo+renesas@jmondi.org> --- drivers/pinctrl/Kconfig | 1 + drivers/pinctrl/Makefile | 1 + drivers/pinctrl/rz-pfc/Kconfig | 18 ++ drivers/pinctrl/rz-pfc/Makefile | 1 + drivers/pinctrl/rz-pfc/pinctrl-rz.c | 447 ++++++++++++++++++++++++++++++++++++ drivers/pinctrl/rz-pfc/pinctrl-rz.h | 114 +++++++++ 6 files changed, 582 insertions(+) create mode 100644 drivers/pinctrl/rz-pfc/Kconfig create mode 100644 drivers/pinctrl/rz-pfc/Makefile create mode 100644 drivers/pinctrl/rz-pfc/pinctrl-rz.c create mode 100644 drivers/pinctrl/rz-pfc/pinctrl-rz.h