Re: [PATCH 3/6] gpio: Add GPIO support for the ST Multi-Function eXpander

From: Amelie DELAUNAY
Date: Mon Feb 19 2018 - 12:14:53 EST




On 02/14/2018 04:30 PM, Andy Shevchenko wrote:
> On Thu, Feb 8, 2018 at 4:27 PM, Amelie Delaunay <amelie.delaunay@xxxxxx> wrote:
>> ST Multi-Function eXpander (MFX) can be used as GPIO expander.
>> It has 16 fast GPIOs and can have 8 extra alternate GPIOs
>> when other MFX features are not enabled.
>
>> +config GPIO_ST_MFX
>> + bool "ST-MFX GPIOs"
>> + depends on MFD_ST_MFX
>> + depends on OF_GPIO
>> + select GPIOLIB_IRQCHIP
>> + help
>> + GPIO driver for STMicroelectronics Multi-Function eXpander.
>> +
>> + This provides GPIO interface supporting inputs and outputs.
>
>> +/*
>> + * STMicroelectronics Multi-Function eXpander (ST-MFX) - GPIO expander driver.
>> + *
>> + * Copyright (C) 2017, STMicroelectronics - All Rights Reserved
>> + * Author(s): Amelie Delaunay <amelie.delaunay@xxxxxx> for STMicroelectronics.
>
>> + * License terms: GPL V2.0.
>> + *
>> + * st-mfx-gpio is free software; you can redistribute it and/or modify it
>> + * under the terms of the GNU General Public License version 2 as published by
>> + * the Free Software Foundation.
>> + *
>> + * st-mfx-gpio is distributed in the hope that it will be useful, but
>> + * WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
>> + * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more
>> + * details.
>> + *
>> + * You should have received a copy of the GNU General Public License along with
>> + * st-mfx-gpio. If not, see <http://www.gnu.org/licenses/>.
>
> Use SPDX instead.
>
OK.
>> + */
>
>> +#include <linux/of_gpio.h>
>
>> +/* MFX has a maximum of 24 gpios, with 8 gpios per bank, so 3 banks maximum */
>> +#define NR_MAX_GPIOS 24
>> +#define NR_GPIOS_PER_BANK 8
>> +#define NR_MAX_BANKS (NR_MAX_GPIOS / NR_GPIOS_PER_BANK)
>
>> +#define get_bank(offset) ((u8)((offset) / NR_GPIOS_PER_BANK))
>> +#define get_mask(offset) ((u8)BIT((offset) % NR_GPIOS_PER_BANK))
>
> I would rather keep it consistent, i.e.
> get_bank() [as is]
> get_mask -> get_shift() [w/o BIT() macro]
>
#define get_shift(offset) ((u8)((offset) % NR_GPIOS_PER_BANK))
u8 mask = (u8)BIT(get_shift(offset));
I'm not sure consistency is more important than readability and
convenience ?

>> +enum { REG_IRQ_SRC, REG_IRQ_EVT, REG_IRQ_TYPE, NR_CACHE_IRQ_REGS };
>
> Please, do one item per line.
>
OK.

>> +/*
>> + * This is MFX-specific flags, overloading Linux-specific of_gpio_flags,
>> + * needed in of_xlate callback.
>
>> + * on MFX: - if GPIO is output:
>
> Split to two lines.
>
OK.

>> + * * (0) means PUSH_PULL
>> + * * OF_GPIO_SINGLE_ENDED (=2) means OPEN-DRAIN
>> + * - if GPIO is input:
>> + * * (0) means NO_PULL
>> + * * OF_MFX_GPI_PUSH_PULL (=2) means PUSH_PULL
>> + *
>> + * * OF_MFX_GPIO_PULL_UP programs pull up resistor on the GPIO,
>> + * whatever its direction, without this flag, depending on
>> + * GPIO type and direction, it programs either no pull or
>> + * pull down resistor.
>> + */
>> +enum of_mfx_gpio_flags {
>> + OF_MFX_GPI_PUSH_PULL = 0x2,
>> + OF_MFX_GPIO_PULL_UP = 0x4,
>
> These have misleading prefix. OF_ is reserved for general OF stuff.
>
You're right. I will fix it in V2.

>> +};
>> +
>> +struct mfx_gpio {
>> + struct gpio_chip gc;
>> + struct mfx *mfx;
>
>> + struct device *dev;
>
> Don't you have it in gc above as a parent device?
>
Yes, I agree to remove it.

>> + struct mutex irq_lock; /* IRQ bus lock */
>> + u32 flags[NR_MAX_GPIOS];
>> + /* Caches of interrupt control registers for bus_lock */
>> + u8 regs[NR_CACHE_IRQ_REGS][NR_MAX_BANKS];
>> + u8 oldregs[NR_CACHE_IRQ_REGS][NR_MAX_BANKS];
>> +};
>
>> +static int mfx_gpio_direction_input(struct gpio_chip *gc,
>> + unsigned int offset)
>> +{
>> + struct mfx_gpio *mfx_gpio = gpiochip_get_data(gc);
>> + struct mfx *mfx = mfx_gpio->mfx;
>> + u8 reg_type = MFX_REG_GPIO_TYPE + get_bank(offset);
>> + u8 reg_pupd = MFX_REG_GPIO_PUPD + get_bank(offset);
>> + u8 reg_dir = MFX_REG_GPIO_DIR + get_bank(offset);
>> + u8 mask = get_mask(offset);
>> + int ret;
>> +
>> + /*
>> + * In case of input direction, there is actually no way to apply some
>> + * configuration in hardware, as it can be done with
>> + * .set_config in case of output direction. That's why we apply
>> + * here the MFX specific-flags.
>> + */
>> + if (mfx_gpio->flags[offset] & OF_MFX_GPI_PUSH_PULL)
>> + ret = mfx_set_bits(mfx, reg_type, mask, mask);
>> + else
>> + ret = mfx_set_bits(mfx, reg_type, mask, 0);
>
>> +
>
> Redundant empty line.
>
OK.

>> + if (ret)
>> + return ret;
>> +
>> + if (mfx_gpio->flags[offset] & OF_MFX_GPIO_PULL_UP)
>> + ret = mfx_set_bits(mfx, reg_pupd, mask, mask);
>> + else
>> + ret = mfx_set_bits(mfx, reg_pupd, mask, 0);
>
>> +
>
> Ditto.
>
OK.

>> + if (ret)
>> + return ret;
>> +
>> + return mfx_set_bits(mfx, reg_dir, mask, 0);
>> +}
>
>> +static void mfx_gpio_dbg_show(struct seq_file *s, struct gpio_chip *gc)
>> +{
>> + struct mfx_gpio *mfx_gpio = gpiochip_get_data(gc);
>> + struct mfx *mfx = mfx_gpio->mfx;
>> + u16 i;
>> +
>> + for (i = 0; i < gc->ngpio; i++) {
>> + int gpio = i + gc->base;
>> + const char *label = gpiochip_is_requested(gc, i);
>> + int dir = mfx_gpio_get_direction(gc, i);
>> + int val = mfx_gpio_get(gc, i);
>> + u8 mask = get_mask(i);
>> + u8 reg;
>> + int type, pupd;
>> + int irqsrc, irqevt, irqtype, irqpending;
>
>> + if (!label)
>> + label = "Unrequested";
>
> I would rather put label = gpiochip_...(); immediately before this
> condition for better readability.
>
You're right.

>> +
>> + seq_printf(s, " gpio-%-3d (%-20.20s) ", gpio, label);
>> +
>
>> + reg = MFX_REG_IRQ_GPI_PENDING + get_bank(i);
>> + irqpending = mfx_reg_read(mfx, reg);
>> + if (irqpending < 0)
>> + return;
>> + irqpending = !!(irqpending & mask);
>> +
>
>> + if (!dir) {
>
> Why not to use
>
> if (dir) {
>
> ?
>
My brain orders things in ascending order, I think it is the reason why
this code manages false cases (=0) before the others.

>> + seq_printf(s, "out %s ", val ? "high" : "low");
>> + if (type)
>> + seq_puts(s, "open drain ");
>> + else
>> + seq_puts(s, "push pull ");
>> + if (pupd && type)
>> + seq_puts(s, "with internal pull-up ");
>> + else
>> + seq_puts(s, "without internal pull-up ");
>> + } else {
>> + seq_printf(s, "in %s ", val ? "high" : "low");
>> + if (type)
>> + seq_printf(s, "with internal pull-%s ",
>> + pupd ? "up" : "down");
>> + else
>> + seq_printf(s, "%s ",
>> + pupd ? "floating" : "analog");
>> + }
>> +
>> + if (irqsrc) {
>> + seq_printf(s, "IRQ generated on %s %s",
>> + !irqevt ?
>> + (!irqtype ? "Low level" : "High level") :
>> + (!irqtype ? "Falling edge" : "Rising edge"),
>> + irqpending ? "(pending)" : "");
>
> Why do you use negative conditions? Use positive ones.
>
OK, I will make the effort in V2.
>> + }
>> +
>> + seq_puts(s, "\n");
>> + }
>> +}
>
>> +static void mfx_gpio_irq_toggle_trigger(struct gpio_chip *gc, int offset)
>> +{
>> + struct mfx_gpio *mfx_gpio = gpiochip_get_data(gc);
>> + struct mfx *mfx = mfx_gpio->mfx;
>> + u8 bank = get_bank(offset);
>> + u8 mask = get_mask(offset);
>> + int val = mfx_gpio_get(gc, offset);
>> +
>
>> + if (val < 0) {
>> + dev_err(mfx_gpio->dev, "can't get value of gpio%d: ret=%d\n",
>> + offset, val);
>
> Is it somehow useful on err level?
>
I can drop it.
>> + return;
>> + }
>
>> +}
>
>> +static int mfx_gpio_irq_set_type(struct irq_data *d, unsigned int type)
>> +{
>> + struct gpio_chip *gc = irq_data_get_irq_chip_data(d);
>> + struct mfx_gpio *mfx_gpio = gpiochip_get_data(gc);
>> + int bank = get_bank(d->hwirq);
>> + int mask = get_mask(d->hwirq);
>> +
>
>> + if ((type & IRQ_TYPE_LEVEL_LOW) || (type & IRQ_TYPE_LEVEL_HIGH))
>
> IRQ_TYPE_LEVEL_MASK?
>
You're right.
>> + mfx_gpio->regs[REG_IRQ_EVT][bank] &= ~mask;
>> + else
>> + mfx_gpio->regs[REG_IRQ_EVT][bank] |= mask;
>> +
>> + if ((type & IRQ_TYPE_EDGE_RISING) || (type & IRQ_TYPE_LEVEL_HIGH))
>
> IRQ_TYPE_EDGE_BOTH ?
>
No, here I test EDGE_RISING and LEVEL_HIGH, so mixing EDGE and LEVEL type.
>> + mfx_gpio->regs[REG_IRQ_TYPE][bank] |= mask;
>> + else
>> + mfx_gpio->regs[REG_IRQ_TYPE][bank] &= ~mask;
>> +
>
>> +}
>
>> +static void mfx_gpio_irq_lock(struct irq_data *d)
>
> Missed annotation that this acquires a lock.
>
It is .irq_bus_lock and .irq_bus_sync_unlock. Description is available
in include/linux/irq.h:
* @irq_bus_lock: function to lock access to slow bus (i2c) chips
* @irq_bus_sync_unlock:function to sync and unlock slow bus (i2c) chips
>> +static void mfx_gpio_irq_sync_unlock(struct irq_data *d)
>
> Ditto for releasing lock.
>
>> +static irqreturn_t mfx_gpio_irq(int irq, void *dev)
>> +{
>> + struct mfx_gpio *mfx_gpio = dev;
>> + struct mfx *mfx = mfx_gpio->mfx;
>> + int num_banks = mfx->num_gpio / 8;
>> + u8 status[num_banks];
>
>> + int ret;
>> + u8 bank;
>
> Swap lines.
>
OK.
>> +
>> + ret = mfx_block_read(mfx, MFX_REG_IRQ_GPI_PENDING, ARRAY_SIZE(status),
>> + status);
>> + if (ret < 0) {
>
>> + dev_err(mfx_gpio->dev, "can't get IRQ_GPI_PENDING: %d\n", ret);
>
> In IRQ context on err level? Hmm...
>
OK I will remove it in V2.
>> + return IRQ_NONE;
>> + }
>> +
>> + for (bank = 0; bank < ARRAY_SIZE(status); bank++) {
>> + u8 stat = status[bank];
>> +
>> + if (!stat)
>> + continue;
>> +
>
>> + while (stat) {
>> + int bit = __ffs(stat);
>
> for_each_set_bit() ?
>
You're right.
>> + int offset = bank * NR_GPIOS_PER_BANK + bit;
>> + int gpio_irq = irq_find_mapping(mfx_gpio->gc.irq.domain,
>> + offset);
>> + int irq_trigger = irq_get_trigger_type(gpio_irq);
>> +
>> + handle_nested_irq(gpio_irq);
>> + stat &= ~(BIT(bit));
>> +
>> + if (irq_trigger & IRQ_TYPE_EDGE_BOTH)
>> + mfx_gpio_irq_toggle_trigger(&mfx_gpio->gc,
>> + offset);
>> + }
>> +
>> + mfx_reg_write(mfx, MFX_REG_IRQ_GPI_ACK + bank, status[bank]);
>> + }
>> +
>> + return IRQ_HANDLED;
>> +}
>

Thanks for review,
Amelie