Re: [PATCH v3 2/2] counter: add GPIO based pulse counters

[Jonathan Cameron]

On Fri, 22 Jan 2021 12:24:34 +0100
Oleksij Rempel <o.rempel@xxxxxxxxxxxxxx> wrote:

> Add simple GPIO base pulse counter. This device is used to measure
> rotation speed of some agricultural devices, so no high frequency on the
> counter pin is expected.
> 
> The maximal measurement frequency depends on the CPU and system load. On
> the idle iMX6S I was able to measure up to 20kHz without count drops.
> 
> Signed-off-by: Oleksij Rempel <o.rempel@xxxxxxxxxxxxxx>

Hi Oleksij,

A few comments inline.

Thanks,

Jonathan

> ---
>  drivers/counter/Kconfig          |   9 ++
>  drivers/counter/Makefile         |   1 +
>  drivers/counter/gpio-pulse-cnt.c | 244 +++++++++++++++++++++++++++++++
>  3 files changed, 254 insertions(+)
>  create mode 100644 drivers/counter/gpio-pulse-cnt.c
> 
> diff --git a/drivers/counter/Kconfig b/drivers/counter/Kconfig
> index 2de53ab0dd25..9ad1d9d49dd1 100644
> --- a/drivers/counter/Kconfig
> +++ b/drivers/counter/Kconfig
> @@ -29,6 +29,15 @@ config 104_QUAD_8
>  	  The base port addresses for the devices may be configured via the base
>  	  array module parameter.
>  
> +config GPIO_PULSE_CNT
> +	tristate "GPIO pulse counter driver"
> +	depends on GPIOLIB
> +	help
> +	  Select this option to enable GPIO pulse counter driver.
> +
> +	  To compile this driver as a module, choose M here: the
> +	  module will be called gpio-pulse-cnt.
> +
>  config STM32_TIMER_CNT
>  	tristate "STM32 Timer encoder counter driver"
>  	depends on MFD_STM32_TIMERS || COMPILE_TEST
> diff --git a/drivers/counter/Makefile b/drivers/counter/Makefile
> index 0a393f71e481..6a5c3fc6f2a0 100644
> --- a/drivers/counter/Makefile
> +++ b/drivers/counter/Makefile
> @@ -6,6 +6,7 @@
>  obj-$(CONFIG_COUNTER) += counter.o
>  
>  obj-$(CONFIG_104_QUAD_8)	+= 104-quad-8.o
> +obj-$(CONFIG_GPIO_PULSE_CNT)	+= gpio-pulse-cnt.o
>  obj-$(CONFIG_STM32_TIMER_CNT)	+= stm32-timer-cnt.o
>  obj-$(CONFIG_STM32_LPTIMER_CNT)	+= stm32-lptimer-cnt.o
>  obj-$(CONFIG_TI_EQEP)		+= ti-eqep.o
> diff --git a/drivers/counter/gpio-pulse-cnt.c b/drivers/counter/gpio-pulse-cnt.c
> new file mode 100644
> index 000000000000..9454345c77ad
> --- /dev/null
> +++ b/drivers/counter/gpio-pulse-cnt.c
> @@ -0,0 +1,244 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Copyright (c) 2021 Pengutronix, Oleksij Rempel <kernel@xxxxxxxxxxxxxx>
> + */
> +
> +#include <linux/counter.h>
> +#include <linux/gpio/consumer.h>
> +#include <linux/interrupt.h>
> +#include <linux/module.h>
> +#include <linux/of_device.h>

why of_device.h?  Probably want mod_devicetable.h instead.

> +#include <linux/platform_device.h>
> +
> +#define GPIO_PULSE_NAME		"gpio-pulse-counter"
> +
> +struct gpio_pulse_priv {
> +	struct counter_device counter;
> +	struct gpio_desc *gpio;
> +	int irq;
> +	bool enabled;
> +	atomic_t count;
> +};
> +
> +static irqreturn_t gpio_pulse_irq_isr(int irq, void *dev_id)
> +{
> +	struct gpio_pulse_priv *priv = dev_id;
> +
> +	if (!priv->enabled)
> +		return IRQ_NONE;
> +
> +	atomic_inc(&priv->count);
> +
> +	return IRQ_HANDLED;
> +}
> +
> +static ssize_t gpio_pulse_count_enable_read(struct counter_device *counter,
> +					    struct counter_count *count,
> +					    void *private, char *buf)
> +{
> +	struct gpio_pulse_priv *priv = counter->priv;
> +
> +	return sysfs_emit(buf, "%d\n", priv->enabled);
> +}
> +
> +static ssize_t gpio_pulse_count_enable_write(struct counter_device *counter,
> +					     struct counter_count *count,
> +					     void *private,
> +					     const char *buf, size_t len)
> +{
> +	struct gpio_pulse_priv *priv = counter->priv;
> +	bool enable;
> +	ssize_t ret;
> +
> +	ret = kstrtobool(buf, &enable);
> +	if (ret)
> +		return ret;
> +
> +	if (priv->enabled == enable)
> +		return len;
> +
> +	priv->enabled = enable;
> +
> +	if (enable)
> +		enable_irq(priv->irq);
> +	else
> +		disable_irq(priv->irq);

As pointed out by Ahmad, this is all racy.
Personally I'd just let the race happen, and don't
bother with the priv->enabled at all.

We aren't going to be dealing with shared interrupts
here so if we get a race, it's between userspace write
getting through to enable / disable the interrupt and
a pulse coming in.  The userspace part is so unpredictable on
timing etc the race isn't pretty meaningless (you might count
one extra, but then if userspace took a bit longer you might
do that anyway).


> +
> +	return len;
> +}
> +
> +static const struct counter_count_ext gpio_pulse_count_ext[] = {
> +	{
> +		.name = "enable",
> +		.read = gpio_pulse_count_enable_read,
> +		.write = gpio_pulse_count_enable_write,
> +	},
> +};
> +
> +static enum counter_synapse_action gpio_pulse_synapse_actions[] = {
> +	COUNTER_SYNAPSE_ACTION_RISING_EDGE,
> +};
> +
> +static int gpio_pulse_action_get(struct counter_device *counter,
> +			    struct counter_count *count,
> +			    struct counter_synapse *synapse,
> +			    size_t *action)
> +{
> +	*action = COUNTER_SYNAPSE_ACTION_RISING_EDGE;
> +
> +	return 0;
> +}
> +
> +static int gpio_pulse_count_read(struct counter_device *counter,
> +				 struct counter_count *count,
> +				 unsigned long *val)
> +{
> +	struct gpio_pulse_priv *priv = counter->priv;
> +
> +	*val = atomic_read(&priv->count);
> +
> +	return 0;
> +}
> +
> +static int gpio_pulse_count_write(struct counter_device *counter,
> +				  struct counter_count *count,
> +				  const unsigned long val)
> +{
> +	struct gpio_pulse_priv *priv = counter->priv;
> +
> +	atomic_set(&priv->count, val);
> +
> +	return 0;
> +}
> +
> +static int gpio_pulse_count_function_get(struct counter_device *counter,
> +					 struct counter_count *count,
> +					 size_t *function)
> +{
> +	*function = COUNTER_COUNT_FUNCTION_INCREASE;
> +
> +	return 0;
> +}
> +
> +static int gpio_pulse_count_signal_read(struct counter_device *counter,
> +					struct counter_signal *signal,
> +					enum counter_signal_value *val)
> +{
> +	struct gpio_pulse_priv *priv = counter->priv;
> +	int ret;
> +
> +	ret = gpiod_get_value(priv->gpio);
> +	if (ret < 0)
> +		return ret;
> +
> +	*val = ret ? COUNTER_SIGNAL_HIGH : COUNTER_SIGNAL_LOW;
> +
> +	return 0;
> +}
> +
> +static const struct counter_ops gpio_pulse_cnt_ops = {
> +	.action_get = gpio_pulse_action_get,
> +	.count_read = gpio_pulse_count_read,
> +	.count_write = gpio_pulse_count_write,
> +	.function_get = gpio_pulse_count_function_get,
> +	.signal_read  = gpio_pulse_count_signal_read,
> +};
> +
> +static struct counter_signal gpio_pulse_signals[] = {
> +	{
> +		.id = 0,
> +		.name = "Channel 0 signal",
> +	},
> +};
> +
> +static struct counter_synapse gpio_pulse_count_synapses[] = {
> +	{
> +		.actions_list = gpio_pulse_synapse_actions,
> +		.num_actions = ARRAY_SIZE(gpio_pulse_synapse_actions),
> +		.signal = &gpio_pulse_signals[0]
> +	},
> +};
> +
> +static enum counter_count_function gpio_pulse_count_functions[] = {
> +	COUNTER_COUNT_FUNCTION_INCREASE,
> +};
> +
> +static struct counter_count gpio_pulse_counts[] = {
> +	{
> +		.id = 0,
> +		.name = "Channel 1 Count",
> +		.functions_list = gpio_pulse_count_functions,
> +		.num_functions = ARRAY_SIZE(gpio_pulse_count_functions),
> +		.synapses = gpio_pulse_count_synapses,
> +		.num_synapses = ARRAY_SIZE(gpio_pulse_count_synapses),
> +		.ext = gpio_pulse_count_ext,
> +		.num_ext = ARRAY_SIZE(gpio_pulse_count_ext),
> +	},
> +};
> +
> +static int gpio_pulse_cnt_probe(struct platform_device *pdev)
> +{
> +	struct device *dev = &pdev->dev;
> +	struct gpio_pulse_priv *priv;
> +	int ret;
> +
> +	priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
> +	if (!priv)
> +		return -ENOMEM;
> +
> +	if (gpiod_count(dev, NULL) != 1) {
> +		dev_err(dev, "Error, need exactly 1 gpio for device\n");
> +		return -EINVAL;
> +	}
> +
> +	priv->gpio = devm_gpiod_get(dev, NULL, GPIOD_IN);
> +	if (IS_ERR(priv->gpio))
> +		return dev_err_probe(dev, PTR_ERR(priv->gpio), "failed to get gpio\n");

My gut feeling here is that it probably makes sense to drop
the direct read of the signal level, and hence allow the driver
to be provided with an interrupt instead of a gpio.

Afterall, not all gpio lines are interrupt capable.

> +
> +	priv->irq = gpiod_to_irq(priv->gpio);
> +	if (priv->irq < 0) {
> +		dev_err(dev, "failed to map GPIO to IRQ: %d\n", priv->irq);
> +		return priv->irq;
> +	}
> +
> +	priv->counter.name = dev_name(dev);
> +	priv->counter.parent = dev;
> +	priv->counter.ops = &gpio_pulse_cnt_ops;
> +	priv->counter.counts = gpio_pulse_counts;
> +	priv->counter.num_counts = ARRAY_SIZE(gpio_pulse_counts);
> +	priv->counter.signals = gpio_pulse_signals;
> +	priv->counter.num_signals = ARRAY_SIZE(gpio_pulse_signals);
> +	priv->counter.priv = priv;
> +
> +	ret = devm_request_irq(dev, priv->irq, gpio_pulse_irq_isr,
> +			       IRQF_TRIGGER_RISING | IRQF_NO_THREAD,
> +			       GPIO_PULSE_NAME, priv);
> +	if (ret)
> +		return ret;
> +
> +	disable_irq(priv->irq);

Race condition here (messy at least) that we can close down.
Note there is ongoing work to try and do this in a nice generic
fashion, but in the meantime call

irq_set_status_flags(priv->irq, IRQ_NOAUTOEN);
before the devm_request_irq()

https://lore.kernel.org/lkml/20210107223926.35284-2-song.bao.hua@xxxxxxxxxxxxx/

There are a lot of cases that series will tidy up, but let us do the best
we can in the meantime!
 
> +
> +	platform_set_drvdata(pdev, priv);
> +
> +	return devm_counter_register(dev, &priv->counter);
> +}
> +
> +static const struct of_device_id gpio_pulse_cnt_of_match[] = {
> +	{ .compatible = "virtual,gpio-pulse-counter", },
> +	{}
> +};
> +MODULE_DEVICE_TABLE(of, gpio_pulse_cnt_of_match);
> +
> +static struct platform_driver gpio_pulse_cnt_driver = {
> +	.probe = gpio_pulse_cnt_probe,
> +	.driver = {
> +		.name = GPIO_PULSE_NAME,
> +		.of_match_table = gpio_pulse_cnt_of_match,
> +	},
> +};
> +module_platform_driver(gpio_pulse_cnt_driver);
> +
> +MODULE_ALIAS("platform:gpio-pulse-counter");
> +MODULE_AUTHOR("Oleksij Rempel <o.rempel@xxxxxxxxxxxxxx>");
> +MODULE_DESCRIPTION("GPIO pulse counter driver");
> +MODULE_LICENSE("GPL v2");