Re: [PATCH v3 2/5] Input: add driver for Focaltech FTS touchscreen

[Jeff LaBundy]

Hi Joel,

Great work so far! It's coming along nicely. Please find my latest
feedback below.

On Fri, Apr 14, 2023 at 09:02:19PM -0500, Joel Selvaraj wrote:
> The Focaltech FTS driver supports several variants of focaltech
> touchscreens found in ~2018 era smartphones including variants found on
> the PocoPhone F1 and the SHIFT6mq which are already present in mainline.
> This driver is loosely based on the original driver from Focaltech
> but has been simplified and largely reworked.
> 
> Co-developed-by: Caleb Connolly <caleb@xxxxxxxxxxxxx>
> Signed-off-by: Caleb Connolly <caleb@xxxxxxxxxxxxx>
> Signed-off-by: Joel Selvaraj <joelselvaraj.oss@xxxxxxxxx>
> ---
>  MAINTAINERS                                   |   8 +
>  drivers/input/touchscreen/Kconfig             |  12 +
>  drivers/input/touchscreen/Makefile            |   1 +
>  drivers/input/touchscreen/focaltech_fts5452.c | 432 ++++++++++++++++++
>  4 files changed, 453 insertions(+)
>  create mode 100644 drivers/input/touchscreen/focaltech_fts5452.c
> 
> diff --git a/MAINTAINERS b/MAINTAINERS
> index 7ec4ce64f66d..1a3ea61e1f52 100644
> --- a/MAINTAINERS
> +++ b/MAINTAINERS
> @@ -8028,6 +8028,14 @@ L:	linux-input@xxxxxxxxxxxxxxx
>  S:	Maintained
>  F:	drivers/input/joystick/fsia6b.c
>  
> +FOCALTECH FTS5452 TOUCHSCREEN DRIVER
> +M:	Joel Selvaraj <joelselvaraj.oss@xxxxxxxxx>
> +M:	Caleb Connolly <caleb@xxxxxxxxxxxxx>
> +L:	linux-input@xxxxxxxxxxxxxxx
> +S:	Maintained
> +F:	Documentation/devicetree/bindings/input/touchscreen/focaltech,fts5452.yaml
> +F:	drivers/input/touchscreen/focaltech_fts5452.c
> +
>  FOCUSRITE SCARLETT GEN 2/3 MIXER DRIVER
>  M:	Geoffrey D. Bennett <g@xxxxx>
>  L:	alsa-devel@xxxxxxxxxxxxxxxx (moderated for non-subscribers)
> diff --git a/drivers/input/touchscreen/Kconfig b/drivers/input/touchscreen/Kconfig
> index 1feecd7ed3cb..11af91504969 100644
> --- a/drivers/input/touchscreen/Kconfig
> +++ b/drivers/input/touchscreen/Kconfig
> @@ -388,6 +388,18 @@ config TOUCHSCREEN_EXC3000
>  	  To compile this driver as a module, choose M here: the
>  	  module will be called exc3000.
>  
> +config TOUCHSCREEN_FOCALTECH_FTS5452
> +	tristate "Focaltech FTS Touchscreen"
> +	depends on I2C
> +	help
> +	  Say Y here to enable support for I2C connected Focaltech FTS
> +	  based touch panels, including the 5452 and 8917 panels.

This language is a bit misleading, as it seems to suggest three or more
models are supported. It seems the title should simply be "FocalTech
FTS5452 touchscreen controller" with the description as "...FocalTech
FTS5452 and compatible touchscreen controllers."

As more are found to be compatible (e.g. FTS8917), the compatible strings
can simply be appended.

> +
> +	  If unsure, say N.
> +
> +	  To compile this driver as a module, choose M here: the
> +	  module will be called focaltech_fts.
> +
>  config TOUCHSCREEN_FUJITSU
>  	tristate "Fujitsu serial touchscreen"
>  	select SERIO
> diff --git a/drivers/input/touchscreen/Makefile b/drivers/input/touchscreen/Makefile
> index 159cd5136fdb..47d78c9cff21 100644
> --- a/drivers/input/touchscreen/Makefile
> +++ b/drivers/input/touchscreen/Makefile
> @@ -45,6 +45,7 @@ obj-$(CONFIG_TOUCHSCREEN_ELO)		+= elo.o
>  obj-$(CONFIG_TOUCHSCREEN_EGALAX)	+= egalax_ts.o
>  obj-$(CONFIG_TOUCHSCREEN_EGALAX_SERIAL)	+= egalax_ts_serial.o
>  obj-$(CONFIG_TOUCHSCREEN_EXC3000)	+= exc3000.o
> +obj-$(CONFIG_TOUCHSCREEN_FOCALTECH_FTS5452)	+= focaltech_fts5452.o
>  obj-$(CONFIG_TOUCHSCREEN_FUJITSU)	+= fujitsu_ts.o
>  obj-$(CONFIG_TOUCHSCREEN_GOODIX)	+= goodix_ts.o
>  obj-$(CONFIG_TOUCHSCREEN_HIDEEP)	+= hideep.o
> diff --git a/drivers/input/touchscreen/focaltech_fts5452.c b/drivers/input/touchscreen/focaltech_fts5452.c
> new file mode 100644
> index 000000000000..abf8a2f271ca
> --- /dev/null
> +++ b/drivers/input/touchscreen/focaltech_fts5452.c
> @@ -0,0 +1,432 @@
> +// SPDX-License-Identifier: GPL-2.0-only
> +/*
> + * FocalTech touchscreen driver.
> + *
> + * Copyright (c) 2010-2017, FocalTech Systems, Ltd., all rights reserved.
> + * Copyright (C) 2018 XiaoMi, Inc.
> + * Copyright (c) 2021 Caleb Connolly <caleb@xxxxxxxxxxxxx>
> + * Copyright (c) 2023 Joel Selvaraj <joelselvaraj.oss@xxxxxxxxx>
> + */

Nit: inconsistent copyright capitalization.

> +
> +#include <linux/delay.h>
> +#include <linux/i2c.h>
> +#include <linux/input.h>
> +#include <linux/input/mt.h>
> +#include <linux/input/touchscreen.h>
> +#include <linux/interrupt.h>
> +#include <linux/kernel.h>
> +#include <linux/module.h>
> +#include <linux/mutex.h>
> +#include <linux/regmap.h>
> +#include <linux/regulator/consumer.h>
> +#include <linux/slab.h>
> +
> +#define FTS_REG_CHIP_ID_H		0xA3
> +#define FTS_REG_CHIP_ID_L		0x9F
> +
> +#define FTS_MAX_POINTS_SUPPORT		10
> +#define FTS_ONE_TOUCH_LEN		6
> +
> +#define FTS_TOUCH_X_H_OFFSET		3
> +#define FTS_TOUCH_X_L_OFFSET		4
> +#define FTS_TOUCH_Y_H_OFFSET		5
> +#define FTS_TOUCH_Y_L_OFFSET		6
> +#define FTS_TOUCH_PRESSURE_OFFSET	7
> +#define FTS_TOUCH_AREA_OFFSET		8
> +#define FTS_TOUCH_TYPE_OFFSET		3
> +#define FTS_TOUCH_ID_OFFSET		5
> +
> +#define FTS_TOUCH_DOWN			0
> +#define FTS_TOUCH_UP			1
> +#define FTS_TOUCH_CONTACT		2
> +
> +#define FTS_INTERVAL_READ_REG_MS	100
> +#define FTS_TIMEOUT_READ_REG_MS		2000
> +
> +#define FTS_DRIVER_NAME			"fts-i2c"
> +
> +static const u16 fts_chip_types[] = {
> +	0x5452,
> +	0x8719,
> +};
> +
> +struct fts_ts_data {
> +	struct i2c_client *client;
> +	struct input_dev *input_dev;
> +	struct regmap *regmap;
> +	int irq;
> +	struct regulator_bulk_data regulators[2];
> +	u8 max_touch_points;
> +	u8 *point_buf;
> +	int point_buf_size;
> +	struct gpio_desc *reset_gpio;
> +	struct touchscreen_properties prop;
> +};
> +
> +struct fts_i2c_chip_data {
> +	int max_touch_points;
> +};

There is no reason to wrap a single member in a struct; just define an array
and point each driver_data member to the appropriate element.

An even better solution, however, would be to merge the device ID into this.
Then you would have a single array of structs with very clear association
between device ID and number of points.

> +
> +int fts_check_status(struct fts_ts_data *data)

This function can be static. It also seems to be inappropriately named. Here
we are checking the device's ID, not its status.

> +{
> +	int error, i = 0, count = 0;
> +	unsigned int val, id;
> +
> +	do {
> +		error = regmap_read(data->regmap, FTS_REG_CHIP_ID_L, &id);
> +		if (error)
> +			dev_err(&data->client->dev, "I2C read failed: %d\n", error);

If this read fails, there is no point in continuing further in this loop. Most
likely the second read would fail as well but if it doesn't, you are computing
the id using an uninitialized variable.

Can you also explain, and possibly add comments, as to why the device ID must
be checked in a retry loop? Is it because the device may be in a deep sleep and
must be hit with I2C traffic a couple of times?

If so, then you likely want to briefly sleep and then start over (i.e. continue)
in the event of an error.

> +
> +		error = regmap_read(data->regmap, FTS_REG_CHIP_ID_H, &val);
> +		if (error)
> +			dev_err(&data->client->dev, "I2C read failed: %d\n", error);

Same problem here.

> +
> +		id |= val << 8;
> +
> +		for (i = 0; i < ARRAY_SIZE(fts_chip_types); i++)
> +			if (id == fts_chip_types[i])
> +				return 0;

This retry loop in general seems a bit non-optimal. If for example the driver
is simply communicating with an incompatible device, there is no need to go
through all N loops.

> +
> +		count++;
> +		msleep(FTS_INTERVAL_READ_REG_MS);
> +	} while ((count * FTS_INTERVAL_READ_REG_MS) < FTS_TIMEOUT_READ_REG_MS);

This multiplication seems unnecessarily complicated; can we not simply have
FTS_MAX_RETRIES or similar?

> +
> +	return -EIO;
> +}
> +
> +static int fts_report_touch(struct fts_ts_data *data)
> +{
> +	struct input_dev *input_dev = data->input_dev;
> +	int base;
> +	unsigned int x, y, z, maj;
> +	u8 slot, type;
> +	int error, i = 0;
> +
> +	u8 *buf = data->point_buf;
> +
> +	memset(buf, 0, data->point_buf_size);
> +
> +	error = regmap_bulk_read(data->regmap, 0, buf, data->point_buf_size);
> +	if (error) {
> +		dev_err(&data->client->dev, "I2C read failed: %d\n", error);
> +		return error;
> +	}
> +
> +	for (i = 0; i < data->max_touch_points; i++) {
> +		base = FTS_ONE_TOUCH_LEN * i;
> +
> +		slot = buf[base + FTS_TOUCH_ID_OFFSET] >> 4;
> +		if (slot >= data->max_touch_points)
> +			break;
> +
> +		x = ((buf[base + FTS_TOUCH_X_H_OFFSET] & 0x0F) << 8) +
> +		    (buf[base + FTS_TOUCH_X_L_OFFSET] & 0xFF);
> +		y = ((buf[base + FTS_TOUCH_Y_H_OFFSET] & 0x0F) << 8) +
> +		    (buf[base + FTS_TOUCH_Y_L_OFFSET] & 0xFF);

Sorry, I did not quite follow the image that was shared in an earlier thread.
It is unclear to me why we cannot represent the interrupt status registers
as an array of __be16 values and then do something like the following:

		x = be16_to_cpu(buf[FTS_TOUCH_X_OFFSET]) & GENMASK(11, 0);

I would be surprised if the mask is even necessary; you would need to refer
to a datasheet however. Perhaps the vendor would be willing to give one to
you if it means they get an upstream driver?

> +
> +		z = buf[base + FTS_TOUCH_PRESSURE_OFFSET];
> +		if (z == 0)
> +			z = 0x3f;
> +
> +		maj = buf[base + FTS_TOUCH_AREA_OFFSET] >> 4;
> +		if (maj == 0)
> +			maj = 0x09;

I think we need some comments and possibly some #defines to explain what is
happening here.

> +
> +		type = buf[base + FTS_TOUCH_TYPE_OFFSET] >> 6;
> +
> +		input_mt_slot(input_dev, slot);
> +		if (type == FTS_TOUCH_DOWN || type == FTS_TOUCH_CONTACT) {
> +			input_mt_report_slot_state(input_dev, MT_TOOL_FINGER, true);
> +			touchscreen_report_pos(data->input_dev, &data->prop, x, y, true);
> +			input_report_abs(input_dev, ABS_MT_PRESSURE, z);
> +			input_report_abs(input_dev, ABS_MT_TOUCH_MAJOR, maj);
> +			input_report_key(data->input_dev, BTN_TOUCH, 1);
> +		} else {
> +			input_report_key(data->input_dev, BTN_TOUCH, 0);
> +			input_mt_report_slot_inactive(input_dev);
> +		}
> +	}
> +	input_mt_sync_frame(input_dev);
> +	input_sync(input_dev);
> +
> +	return 0;
> +}
> +
> +static irqreturn_t fts_ts_interrupt(int irq, void *dev_id)
> +{
> +	struct fts_ts_data *data = dev_id;
> +
> +	return fts_report_touch(data) ? IRQ_NONE : IRQ_HANDLED;
> +}
> +
> +static void fts_power_off(void *d)
> +{
> +	struct fts_ts_data *data = d;
> +
> +	regulator_bulk_disable(ARRAY_SIZE(data->regulators), data->regulators);
> +}
> +
> +static int fts_start(struct fts_ts_data *data)
> +{
> +	int error;
> +
> +	error = regulator_bulk_enable(ARRAY_SIZE(data->regulators),
> +				      data->regulators);
> +	if (error) {
> +		dev_err(&data->client->dev, "failed to enable regulators\n");
> +		return error;
> +	}
> +
> +	gpiod_set_value_cansleep(data->reset_gpio, 0);
> +	msleep(200);

Same here with respect to comments; what happens during these first 200 ms after
reset is released? Does the interrupt pin toggle several times? 200 ms is also
quite a while to wait each time the input handler opens the device; is it really
necessary?

> +
> +	enable_irq(data->irq);
> +
> +	return 0;
> +}
> +
> +static int fts_stop(struct fts_ts_data *data)
> +{
> +	disable_irq(data->irq);
> +	gpiod_set_value_cansleep(data->reset_gpio, 1);
> +	fts_power_off(data);
> +
> +	return 0;
> +}
> +
> +static int fts_input_open(struct input_dev *dev)
> +{
> +	struct fts_ts_data *data = input_get_drvdata(dev);
> +	int error;
> +
> +	error = fts_start(data);
> +	if (error)
> +		return error;
> +
> +	error = fts_check_status(data);
> +	if (error) {
> +		dev_err(&data->client->dev, "Failed to start or unsupported chip");
> +		return error;
> +	}

It seems unnecessary and wasteful to check the device ID every time the input
handler opens the device. We also don't want to go through all the trouble of
registering the device, only to find out later it wasn't even the right part.

Instead, you should power up the device during probe, validate its ID and then
power it back down.

> +
> +	return 0;
> +}
> +
> +static void fts_input_close(struct input_dev *dev)
> +{
> +	struct fts_ts_data *data = input_get_drvdata(dev);
> +
> +	fts_stop(data);
> +}
> +
> +static int fts_input_init(struct fts_ts_data *data)
> +{
> +	struct device *dev = &data->client->dev;
> +	struct input_dev *input_dev;
> +	int error = 0;

No need to initialize this, only for it to get overwritten later.

> +
> +	input_dev = devm_input_allocate_device(dev);
> +	if (!input_dev)
> +		return -ENOMEM;
> +
> +	data->input_dev = input_dev;
> +
> +	input_dev->name = FTS_DRIVER_NAME;
> +	input_dev->id.bustype = BUS_I2C;
> +	input_dev->dev.parent = dev;
> +	input_dev->open = fts_input_open;
> +	input_dev->close = fts_input_close;
> +	input_set_drvdata(input_dev, data);
> +
> +	input_set_capability(input_dev, EV_ABS, ABS_MT_POSITION_X);
> +	input_set_capability(input_dev, EV_ABS, ABS_MT_POSITION_Y);
> +	input_set_abs_params(input_dev, ABS_MT_TOUCH_MAJOR, 0, 255, 0, 0);
> +	input_set_abs_params(input_dev, ABS_MT_PRESSURE, 0, 255, 0, 0);
> +
> +	touchscreen_parse_properties(input_dev, true, &data->prop);
> +	if (!data->prop.max_x || !data->prop.max_y) {
> +		dev_err(dev,
> +			"touchscreen-size-x and/or touchscreen-size-y not set in device properties\n");

"Device properties" is vague; one could interpret it to mean the controller's
embedded FW. Just cut the message off at "...not set".

> +		return -EINVAL;
> +	}
> +
> +	error = input_mt_init_slots(input_dev, data->max_touch_points,
> +				    INPUT_MT_DIRECT);
> +	if (error)
> +		return error;
> +
> +	data->point_buf_size = (data->max_touch_points * FTS_ONE_TOUCH_LEN) + 3;
> +	data->point_buf = devm_kzalloc(dev, data->point_buf_size, GFP_KERNEL);
> +	if (!data->point_buf)
> +		return -ENOMEM;
> +
> +	error = input_register_device(input_dev);
> +	if (error) {
> +		dev_err(dev, "Failed to register input device\n");
> +		return error;
> +	}
> +
> +	return 0;
> +}
> +
> +static const struct regmap_config fts_ts_i2c_regmap_config = {
> +	.reg_bits = 8,
> +	.val_bits = 8,
> +};
> +
> +static int fts_ts_probe(struct i2c_client *client)
> +{
> +	const struct i2c_device_id *id = i2c_client_get_device_id(client);
> +	const struct fts_i2c_chip_data *chip_data;
> +	struct fts_ts_data *data;
> +	int error = 0;
> +
> +	if (!i2c_check_functionality(client->adapter, I2C_FUNC_I2C)) {
> +		dev_err(&client->dev, "I2C not supported");
> +		return -ENODEV;
> +	}
> +
> +	if (!client->irq) {
> +		dev_err(&client->dev, "No irq specified\n");
> +		return -EINVAL;
> +	}
> +
> +	data = devm_kzalloc(&client->dev, sizeof(*data), GFP_KERNEL);
> +	if (!data)
> +		return -ENOMEM;
> +
> +	chip_data = device_get_match_data(&client->dev);
> +	if (!chip_data)
> +		chip_data = (const struct fts_i2c_chip_data *)id->driver_data;
> +	if (!chip_data || !chip_data->max_touch_points) {
> +		dev_err(&client->dev, "invalid or missing chip data\n");
> +		return -EINVAL;
> +	}
> +	if (chip_data->max_touch_points > FTS_MAX_POINTS_SUPPORT) {
> +		dev_err(&client->dev,
> +			"invalid chip data, max_touch_points should be less than or equal to %d\n",
> +			FTS_MAX_POINTS_SUPPORT);
> +		return -EINVAL;
> +	}

This check is not necessary; if someone adds an invalid max_touch_points, then the
driver was updated incorrectly. There is no need to check it at every runtime.

> +	data->max_touch_points = chip_data->max_touch_points;
> +
> +	data->client = client;
> +	i2c_set_clientdata(client, data);
> +
> +	data->reset_gpio = devm_gpiod_get_optional(&client->dev, "reset", GPIOD_OUT_HIGH);
> +	if (IS_ERR(data->reset_gpio)) {
> +		error = PTR_ERR(data->reset_gpio);
> +		dev_err(&client->dev, "Failed to request reset gpio, error %d\n", error);
> +		return error;
> +	}
> +
> +	data->regmap = devm_regmap_init_i2c(client, &fts_ts_i2c_regmap_config);
> +	if (IS_ERR(data->regmap)) {
> +		error = PTR_ERR(data->regmap);
> +		dev_err(&client->dev, "regmap allocation failed, error %d\n", error);
> +		return error;
> +	}
> +
> +	/*
> +	 * AVDD is the analog voltage supply (2.6V to 3.3V)
> +	 * VDDIO is the digital voltage supply (1.8V)
> +	 */
> +	data->regulators[0].supply = "avdd";
> +	data->regulators[1].supply = "vddio";
> +	error = devm_regulator_bulk_get(&client->dev, ARRAY_SIZE(data->regulators),
> +					data->regulators);
> +	if (error) {
> +		dev_err(&client->dev, "Failed to get regulators %d\n", error);
> +		return error;
> +	}
> +
> +	error = devm_add_action_or_reset(&client->dev, fts_power_off, data);
> +	if (error) {
> +		dev_err(&client->dev, "failed to install power off handler\n");
> +		return error;
> +	}

Christophe makes a great point. If this or any other call throughout the rest of
probe as you have written it fails, you will try to disable a disabled regulator.

The same will happen when the driver is torn down, as the input handler should
have already powered down the device by way of the close callback. Did you build
this driver as a module and test removal? I suspect you will get a stack trace.

I think the call needs to go away altogether.

> +
> +	error = devm_request_threaded_irq(&client->dev, client->irq, NULL,
> +					  fts_ts_interrupt, IRQF_ONESHOT,
> +					  client->name, data);
> +	if (error) {
> +		dev_err(&client->dev, "Failed to request IRQ: %d\n", error);
> +		return error;
> +	}
> +
> +	error = fts_input_init(data);
> +	if (error)
> +		return error;
> +
> +	return 0;

This is idiomatic, but I find "return fts_input_init(data);" to be simpler.

> +}
> +
> +static int fts_pm_suspend(struct device *dev)
> +{
> +	struct fts_ts_data *data = dev_get_drvdata(dev);
> +
> +	mutex_lock(&data->input_dev->mutex);
> +
> +	if (input_device_enabled(data->input_dev))
> +		fts_stop(data);
> +
> +	mutex_unlock(&data->input_dev->mutex);
> +
> +	return 0;
> +}
> +
> +static int fts_pm_resume(struct device *dev)
> +{
> +	struct fts_ts_data *data = dev_get_drvdata(dev);
> +	int error = 0;

Same here, there is no point in initializating this.

> +
> +	mutex_lock(&data->input_dev->mutex);
> +
> +	if (input_device_enabled(data->input_dev))
> +		error = fts_start(data);
> +
> +	mutex_unlock(&data->input_dev->mutex);
> +
> +	return error;
> +}
> +
> +static DEFINE_SIMPLE_DEV_PM_OPS(fts_dev_pm_ops, fts_pm_suspend, fts_pm_resume);
> +
> +static const struct fts_i2c_chip_data fts5452_chip_data = {
> +	.max_touch_points = 5,
> +};
> +
> +static const struct fts_i2c_chip_data fts8719_chip_data = {
> +	.max_touch_points = 10,
> +};
> +
> +static const struct i2c_device_id fts_i2c_id[] = {
> +	{ .name = "fts5452", .driver_data = (long)&fts5452_chip_data },
> +	{ .name = "fts8719", .driver_data = (long)&fts8719_chip_data },
> +	{ /* sentinel */ }
> +};
> +MODULE_DEVICE_TABLE(i2c, fts_i2c_id);
> +
> +static const struct of_device_id fts_of_match[] = {
> +	{ .compatible = "focaltech,fts5452", .data = &fts5452_chip_data },
> +	{ .compatible = "focaltech,fts8719", .data = &fts8719_chip_data },
> +	{ /* sentinel */ }
> +};
> +
> +MODULE_DEVICE_TABLE(of, fts_of_match);
> +
> +static struct i2c_driver fts_ts_driver = {
> +	.probe_new = fts_ts_probe,
> +	.id_table = fts_i2c_id,
> +	.driver = {
> +		.name = FTS_DRIVER_NAME,
> +		.pm = pm_sleep_ptr(&fts_dev_pm_ops),
> +		.of_match_table = fts_of_match,
> +	},
> +};
> +module_i2c_driver(fts_ts_driver);
> +
> +MODULE_AUTHOR("Joel Selvaraj <joelselvaraj.oss@xxxxxxxxx>");
> +MODULE_AUTHOR("Caleb Connolly <caleb@xxxxxxxxxxxxx>");
> +MODULE_DESCRIPTION("Focaltech Touchscreen Driver");

Nit: mixing 'FocalTech' and 'Focaltech' throughout.

> +MODULE_LICENSE("GPL");
> -- 
> 2.40.0
> 

Kind regards,
Jeff LaBundy