Re: [PATCH 4/4] Input: goodix - Add pen support

[Hans de Goede <hdegoede@redhat.com>]

Hi Dmitry,

Thank you for the review.

On 12/7/21 18:50, Dmitry Torokhov wrote:
> On Mon, Dec 06, 2021 at 05:47:47PM +0100, Hans de Goede wrote:
>> Some Goodix touchscreens have support for a (Goodix) active pen, add
>> support for this. The info on how to detect when a pen is down and to
>> detect when the stylus buttons are pressed was lifted from the out
>> of tree Goodix driver with pen support written by Adya:
>> https://gitlab.com/AdyaAdya/goodix-touchscreen-linux-driver/
>>
>> Since there is no way to tell if pen support is present, the registering
>> of the pen input_dev is delayed till the first pen event is detected.
>>
>> This has been tested on a Trekstor Surftab duo W1, a Chuwi Hi13 and
>> a Cyberbook T116 tablet.
>>
>> Link: https://gitlab.com/AdyaAdya/goodix-touchscreen-linux-driver/
>> BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=202161
>> BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=204513
>> Signed-off-by: Hans de Goede <hdegoede@redhat.com>
>> ---
>>  drivers/input/touchscreen/goodix.c | 122 ++++++++++++++++++++++++++++-
>>  drivers/input/touchscreen/goodix.h |   1 +
>>  2 files changed, 121 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/input/touchscreen/goodix.c b/drivers/input/touchscreen/goodix.c
>> index 2d38a941e7e4..691e4505cf4a 100644
>> --- a/drivers/input/touchscreen/goodix.c
>> +++ b/drivers/input/touchscreen/goodix.c
>> @@ -298,6 +298,107 @@ static int goodix_ts_read_input_report(struct goodix_ts_data *ts, u8 *data)
>>  	return -ENOMSG;
>>  }
>>  
>> +static struct input_dev *goodix_create_pen_input(struct goodix_ts_data *ts)
>> +{
>> +	struct device *dev = &ts->client->dev;
>> +	struct input_dev *input;
>> +
>> +	input = devm_input_allocate_device(dev);
>> +	if (!input)
>> +		return NULL;
>> +
>> +	input_alloc_absinfo(input);
>> +	if (!input->absinfo) {
>> +		input_free_device(input);
>> +		return NULL;
>> +	}
> 
> Please drop this as input_abs_set_max() will do allocation and
> input_register_device() will reject device with ABS_* events without
> absinfo allocated.
> 
>> +
>> +	input->absinfo[ABS_X] = ts->input_dev->absinfo[ABS_MT_POSITION_X];
> 
> 	input_abs_set_max(input, ABS_X,
> 		input_abs_get_max(input, ABS_MT_POSITION_X);
> 
> or even maybe
> 
> 	input_set_abs_params(input, ABS_X,
> 		0, input_abs_get_max(input, ABS_MT_POSITION_X), 0, 0);

The reason why I'm just copying the entire absinfo struct
(and thus need the NULL check above) is because this driver uses
touchscreen_parse_properties(), so the min and fuzz values
might (theoretically) also be set through device-properties and
I wanted to cover that.

Since you don't like the above approach, I will go with the following
for the next version:

	input_set_abs_params(input, ABS_X,
			     input_abs_get_min(ts->input_dev, ABS_MT_POSITION_X),
			     input_abs_get_max(ts->input_dev, ABS_MT_POSITION_X),
			     input_abs_get_fuzz(ts->input_dev, ABS_MT_POSITION_X),
			     input_abs_get_flat(ts->input_dev, ABS_MT_POSITION_X));

(and the same for the Y axis).


> 
> 
>> +	input->absinfo[ABS_Y] = ts->input_dev->absinfo[ABS_MT_POSITION_Y];
>> +	__set_bit(ABS_X, input->absbit);
>> +	__set_bit(ABS_Y, input->absbit);
> 
> This might not be needed, depending...

Ack, will drop.

> 
>> +	input_set_abs_params(input, ABS_PRESSURE, 0, 255, 0, 0);
>> +
>> +	input_set_capability(input, EV_KEY, BTN_TOUCH);
>> +	input_set_capability(input, EV_KEY, BTN_TOOL_PEN);
>> +	input_set_capability(input, EV_KEY, BTN_STYLUS);
>> +	input_set_capability(input, EV_KEY, BTN_STYLUS2);
>> +	__set_bit(INPUT_PROP_DIRECT, input->propbit);
>> +	/*
>> +	 * The resolution of these touchscreens is about 10 units/mm, the actual
>> +	 * resolution does not matter much since we set INPUT_PROP_DIRECT.
>> +	 * Userspace wants something here though, so just set it to 10 units/mm.
>> +	 */
>> +	input_abs_set_res(input, ABS_X, 10);
>> +	input_abs_set_res(input, ABS_Y, 10);
> 
> Could it be moved next to setting up axes?

Ack, will do.

> 
>> +
>> +	input->name = "Goodix Active Pen";
>> +	input->phys = "input/pen";
>> +	input->id.bustype = BUS_I2C;
>> +	if (kstrtou16(ts->id, 10, &input->id.product))
>> +		input->id.product = 0x1001;
>> +	input->id.version = ts->version;
>> +
>> +	if (input_register_device(input) != 0) {
>> +		input_free_device(input);
> 
> Warrants a comment on why we need to free devm.
> 
> Is it going to be safely destroyed on removal? It is likely to happen
> very first thing, before we deal with interrupts, etc.

That is a very good point, since I've chosen to create the pen
input device on the first pen event, it happens after the interrupts
have been registered. So that means the driver can no longer rely
on the devm reverse-teardown order to get things right.

For the next version I will always allocate the device and only
register it on the fly. Since the devres is created when allocating
this will fix this ordering problem. The device will then still
unregister before the irq gets disabled, but it will not be
free-ed, so the irq can still safely use it.

>> +		return NULL;
>> +	}
>> +
>> +	return input;
>> +}
>> +
>> +static void goodix_ts_report_pen_down(struct goodix_ts_data *ts, u8 *data)
>> +{
>> +	int input_x, input_y, input_w;
>> +	u8 key_value;
>> +
>> +	if (!ts->input_pen) {
>> +		ts->input_pen = goodix_create_pen_input(ts);
>> +		if (!ts->input_pen)
>> +			return;
>> +	}
>> +
>> +	if (ts->contact_size == 9) {
>> +		input_x = get_unaligned_le16(&data[4]);
>> +		input_y = get_unaligned_le16(&data[6]);
>> +		input_w = get_unaligned_le16(&data[8]);
>> +	} else {
>> +		input_x = get_unaligned_le16(&data[2]);
>> +		input_y = get_unaligned_le16(&data[4]);
>> +		input_w = get_unaligned_le16(&data[6]);
>> +	}
>> +
>> +	touchscreen_report_pos(ts->input_pen, &ts->prop, input_x, input_y, false);
>> +	input_report_abs(ts->input_pen, ABS_PRESSURE, input_w);
>> +
>> +	input_report_key(ts->input_pen, BTN_TOUCH, 1);
>> +	input_report_key(ts->input_pen, BTN_TOOL_PEN, 1);
>> +
>> +	if (data[0] & GOODIX_HAVE_KEY) {
>> +		key_value = data[1 + ts->contact_size];
>> +		input_report_key(ts->input_pen, BTN_STYLUS, key_value & 0x10);
>> +		input_report_key(ts->input_pen, BTN_STYLUS2, key_value & 0x20);
> 
> Use BIT?

Ack.

> 
>> +	} else {
>> +		input_report_key(ts->input_pen, BTN_STYLUS, 0);
>> +		input_report_key(ts->input_pen, BTN_STYLUS2, 0);
>> +	}
>> +
>> +	input_sync(ts->input_pen);
>> +}
>> +
>> +static void goodix_ts_report_pen_up(struct goodix_ts_data *ts)
>> +{
>> +	if (!ts->input_pen)
>> +		return;
>> +
>> +	input_report_key(ts->input_pen, BTN_TOUCH, 0);
>> +	input_report_key(ts->input_pen, BTN_TOOL_PEN, 0);
>> +	input_report_key(ts->input_pen, BTN_STYLUS, 0);
>> +	input_report_key(ts->input_pen, BTN_STYLUS2, 0);
>> +
>> +	input_sync(ts->input_pen);
>> +}
>> +
>>  static void goodix_ts_report_touch_8b(struct goodix_ts_data *ts, u8 *coor_data)
>>  {
>>  	int id = coor_data[0] & 0x0F;
>> @@ -328,6 +429,14 @@ static void goodix_ts_report_touch_9b(struct goodix_ts_data *ts, u8 *coor_data)
>>  	input_report_abs(ts->input_dev, ABS_MT_WIDTH_MAJOR, input_w);
>>  }
>>  
>> +static void goodix_ts_release_keys(struct goodix_ts_data *ts)
>> +{
>> +	int i;
>> +
>> +	for (i = 0; i < GOODIX_MAX_KEYS; i++)
>> +		input_report_key(ts->input_dev, ts->keymap[i], 0);
>> +}
>> +
>>  static void goodix_ts_report_key(struct goodix_ts_data *ts, u8 *data)
>>  {
>>  	int touch_num;
>> @@ -342,8 +451,7 @@ static void goodix_ts_report_key(struct goodix_ts_data *ts, u8 *data)
>>  				input_report_key(ts->input_dev,
>>  						 ts->keymap[i], 1);
>>  	} else {
>> -		for (i = 0; i < GOODIX_MAX_KEYS; i++)
>> -			input_report_key(ts->input_dev, ts->keymap[i], 0);
>> +		goodix_ts_release_keys(ts);
>>  	}
>>  }
>>  
>> @@ -365,6 +473,15 @@ static void goodix_process_events(struct goodix_ts_data *ts)
>>  	if (touch_num < 0)
>>  		return;
>>  
>> +	/* The pen being down is always reported as a single touch */
>> +	if (touch_num == 1 && (point_data[1] & 0x80)) {
>> +		goodix_ts_report_pen_down(ts, point_data);
>> +		goodix_ts_release_keys(ts);
>> +		goto sync; /* Release any previousle registered touches */
>> +	} else {
> 
> Not sure why we need else with goto...

It is not needed, I wrote it this way for readability, one is the
pen down path and the other the pen up path, so having that as
an if ... else ... feels more natural to me.

I plan to keep this for the next version, but if you strongly dislike this
let me know and/or feel free to modify this when you merge the next version.

Regards,

Hans


> 
>> +		goodix_ts_report_pen_up(ts);
>> +	}
>> +
>>  	goodix_ts_report_key(ts, point_data);
>>  
>>  	for (i = 0; i < touch_num; i++)
>> @@ -375,6 +492,7 @@ static void goodix_process_events(struct goodix_ts_data *ts)
>>  			goodix_ts_report_touch_8b(ts,
>>  				&point_data[1 + ts->contact_size * i]);
>>  
>> +sync:
>>  	input_mt_sync_frame(ts->input_dev);
>>  	input_sync(ts->input_dev);
>>  }
>> diff --git a/drivers/input/touchscreen/goodix.h b/drivers/input/touchscreen/goodix.h
>> index 02065d1c3263..fa8602e78a64 100644
>> --- a/drivers/input/touchscreen/goodix.h
>> +++ b/drivers/input/touchscreen/goodix.h
>> @@ -76,6 +76,7 @@ struct goodix_chip_data {
>>  struct goodix_ts_data {
>>  	struct i2c_client *client;
>>  	struct input_dev *input_dev;
>> +	struct input_dev *input_pen;
>>  	const struct goodix_chip_data *chip;
>>  	const char *firmware_name;
>>  	struct touchscreen_properties prop;
>> -- 
>> 2.33.1
>>
> 
> Thanks.
>