Re: [PATCH v9 13/15] leds: lp55xx: Update the lp55xx to use the multi color framework
From: Dan Murphy
Date: Tue Oct 01 2019 - 08:00:51 EST
Jacek
On 9/30/19 11:31 AM, Dan Murphy wrote:
Jacek
On 9/29/19 8:01 AM, Jacek Anaszewski wrote:
Dan,
On 9/26/19 2:02 PM, Dan Murphy wrote:
Jacek
On 9/25/19 5:00 PM, Jacek Anaszewski wrote:
Dan,
On 9/25/19 7:46 PM, Dan Murphy wrote:
Update the lp5523 to allow the use of the multi color framework.
Signed-off-by: Dan Murphy <dmurphy@xxxxxx>
---
ÂÂ drivers/leds/KconfigÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ |ÂÂ 1 +
ÂÂ drivers/leds/leds-lp5523.cÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ |Â 13 ++
ÂÂ drivers/leds/leds-lp55xx-common.cÂÂÂÂÂÂÂÂ | 150
++++++++++++++++++----
ÂÂ drivers/leds/leds-lp55xx-common.hÂÂÂÂÂÂÂÂ |Â 11 ++
ÂÂ include/linux/platform_data/leds-lp55xx.h |ÂÂ 6 +
ÂÂ 5 files changed, 157 insertions(+), 24 deletions(-)
diff --git a/drivers/leds/Kconfig b/drivers/leds/Kconfig
index 84f60e35c5ee..dc3d9f2194cd 100644
--- a/drivers/leds/Kconfig
+++ b/drivers/leds/Kconfig
@@ -377,6 +377,7 @@ config LEDS_LP50XX
ÂÂ config LEDS_LP55XX_COMMON
ÂÂÂÂÂÂ tristate "Common Driver for TI/National
LP5521/5523/55231/5562/8501"
ÂÂÂÂÂÂ depends on LEDS_LP5521 || LEDS_LP5523 || LEDS_LP5562 ||
LEDS_LP8501
+ÂÂÂ depends on LEDS_CLASS_MULTI_COLOR && OF
ÂÂÂÂÂÂ select FW_LOADER
ÂÂÂÂÂÂ select FW_LOADER_USER_HELPER
ÂÂÂÂÂÂ help
diff --git a/drivers/leds/leds-lp5523.c b/drivers/leds/leds-lp5523.c
index d0b931a136b9..8b2cdb98fed6 100644
--- a/drivers/leds/leds-lp5523.c
+++ b/drivers/leds/leds-lp5523.c
@@ -791,6 +791,18 @@ static ssize_t store_master_fader_leds(struct
device *dev,
ÂÂÂÂÂÂ return ret;
ÂÂ }
ÂÂ +static int lp5523_led_intensity(struct lp55xx_led *led, int
chan_num)
Why do we need this function? brightness op will not suffice?
I looked at this before sending it in. This API adds the chan_num to
write to.
The brightness_fn does not it takes it from the led structure.
+{
+ÂÂÂ struct lp55xx_chip *chip = led->chip;
+ÂÂÂ int ret;
+
+ÂÂÂ mutex_lock(&chip->lock);
+ÂÂÂ ret = lp55xx_write(chip, LP5523_REG_LED_PWM_BASE + chan_num,
+ÂÂÂÂÂÂÂÂÂÂÂÂ led->brightness);
+ÂÂÂ mutex_unlock(&chip->lock);
+ÂÂÂ return ret;
+}
+
ÂÂ static int lp5523_led_brightness(struct lp55xx_led *led)
ÂÂ {
ÂÂÂÂÂÂ struct lp55xx_chip *chip = led->chip;
@@ -857,6 +869,7 @@ static struct lp55xx_device_config lp5523_cfg = {
 .max_channel = LP5523_MAX_LEDS,
ÂÂÂÂÂÂ .post_init_deviceÂÂ = lp5523_post_init_device,
ÂÂÂÂÂÂ .brightness_fnÂÂÂÂÂ = lp5523_led_brightness,
+ÂÂÂ .color_intensity_fn = lp5523_led_intensity,
ÂÂÂÂÂÂ .set_led_currentÂÂÂ = lp5523_set_led_current,
ÂÂÂÂÂÂ .firmware_cbÂÂÂÂÂÂÂ = lp5523_firmware_loaded,
ÂÂÂÂÂÂ .run_engineÂÂÂÂÂÂÂÂ = lp5523_run_engine,
diff --git a/drivers/leds/leds-lp55xx-common.c
b/drivers/leds/leds-lp55xx-common.c
index 44ced02b49f9..0e4b3a9d3047 100644
--- a/drivers/leds/leds-lp55xx-common.c
+++ b/drivers/leds/leds-lp55xx-common.c
@@ -136,9 +136,26 @@ static int lp55xx_set_brightness(struct
led_classdev *cdev,
ÂÂ {
ÂÂÂÂÂÂ struct lp55xx_led *led = cdev_to_lp55xx_led(cdev);
ÂÂÂÂÂÂ struct lp55xx_device_config *cfg = led->chip->cfg;
+ÂÂÂ int brightness_val[LP55XX_MAX_GROUPED_CHAN];
+ÂÂÂ int ret;
+ÂÂÂ int i;
+
+ÂÂÂ if (led->mc_cdev.num_leds > 1) {
+ÂÂÂÂÂÂÂ led_mc_calc_brightness(&led->mc_cdev,
+ÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂÂ brightness, brightness_val);
+ÂÂÂÂÂÂÂ for (i = 0; i < led->mc_cdev.num_leds; i++) {
+ÂÂÂÂÂÂÂÂÂÂÂ led->brightness = brightness_val[i];
+ÂÂÂÂÂÂÂÂÂÂÂ ret = cfg->color_intensity_fn(led,
+ led->grouped_channels[i]);
Now we will have three separate calls for each color component
(and possibly sleeping on mutex on contention).
Probably LED mc class use case will need a bit different design.
Also, instead of grouped_channels we could possibly have
led_mc_get_color_id(&led->mc_dev, i)
color_id and grouped_channels are not a 1:1 mapping
OK, they're channel numbers.
which would map entry position index to color_id.
I will stop reviewing here and will continue after taking
deeper look at this lp55xx design.
I've analyzed that design in greater detail and have started
to wonder why you can't pass two arrays to the new op:
channel numbers and corresponding calculated channel intensities?
OK so I coded this up. And there is not a clean way to do this since
the channels and colors are not correlated between two structures. In
order to do this we would have to expand the grouped_channel and
channel_color arrays in the lp55xx_led to LED_COLOR_ID_MAX and put the
channel and color_component value in the position in the array that is
associated with the color ID. Then the driver can loop through the
available colors from the MC class and get the channel and
color_component information.
I spent the day trying to figure out the cleanest way to do this. What
I did was I added a struct to the MC framework (struct name will
probably need to be changed)
This also solves the problem that you brought up on the caller knowing
what ID and brightness goes to what.
struct led_mc_color_conversion {
ÂÂÂ int color_id;
ÂÂÂ int brightness;
};
For devices that need the color conversion to take place the driver will
pass in an array of structs to be filled out by the MC framework. Once
the MCFW populates the array the device driver can map the color IDs to
the the associated output. So for the LP55xx I was able to map the
color ID to the channel number.
I added a brightness value to the intensity function so now the caller
passes in the channel number and the brightness. This should be all the
child cares about.
I think it is ok that the common code calls the intensity function x
number of times. Otherwise each child will need to have the same code
to de-reference the struct seems a bit error wrought IMO.
Dan
Beyond that in coding this and thinking about the design it is better
to have the lp55xx_common code to do all the heavy lifting and the
children to just perform the action on the device itself. So doing the
loop in the common code like we have now and calling the intensity_fn
based on x number of LEDs. The child subscriber to the common code
should have to worry about the color it should only care about the
channel and the value.
Dan