backlight: add ktd3137 support

From: Roman Vivchar <[email protected]>

Kinetic KTD3137 is an LED IC, used in various mobile devices for the display backlight. Add devicetree bindings documentation for this chip.

Datasheet: https://www.kinet-ic.com/uploads/KTD3137-04a.pdf

Signed-off-by Roman Vivchar <[email protected]>

---
 .../bindings/leds/backlight/kinetic,ktd3137.yaml   | 64 ++++++++++++++++++++++
 MAINTAINERS                                        |  5 ++
 2 files changed, 69 insertions(+)

Documentation/devicetree/bindings/leds/backlight/kinetic,ktd3137.yaml new file +15 -0

new file mode 100644
index 000000000000..140615640a9a
@@ -0,0 +1,64 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/leds/backlight/kinetic,ktd3137.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Kinetic Technologies KTD3137 LED Backlight
+
+maintainers:
+  - Roman Vivchar <[email protected]>
+
+description:
+  The Kinetic Technologies KTD3137 is an I2C LED backlight driver for medium
+  to large size LCD panels.
<snip>
+  led-max-microamp:

Quoted 39 lines · On Mon, Nov 02, 2026 at 09:58:48PM +0200, Roman Vivchar via B4 Relay wrote:

From: Roman Vivchar <[email protected]>

Kinetic KTD3137 is an LED IC, used in various mobile devices for the display backlight. Add devicetree bindings documentation for this chip.

Datasheet: https://www.kinet-ic.com/uploads/KTD3137-04a.pdf

Signed-off-by Roman Vivchar <[email protected]>

---
 .../bindings/leds/backlight/kinetic,ktd3137.yaml   | 64 ++++++++++++++++++++++
 MAINTAINERS                                        |  5 ++
 2 files changed, 69 insertions(+)

Documentation/devicetree/bindings/leds/backlight/kinetic,ktd3137.yaml new file +19 -0

new file mode 100644
index 000000000000..140615640a9a
@@ -0,0 +1,64 @@
+# SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/leds/backlight/kinetic,ktd3137.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: Kinetic Technologies KTD3137 LED Backlight
+
+maintainers:
+  - Roman Vivchar <[email protected]>
+
+description:
+  The Kinetic Technologies KTD3137 is an I2C LED backlight driver for medium
+  to large size LCD panels.
<snip>
+  led-max-microamp:
+    minimum: 5000
+    maximum: 29800
+    default: 20200
+

Documentation/devicetree/bindings/leds/backlight/kinetic,ktd3137.yaml line 44

Quoted 6 lines

Documentation/devicetree/bindings/leds/backlight/kinetic,ktd3137.yaml new file +6 -0

+  vin-supply: true
+
+required:
+  - compatible
+  - reg
+  - vin-supply

Why is vin-supply required?

Daniel.

Hi Daniel,

Quoted 15 lines · Daniel Thompson

Documentation/devicetree/bindings/leds/backlight/kinetic,ktd3137.yaml line 44

Quoted 11 lines · On Mon, Nov 02, 2026 at 09:58:48PM +0200, Roman Vivchar via B4 Relay wrote:

Documentation/devicetree/bindings/leds/backlight/kinetic,ktd3137.yaml new file +11 -0

+  led-max-microamp:
+    minimum: 5000
+    maximum: 29800
+    default: 20200
+
+  vin-supply: true
+
+required:
+  - compatible
+  - reg
+  - vin-supply

Why is vin-supply required?

The chip cannot function without VIN power supply. Even if it's on the always-on power rail on certain board, there's always a chance it might be accidentally left disabled on the other. It's better to document supply in the bindings as far as I understand.

Quoted 19 lines · Roman Vivchar

Hi Daniel,

Quoted 15 lines · Daniel Thompson

Documentation/devicetree/bindings/leds/backlight/kinetic,ktd3137.yaml line 44

Quoted 11 lines · On Mon, Nov 02, 2026 at 09:58:48PM +0200, Roman Vivchar via B4 Relay wrote:

Documentation/devicetree/bindings/leds/backlight/kinetic,ktd3137.yaml new file +11 -0

+  led-max-microamp:
+    minimum: 5000
+    maximum: 29800
+    default: 20200
+
+  vin-supply: true
+
+required:
+  - compatible
+  - reg
+  - vin-supply

Why is vin-supply required?

Quoted 4 lines

The chip cannot function without VIN power supply. Even if it's on the always-on power rail on certain board, there's always a chance it might be accidentally left disabled on the other. It's better to document supply in the bindings as far as I understand.

You haven't just documented the binding, you have also made the property mandatory by adding it to required.

So what I am asking is why is this property listed as required. You didn't implement the driver in a way that makes the supply mandatory.

Daniel.

Quoted 25 lines · Daniel Thompson

Quoted 19 lines · Roman Vivchar

Hi Daniel,

Quoted 15 lines · Daniel Thompson

Documentation/devicetree/bindings/leds/backlight/kinetic,ktd3137.yaml line 44

Quoted 11 lines · On Mon, Nov 02, 2026 at 09:58:48PM +0200, Roman Vivchar via B4 Relay wrote:

Documentation/devicetree/bindings/leds/backlight/kinetic,ktd3137.yaml new file +11 -0

+  led-max-microamp:
+    minimum: 5000
+    maximum: 29800
+    default: 20200
+
+  vin-supply: true
+
+required:
+  - compatible
+  - reg
+  - vin-supply

Why is vin-supply required?

Quoted 4 lines

The chip cannot function without VIN power supply. Even if it's on the always-on power rail on certain board, there's always a chance it might be accidentally left disabled on the other. It's better to document supply in the bindings as far as I understand.

Quoted 5 lines

You haven't just documented the binding, you have also made the property mandatory by adding it to required.

So what I am asking is why is this property listed as required. You didn't implement the driver in a way that makes the supply mandatory.

Not sure I follow.

In the driver I've used devm_regulator_get_enable(), rather than _optional(), since my understanding is that the non-optional helper is a standard kernel API for required supplies (with regulator core warning and fallback to dummy regulator in case DT omits it).

Do you suggest dropping vin-supply from required: ? On my board it's connected to an always-on supply, though I can't tell about others. I can proceed with this if the DT maintainers are okay with that.

Quoted 5 lines · Roman Vivchar

In the driver I've used devm_regulator_get_enable(), rather than _optional(), since my understanding is that the non-optional helper is a standard kernel API for required supplies (with regulator core warning and fallback to dummy regulator in case DT omits it).

That dummy fallback essentially makes it optional. On a board with an always-on supply then there is no need to include vin-supply in the DTB. The driver will work just fine without it.

Quoted 7 lines · Daniel Thompson

Quoted 5 lines · Roman Vivchar

In the driver I've used devm_regulator_get_enable(), rather than _optional(), since my understanding is that the non-optional helper is a standard kernel API for required supplies (with regulator core warning and fallback to dummy regulator in case DT omits it).

Quoted 3 lines

That dummy fallback essentially makes it optional. On a board with an always-on supply then there is no need to include vin-supply in the DTB. The driver will work just fine without it.

There is a need, because this device needs a supply. How you describe that supply in DTS is a different story - can be always-on, can be missing etc.

Documentation/devicetree/bindings/leds/backlight/kinetic,ktd3137.yaml line 44

Quoted 12 lines · On Mon, Nov 02, 2026 at 09:58:48PM +0200, Roman Vivchar via B4 Relay wrote:

<snip>

Documentation/devicetree/bindings/leds/backlight/kinetic,ktd3137.yaml new file +11 -0

+  led-max-microamp:
+    minimum: 5000
+    maximum: 29800
+    default: 20200
+
+  vin-supply: true
+
+required:
+  - compatible
+  - reg
+  - vin-supply

Why is vin-supply required?

From: Roman Vivchar <[email protected]>

Add support for the KTD3137 display backlight IC. It's used in various Xiaomi phones, such as Xiaomi Redmi Note 9.

Signed-off-by Roman Vivchar <[email protected]>

---
 MAINTAINERS                       |   1 +
 drivers/video/backlight/Kconfig   |  11 ++
 drivers/video/backlight/Makefile  |   1 +
 drivers/video/backlight/ktd3137.c | 262 ++++++++++++++++++++++++++++++++++++++
 4 files changed, 275 insertions(+)

MAINTAINERS +1 -0

index 24dd5b43f167..23114fd09445 100644
@@ -14622,6 +14622,7 @@ KTD3137 BACKLIGHT DRIVER
 M:	Roman Vivchar <[email protected]>
 S:	Maintained
 F:	Documentation/devicetree/bindings/leds/backlight/kinetic,ktd3137.yaml
+F:	drivers/video/backlight/ktd3137.c
 
 KTEST
 M:	Steven Rostedt <[email protected]>

drivers/video/backlight/Kconfig +11 -0

index 7c66b8840d88..0196b8597367 100644
@@ -199,6 +199,17 @@ config BACKLIGHT_KTD2801
 	  Say Y to enable the backlight driver for the Kinetic KTD2801 1-wire
 	  GPIO-controlled backlight found in Samsung Galaxy Core Prime VE LTE.
 
+config BACKLIGHT_KTD3137
+	tristate "Backlight Driver for Kinetic KTD3137"
+	depends on I2C
+	select REGMAP_I2C
+	help
+	  Say Y to enable the backlight driver for the Kinetic KTD3137
+	  found in various Xiaomi smartphones.
+
+	  To compile this driver as a module, choose M here: the module
+	  will be called ktd3137.
+
 config BACKLIGHT_KTZ8866
 	tristate "Backlight Driver for Kinetic KTZ8866"
 	depends on I2C

drivers/video/backlight/Makefile +1 -0

index 34469711c6cd..b2d3b69e9e32 100644
@@ -38,6 +38,7 @@ obj-$(CONFIG_BACKLIGHT_HP700)		+= jornada720_bl.o
 obj-$(CONFIG_BACKLIGHT_IPAQ_MICRO)	+= ipaq_micro_bl.o
 obj-$(CONFIG_BACKLIGHT_KTD253)		+= ktd253-backlight.o
 obj-$(CONFIG_BACKLIGHT_KTD2801)		+= ktd2801-backlight.o
+obj-$(CONFIG_BACKLIGHT_KTD3137)		+= ktd3137.o
 obj-$(CONFIG_BACKLIGHT_KTZ8866)		+= ktz8866.o
 obj-$(CONFIG_BACKLIGHT_LM3509)		+= lm3509_bl.o
 obj-$(CONFIG_BACKLIGHT_LM3533)		+= lm3533_bl.o

drivers/video/backlight/ktd3137.c new file +0 -0

new file mode 100644
index 000000000000..a3ff334b8d46

drivers/video/backlight/ktd3137.c new file +59 -0

+	if (ret)
+		return ret;
+
+	ret = regmap_write(ktd->regmap, KTD3137_REG_PWM, KTD3137_PWM_DEFAULT);
+	if (ret)
+		return ret;
+
+	ret = regmap_write(ktd->regmap, KTD3137_REG_RAMP_ON, KTD3137_RAMP_DEFAULT);
+	if (ret)
+		return ret;
+
+	ret = regmap_write(ktd->regmap, KTD3137_REG_TRANS_RAMP, 0x00);
+	if (ret)
+		return ret;
+
+	return regmap_update_bits(ktd->regmap, KTD3137_REG_MODE,
+				  KTD3137_MODE_CURRENT_MASK, ktd->led_current);
+}
<snip>
+static int ktd3137_probe(struct i2c_client *client)
+{
+	struct backlight_device *backlight_dev;
+	struct backlight_properties props = { };
+	struct ktd3137 *ktd;
+	u32 val;
+	int ret;
+
+	ktd = devm_kzalloc(&client->dev, sizeof(*ktd), GFP_KERNEL);
+	if (!ktd)
+		return -ENOMEM;
+
+	ktd->dev = &client->dev;
+
+	ret = devm_regulator_get_enable(ktd->dev, "vin");
+	if (ret)
+		return dev_err_probe(ktd->dev, ret, "failed to power up\n");
+
+	ktd->enable_gpio = devm_gpiod_get_optional(ktd->dev, "enable",
+						   GPIOD_OUT_HIGH);
+	if (IS_ERR(ktd->enable_gpio))
+		return dev_err_probe(ktd->dev, PTR_ERR(ktd->enable_gpio),
+				     "failed to get enable gpio\n");
+
+	/* HWEN should stay HIGH for at least 150us before I2C can be used */
+	fsleep(150);
+
+	ktd->regmap = devm_regmap_init_i2c(client, &ktd3137_regmap_config);
+	if (IS_ERR(ktd->regmap))
+		return dev_err_probe(ktd->dev, PTR_ERR(ktd->regmap),
+				     "failed to init regmap\n");
+
+	ret = regmap_read(ktd->regmap, KTD3137_REG_DEV_ID, &val);
+	if (ret)
+		return dev_err_probe(ktd->dev, ret, "failed to read chip id\n");
+
+	if (val != KTD3137_DEV_ID_A && val != KTD3137_DEV_ID_B)
+		return dev_err_probe(ktd->dev, -ENODEV,
+				     "unsupported chip id: 0x%02x\n", val);
+
+	ret = ktd3137_parse_dt(ktd);

drivers/video/backlight/ktd3137.c new file +4 -0

+
+	i2c_set_clientdata(client, backlight_dev);
+	return backlight_update_status(backlight_dev);
+}

+}

Quoted 4 lines · On Mon, Nov 02, 2026 at 09:58:49PM +0200, Roman Vivchar via B4 Relay wrote:

From: Roman Vivchar <[email protected]>

Add support for the KTD3137 display backlight IC. It's used in various Xiaomi phones, such as Xiaomi Redmi Note 9.

drivers/video/backlight/ktd3137.c line 95

Quoted 6 lines

drivers/video/backlight/ktd3137.c new file +6 -0

+static int ktd3137_init(struct ktd3137 *ktd)
+{
+	u32 val = KTD3137_CONTROL_DEFAULT | KTD3137_CONTROL_LINEAR;
+	int ret;
+
+	ret = regmap_write(ktd->regmap, KTD3137_REG_CONTROL, val);

Nitpicking but val seems a bit pointless here.

Quoted 8 lines

drivers/video/backlight/ktd3137.c new file +6 -0

+	if (ret)
+		return ret;
+
+	ret = regmap_write(ktd->regmap, KTD3137_REG_PWM, KTD3137_PWM_DEFAULT);
+	if (ret)
+		return ret;

drivers/video/backlight/ktd3137.c line 210

Quoted 4 lines

drivers/video/backlight/ktd3137.c new file +4 -0

+
+	props.type = BACKLIGHT_RAW;
+	props.max_brightness = KTD3137_MAX_BRIGHTNESS;
+	props.brightness = KTD3137_DEFAULT_BRIGHTNESS;

No need for KTD3137_DEFAULT_BRIGHTNESS. Just use KTD3137_MAX_BRIGHTNESS.

drivers/video/backlight/ktd3137.c line 221

Quoted 13 lines

drivers/video/backlight/ktd3137.c new file +11 -0

+	props.scale = BACKLIGHT_SCALE_LINEAR;
+	device_property_read_u32(ktd->dev, "default-brightness", &props.brightness);
+
+	backlight_dev = devm_backlight_device_register(ktd->dev,
+						       "ktd3137-backlight",
+						       ktd->dev, ktd,
+						       &ktd3137_backlight_ops,
+						       &props);
+	if (IS_ERR(backlight_dev))
+		return dev_err_probe(ktd->dev, PTR_ERR(backlight_dev),
+				     "failed to register backlight device\n");

dev_err_cast_probe()

Quoted 6 lines

drivers/video/backlight/ktd3137.c new file +4 -0

+
+	i2c_set_clientdata(client, backlight_dev);
+	return backlight_update_status(backlight_dev);
+}

drivers/video/backlight/ktd3137.c line 235

Quoted 10 lines

drivers/video/backlight/ktd3137.c new file +10 -0

+
+static void ktd3137_remove(struct i2c_client *client)
+{
+	struct backlight_device *backlight_dev = i2c_get_clientdata(client);
+	struct ktd3137 *ktd = bl_get_data(backlight_dev);
+
+	backlight_dev->props.brightness = 0;
+	backlight_update_status(backlight_dev);
+
+	gpiod_set_value_cansleep(ktd->enable_gpio, 0);

Is this needed?

Quoted 2 lines

+}

Daniel.

Quoted 12 lines · Daniel Thompson

drivers/video/backlight/ktd3137.c line 235

Quoted 9 lines · On Mon, Nov 02, 2026 at 09:58:49PM +0200, Roman Vivchar via B4 Relay wrote:

drivers/video/backlight/ktd3137.c new file +9 -0

+static void ktd3137_remove(struct i2c_client *client)
+{
+	struct backlight_device *backlight_dev = i2c_get_clientdata(client);
+	struct ktd3137 *ktd = bl_get_data(backlight_dev);
+
+	backlight_dev->props.brightness = 0;
+	backlight_update_status(backlight_dev);
+
+	gpiod_set_value_cansleep(ktd->enable_gpio, 0);

Is this needed?

The datasheet says "When HWEN is set high, ... if SCL/SDA’s pull up voltage is much less than VIN voltage, it can cause a small leakage current ..., there will be around 7uA additional leakage current from VIN in this standby mode."

It's better to put chip into shutdown mode, which won't cause power leakage according to the datasheet, and ensures that I2C registers will be reset in case chip needs to be cleanly re-initialized.