[PATCH v6 0/2] pwm: add support for NXPs high-side switch MC33XS2410

[PATCH v6 0/2] pwm: add support for NXPs high-side switch MC33XS2410

[Dimitri Fedrau]

The MC33XS2410 is a four channel high-side switch. Featuring advanced
monitoring and control function, the device is operational from 3.0 V to
60 V. The device is controlled by SPI port for configuration.

Changes in V2:
  - fix title in devicetree binding
  - fix commit message in devicetree binding patch
  - remove external clock from pwms and create clocks property
  - switch to unevaluatedProperties: false
  - add missing properties for complete example:
    - pwm-names
    - pwms
    - interrupts
    - clocks

Changes in V3:
  - Add description of the general behaviour of the device (limitations)
  - Drop unused defines
  - Add ranges comments for defines with parameters
  - Drop MC33XS2410_PERIOD_MAX, MC33XS2410_PERIOD_MIN defines
  - Drop mc33xs2410_period variable
  - Round down when calculating period and duty cycle
  - Use switch instead of loop when calculating frequency
  - Removed ret variable in mc33xs2410_pwm_get_freq
  - Handle all accesses in a single call to spi_sync_transfer
  - Fix comments in function mc33xs2410_pwm_get_period
  - Fix call pwm_set_relative_duty_cycle(state, ret, 255), instead
    pwm_set_relative_duty_cycle(state, val[1] + 1, 256);
  - Use devm_pwmchip_alloc
  - Fix typo s/Transitition/Transition/
  - Drop driver_data
  - Removed patch for direct inputs from series
  - Tested with PWM_DEBUG enabled, didn't before !

Changes in V4:
  - removed include of math.h, already included in math64.h
  - removed include of mutex.h, no mutexes are used
  - added include of bitfield.h(FIELD_GET, FIELD_PREP), fixes errors
    discovered by kernel test robot
  - cast parameters in DIV_ROUND_UP to u32, fixes errors discovered by
    kernel test robot

Changes in V5:
  - Fix comment in mc33xs2410_pwm_get_freq, selecting steps instead of
    period
  - Add comment in mc33xs2410_pwm_get_relative_duty_cycle that duty_cycle
    cannot overflow and and period is not zero, guaranteed by the caller
  - Hardware emits a low level when disabled, disable if duty_cycle = 0 is
    requested.
  - Fix complaints when PWM_DEBUG enabled, round-down division in
    mc33xs2410_pwm_apply and round-up in mc33xs2410_pwm_get_state.
  - Add comment for disabling watchdog in probe

Changes in V6:
  - Add link to manual
  - Redefine MC33XS2410_GLB_CTRL_MODE_MASK as MC33XS2410_GLB_CTRL_MODE and
    MC33XS2410_GLB_CTRL_NORMAL_MODE as MC33XS2410_GLB_CTRL_MODE_NORMAL
  - Remove define MC33XS2410_MIN_PERIOD_STEP(x) as there is no need for
    parameters and use instead MC33XS2410_MIN_PERIOD
  - Rename function to_pwm_mc33xs2410_chip to mc33xs2410_from_chip
  - Add comment in mc33xs2410_pwm_get_freq why count should be rounded up
  - Fix incorrect comment in mc33xs2410_pwm_get_period
  - Rename steps variable to doubled_steps in function
    mc33xs2410_pwm_get_freq
  - remove cast in mc33xs2410_pwm_set_relative_duty_cycle
  - remove duty_cycle from if in mc33xs2410_pwm_set_relative_duty_cycle

Dimitri Fedrau (2):
  dt-bindings: pwm: add support for MC33XS2410
  pwm: add support for NXPs high-side switch MC33XS2410

 .../bindings/pwm/nxp,mc33xs2410.yaml          | 118 +++++
 drivers/pwm/Kconfig                           |  12 +
 drivers/pwm/Makefile                          |   1 +
 drivers/pwm/pwm-mc33xs2410.c                  | 422 ++++++++++++++++++
 4 files changed, 553 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/pwm/nxp,mc33xs2410.yaml
 create mode 100644 drivers/pwm/pwm-mc33xs2410.c

-- 
2.39.5

[PATCH v6 1/2] dt-bindings: pwm: add support for MC33XS2410

[Dimitri Fedrau]

Adding documentation for NXPs MC33XS2410 high side switch.

Reviewed-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Signed-off-by: Dimitri Fedrau <dima.fedrau@gmail.com>
---
 .../bindings/pwm/nxp,mc33xs2410.yaml          | 118 ++++++++++++++++++
 1 file changed, 118 insertions(+)
 create mode 100644 Documentation/devicetree/bindings/pwm/nxp,mc33xs2410.yaml

diff --git a/Documentation/devicetree/bindings/pwm/nxp,mc33xs2410.yaml b/Documentation/devicetree/bindings/pwm/nxp,mc33xs2410.yaml
new file mode 100644
index 000000000000..1729fe5c3dfb
--- /dev/null
+++ b/Documentation/devicetree/bindings/pwm/nxp,mc33xs2410.yaml
@@ -0,0 +1,118 @@
+# SPDX-License-Identifier: GPL-2.0-only OR BSD-2-Clause
+%YAML 1.2
+---
+$id: http://devicetree.org/schemas/pwm/nxp,mc33xs2410.yaml#
+$schema: http://devicetree.org/meta-schemas/core.yaml#
+
+title: High-side switch MC33XS2410
+
+maintainers:
+  - Dimitri Fedrau <dima.fedrau@gmail.com>
+
+allOf:
+  - $ref: pwm.yaml#
+  - $ref: /schemas/spi/spi-peripheral-props.yaml#
+
+properties:
+  compatible:
+    const: nxp,mc33xs2410
+
+  reg:
+    maxItems: 1
+
+  spi-max-frequency:
+    maximum: 10000000
+
+  spi-cpha: true
+
+  spi-cs-setup-delay-ns:
+    minimum: 100
+    default: 100
+
+  spi-cs-hold-delay-ns:
+    minimum: 10
+    default: 10
+
+  spi-cs-inactive-delay-ns:
+    minimum: 300
+    default: 300
+
+  reset-gpios:
+    description:
+      GPIO connected to the active low reset pin.
+    maxItems: 1
+
+  "#pwm-cells":
+    const: 3
+
+  pwm-names:
+    items:
+      - const: di0
+      - const: di1
+      - const: di2
+      - const: di3
+
+  pwms:
+    description:
+      Direct inputs(di0-3) are used to directly turn-on or turn-off the
+      outputs.
+    maxItems: 4
+
+  interrupts:
+    maxItems: 1
+
+  clocks:
+    description:
+      The external clock can be used if the internal clock doesn't meet
+      timing requirements over temperature and voltage operating range.
+    maxItems: 1
+
+  vdd-supply:
+    description:
+      Logic supply voltage
+
+  vspi-supply:
+    description:
+      Supply voltage for SPI
+
+  vpwr-supply:
+    description:
+      Power switch supply
+
+required:
+  - compatible
+  - reg
+
+unevaluatedProperties: false
+
+examples:
+  - |
+    #include <dt-bindings/gpio/gpio.h>
+    #include <dt-bindings/interrupt-controller/irq.h>
+    spi {
+        #address-cells = <1>;
+        #size-cells = <0>;
+
+       pwm@0 {
+           compatible = "nxp,mc33xs2410";
+           reg = <0x0>;
+           spi-max-frequency = <4000000>;
+           spi-cpha;
+           spi-cs-setup-delay-ns = <100>;
+           spi-cs-hold-delay-ns = <10>;
+           spi-cs-inactive-delay-ns = <300>;
+           reset-gpios = <&gpio3 22 GPIO_ACTIVE_LOW>;
+           #pwm-cells = <3>;
+           pwm-names = "di0", "di1", "di2", "di3";
+           pwms = <&pwm0 0 1000000>,
+                  <&pwm1 0 1000000>,
+                  <&pwm2 0 1000000>,
+                  <&pwm3 0 1000000>;
+           interrupt-parent = <&gpio0>;
+           interrupts = <31 IRQ_TYPE_LEVEL_LOW>;
+           clocks = <&clk_ext_fixed>;
+           vdd-supply = <&reg_3v3>;
+           vspi-supply = <&reg_3v3>;
+           vpwr-supply = <&reg_24v0>;
+       };
+    };
-- 
2.39.5

[PATCH v6 2/2] pwm: add support for NXPs high-side switch MC33XS2410

[Dimitri Fedrau]

The MC33XS2410 is a four channel high-side switch. Featuring advanced
monitoring and control function, the device is operational from 3.0 V to
60 V. The device is controlled by SPI port for configuration.

Signed-off-by: Dimitri Fedrau <dima.fedrau@gmail.com>
---
 drivers/pwm/Kconfig          |  12 +
 drivers/pwm/Makefile         |   1 +
 drivers/pwm/pwm-mc33xs2410.c | 422 +++++++++++++++++++++++++++++++++++
 3 files changed, 435 insertions(+)
 create mode 100644 drivers/pwm/pwm-mc33xs2410.c

diff --git a/drivers/pwm/Kconfig b/drivers/pwm/Kconfig
index 0915c1e7df16..f513513f9b2f 100644
--- a/drivers/pwm/Kconfig
+++ b/drivers/pwm/Kconfig
@@ -411,6 +411,18 @@ config PWM_LPSS_PLATFORM
 	  To compile this driver as a module, choose M here: the module
 	  will be called pwm-lpss-platform.
 
+config PWM_MC33XS2410
+	tristate "MC33XS2410 PWM support"
+	depends on OF
+	depends on SPI
+	help
+	  NXP MC33XS2410 high-side switch driver. The MC33XS2410 is a four
+	  channel high-side switch. The device is operational from 3.0 V
+	  to 60 V. The device is controlled by SPI port for configuration.
+
+	  To compile this driver as a module, choose M here: the module
+	  will be called pwm-mc33xs2410.
+
 config PWM_MESON
 	tristate "Amlogic Meson PWM driver"
 	depends on ARCH_MESON || COMPILE_TEST
diff --git a/drivers/pwm/Makefile b/drivers/pwm/Makefile
index 9081e0c0e9e0..c75deeeace40 100644
--- a/drivers/pwm/Makefile
+++ b/drivers/pwm/Makefile
@@ -36,6 +36,7 @@ obj-$(CONFIG_PWM_LPC32XX)	+= pwm-lpc32xx.o
 obj-$(CONFIG_PWM_LPSS)		+= pwm-lpss.o
 obj-$(CONFIG_PWM_LPSS_PCI)	+= pwm-lpss-pci.o
 obj-$(CONFIG_PWM_LPSS_PLATFORM)	+= pwm-lpss-platform.o
+obj-$(CONFIG_PWM_MC33XS2410)	+= pwm-mc33xs2410.o
 obj-$(CONFIG_PWM_MESON)		+= pwm-meson.o
 obj-$(CONFIG_PWM_MEDIATEK)	+= pwm-mediatek.o
 obj-$(CONFIG_PWM_MICROCHIP_CORE)	+= pwm-microchip-core.o
diff --git a/drivers/pwm/pwm-mc33xs2410.c b/drivers/pwm/pwm-mc33xs2410.c
new file mode 100644
index 000000000000..f9a334a5e69b
--- /dev/null
+++ b/drivers/pwm/pwm-mc33xs2410.c
@@ -0,0 +1,422 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2024 Liebherr-Electronics and Drives GmbH
+ *
+ * Reference Manual : https://www.nxp.com/docs/en/data-sheet/MC33XS2410.pdf
+ *
+ * Limitations:
+ * - Supports frequencies between 0.5Hz and 2048Hz with following steps:
+ *   - 0.5 Hz steps from 0.5 Hz to 32 Hz
+ *   - 2 Hz steps from 2 Hz to 128 Hz
+ *   - 8 Hz steps from 8 Hz to 512 Hz
+ *   - 32 Hz steps from 32 Hz to 2048 Hz
+ * - Cannot generate a 0 % duty cycle.
+ * - Always produces low output if disabled.
+ * - Configuration isn't atomic. When changing polarity, duty cycle or period
+ *   the data is taken immediately, counters not being affected, resulting in a
+ *   behavior of the output pin that is neither the old nor the new state,
+ *   rather something in between.
+ */
+
+#include <linux/bitfield.h>
+#include <linux/delay.h>
+#include <linux/err.h>
+#include <linux/math64.h>
+#include <linux/minmax.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/pwm.h>
+
+#include <asm/unaligned.h>
+
+#include <linux/spi/spi.h>
+
+#define MC33XS2410_GLB_CTRL		0x00
+#define MC33XS2410_GLB_CTRL_MODE	GENMASK(7, 6)
+#define MC33XS2410_GLB_CTRL_MODE_NORMAL	FIELD_PREP(MC33XS2410_GLB_CTRL_MODE, 1)
+#define MC33XS2410_PWM_CTRL1		0x05
+#define MC33XS2410_PWM_CTRL1_POL_INV(x)	BIT(x)
+#define MC33XS2410_PWM_CTRL3		0x07
+/* x in { 0 ... 3 } */
+#define MC33XS2410_PWM_CTRL3_EN(x)	BIT(4 + (x))
+#define MC33XS2410_PWM_FREQ1		0x08
+/* x in { 1 ... 4 } */
+#define MC33XS2410_PWM_FREQ(x)		(MC33XS2410_PWM_FREQ1 + (x - 1))
+#define MC33XS2410_PWM_FREQ_STEP_MASK	GENMASK(7, 6)
+#define MC33XS2410_PWM_FREQ_COUNT_MASK	GENMASK(5, 0)
+#define MC33XS2410_PWM_DC1		0x0c
+/* x in { 1 ... 4 } */
+#define MC33XS2410_PWM_DC(x)		(MC33XS2410_PWM_DC1 + (x - 1))
+#define MC33XS2410_WDT			0x14
+
+#define MC33XS2410_WR			BIT(7)
+#define MC33XS2410_RD_CTRL		BIT(7)
+#define MC33XS2410_RD_DATA_MASK		GENMASK(13, 0)
+
+#define MC33XS2410_MIN_PERIOD		488282
+#define MC33XS2410_MAX_PERIOD_STEP0	2000000000
+/* x in { 0 ... 3 } */
+#define MC33XS2410_MAX_PERIOD_STEP(x)	(MC33XS2410_MAX_PERIOD_STEP0 >> (2 * x))
+
+#define MC33XS2410_MAX_TRANSFERS	5
+#define MC33XS2410_WORD_LEN		2
+
+struct mc33xs2410_pwm {
+	struct spi_device *spi;
+};
+
+static inline struct mc33xs2410_pwm *mc33xs2410_from_chip(struct pwm_chip *chip)
+{
+	return pwmchip_get_drvdata(chip);
+}
+
+static int mc33xs2410_xfer_regs(struct spi_device *spi, bool read, u8 *reg,
+				u16 *val, bool *ctrl, int len)
+{
+	struct spi_transfer t[MC33XS2410_MAX_TRANSFERS] = { { 0 } };
+	u8 tx[MC33XS2410_MAX_TRANSFERS * MC33XS2410_WORD_LEN];
+	u8 rx[MC33XS2410_MAX_TRANSFERS * MC33XS2410_WORD_LEN];
+	int i, ret, reg_i, val_i;
+
+	if (!len)
+		return 0;
+
+	if (read)
+		len++;
+
+	if (len > MC33XS2410_MAX_TRANSFERS)
+		return -EINVAL;
+
+	for (i = 0; i < len; i++) {
+		reg_i = i * MC33XS2410_WORD_LEN;
+		val_i = reg_i + 1;
+		if (read) {
+			if (i < len - 1) {
+				tx[reg_i] = reg[i];
+				tx[val_i] = ctrl[i] ? MC33XS2410_RD_CTRL : 0;
+				t[i].tx_buf = &tx[reg_i];
+			}
+
+			if (i > 0)
+				t[i].rx_buf = &rx[reg_i - MC33XS2410_WORD_LEN];
+		} else {
+			tx[reg_i] = reg[i] | MC33XS2410_WR;
+			tx[val_i] = val[i];
+			t[i].tx_buf = &tx[reg_i];
+		}
+
+		t[i].len = MC33XS2410_WORD_LEN;
+		t[i].cs_change = 1;
+	}
+
+	t[len - 1].cs_change = 0;
+
+	ret = spi_sync_transfer(spi, &t[0], len);
+	if (ret < 0)
+		return ret;
+
+	if (read) {
+		for (i = 0; i < len - 1; i++) {
+			reg_i = i * MC33XS2410_WORD_LEN;
+			val[i] = FIELD_GET(MC33XS2410_RD_DATA_MASK,
+					   get_unaligned_be16(&rx[reg_i]));
+		}
+	}
+
+	return 0;
+}
+
+static
+int mc33xs2410_write_regs(struct spi_device *spi, u8 *reg, u16 *val, int len)
+{
+
+	return mc33xs2410_xfer_regs(spi, false, reg, val, NULL, len);
+}
+
+static int mc33xs2410_read_regs(struct spi_device *spi, u8 *reg, bool *ctrl,
+				u16 *val, u8 len)
+{
+	return mc33xs2410_xfer_regs(spi, true, reg, val, ctrl, len);
+}
+
+
+static int mc33xs2410_write_reg(struct spi_device *spi, u8 reg, u16 val)
+{
+	return mc33xs2410_write_regs(spi, &reg, &val, 1);
+}
+
+static
+int mc33xs2410_read_reg(struct spi_device *spi, u8 reg, u16 *val, bool ctrl)
+{
+	return mc33xs2410_read_regs(spi, &reg, &ctrl, val, 1);
+}
+
+static int mc33xs2410_read_reg_ctrl(struct spi_device *spi, u8 reg, u16 *val)
+{
+	return mc33xs2410_read_reg(spi, reg, val, true);
+}
+
+static
+int mc33xs2410_modify_reg(struct spi_device *spi, u8 reg, u16 mask, u16 val)
+{
+	u16 tmp;
+	int ret;
+
+	ret = mc33xs2410_read_reg_ctrl(spi, reg, &tmp);
+	if (ret < 0)
+		return ret;
+
+	tmp &= ~mask;
+	tmp |= val & mask;
+
+	return mc33xs2410_write_reg(spi, reg, tmp);
+}
+
+static u8 mc33xs2410_pwm_get_freq(u64 period)
+{
+	u8 step, count;
+
+	/*
+	 * Check which step is appropriate for the given period, starting with
+	 * the highest frequency(lowest period). Higher frequencies are
+	 * represented with better resolution by the device. Therefore favor
+	 * frequency range with the better resolution to minimize error
+	 * introduced by the frequency steps.
+	 */
+
+	switch (period) {
+	case MC33XS2410_MIN_PERIOD ... MC33XS2410_MAX_PERIOD_STEP(3):
+		step = 3;
+		break;
+	case MC33XS2410_MAX_PERIOD_STEP(3) + 1 ... MC33XS2410_MAX_PERIOD_STEP(2):
+		step = 2;
+		break;
+	case MC33XS2410_MAX_PERIOD_STEP(2) + 1 ... MC33XS2410_MAX_PERIOD_STEP(1):
+		step = 1;
+		break;
+	case MC33XS2410_MAX_PERIOD_STEP(1) + 1 ... MC33XS2410_MAX_PERIOD_STEP(0):
+		step = 0;
+		break;
+	}
+
+	/*
+	 * Round up here because a higher count results in a higher frequency
+	 * and so a smaller period.
+	 */
+	count = DIV_ROUND_UP((u32)MC33XS2410_MAX_PERIOD_STEP(step), (u32)period);
+	return FIELD_PREP(MC33XS2410_PWM_FREQ_STEP_MASK, step) |
+	       FIELD_PREP(MC33XS2410_PWM_FREQ_COUNT_MASK, count - 1);
+}
+
+static u64 mc33xs2410_pwm_get_period(u8 reg)
+{
+	u32 freq, code, doubled_steps;
+
+	/*
+	 * steps:
+	 *   - 0 = 0.5Hz
+	 *   - 1 = 2Hz
+	 *   - 2 = 8Hz
+	 *   - 3 = 32Hz
+	 * frequency = (code + 1) x steps.
+	 *
+	 * To avoid losing precision in case steps value is zero, scale the
+	 * steps value for now by two and keep it in mind when calculating the
+	 * period that the frequency had been doubled.
+	 */
+	doubled_steps = 1 << (FIELD_GET(MC33XS2410_PWM_FREQ_STEP_MASK, reg) * 2);
+	code = FIELD_GET(MC33XS2410_PWM_FREQ_COUNT_MASK, reg);
+	freq = (code + 1) * doubled_steps;
+
+	/* Convert frequency to period, considering the doubled frequency. */
+	return DIV_ROUND_UP((u32)(2 * NSEC_PER_SEC), freq);
+}
+
+static int mc33xs2410_pwm_get_relative_duty_cycle(u64 period, u64 duty_cycle)
+{
+	/*
+	 * duty_cycle cannot overflow and period is not zero, since this is
+	 * guaranteed by the caller.
+	 */
+	duty_cycle *= 256;
+	duty_cycle = div64_u64(duty_cycle, period);
+
+	return duty_cycle - 1;
+}
+
+static void mc33xs2410_pwm_set_relative_duty_cycle(struct pwm_state *state,
+						   u16 duty_cycle)
+{
+	if (!state->enabled)
+		state->duty_cycle = 0;
+	else
+		state->duty_cycle = DIV_ROUND_UP_ULL((duty_cycle + 1) * state->period, 256);
+}
+
+static int mc33xs2410_pwm_apply(struct pwm_chip *chip, struct pwm_device *pwm,
+				const struct pwm_state *state)
+{
+	struct mc33xs2410_pwm *mc33xs2410 = mc33xs2410_from_chip(chip);
+	struct spi_device *spi = mc33xs2410->spi;
+	u8 reg[4] = {
+			MC33XS2410_PWM_FREQ(pwm->hwpwm + 1),
+			MC33XS2410_PWM_DC(pwm->hwpwm + 1),
+			MC33XS2410_PWM_CTRL1,
+			MC33XS2410_PWM_CTRL3
+		    };
+	bool ctrl[2] = { true, true };
+	u64 period, duty_cycle;
+	int ret, rel_dc;
+	u16 val[4];
+	u8 mask;
+
+	period = min(state->period, MC33XS2410_MAX_PERIOD_STEP(0));
+	if (period < MC33XS2410_MIN_PERIOD)
+		return -EINVAL;
+
+	ret = mc33xs2410_read_regs(spi, &reg[2], &ctrl[0], &val[2], 2);
+	if (ret < 0)
+		return ret;
+
+	/* frequency */
+	val[0] = mc33xs2410_pwm_get_freq(period);
+	/* Continue calculations with the possibly truncated period */
+	period = mc33xs2410_pwm_get_period(val[0]);
+
+	/* duty cycle */
+	duty_cycle = min(period, state->duty_cycle);
+	rel_dc = mc33xs2410_pwm_get_relative_duty_cycle(period, duty_cycle);
+	val[1] = rel_dc < 0 ? 0 : rel_dc;
+
+	/* polarity */
+	mask = MC33XS2410_PWM_CTRL1_POL_INV(pwm->hwpwm);
+	val[2] = (state->polarity == PWM_POLARITY_INVERSED) ?
+		 (val[2] | mask) : (val[2] & ~mask);
+
+	/* enable output */
+	mask = MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm);
+	val[3] = (state->enabled && rel_dc >= 0) ? (val[3] | mask) :
+						   (val[3] & ~mask);
+
+	return mc33xs2410_write_regs(spi, reg, val, 4);
+}
+
+static int mc33xs2410_pwm_get_state(struct pwm_chip *chip,
+				    struct pwm_device *pwm,
+				    struct pwm_state *state)
+{
+	struct mc33xs2410_pwm *mc33xs2410 = mc33xs2410_from_chip(chip);
+	struct spi_device *spi = mc33xs2410->spi;
+	u8 reg[4] = {
+			MC33XS2410_PWM_FREQ(pwm->hwpwm + 1),
+			MC33XS2410_PWM_DC(pwm->hwpwm + 1),
+			MC33XS2410_PWM_CTRL1,
+			MC33XS2410_PWM_CTRL3,
+		    };
+	bool ctrl[4] = { true, true, true, true };
+	u16 val[4];
+	int ret;
+
+	ret = mc33xs2410_read_regs(spi, reg, ctrl, val, 4);
+	if (ret < 0)
+		return ret;
+
+	state->period = mc33xs2410_pwm_get_period(val[0]);
+	state->polarity = (val[2] & MC33XS2410_PWM_CTRL1_POL_INV(pwm->hwpwm)) ?
+			  PWM_POLARITY_INVERSED : PWM_POLARITY_NORMAL;
+	state->enabled = !!(val[3] & MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm));
+	mc33xs2410_pwm_set_relative_duty_cycle(state, val[1]);
+	return 0;
+}
+
+static const struct pwm_ops mc33xs2410_pwm_ops = {
+	.apply = mc33xs2410_pwm_apply,
+	.get_state = mc33xs2410_pwm_get_state,
+};
+
+static int mc33xs2410_reset(struct device *dev)
+{
+	struct gpio_desc *reset_gpio;
+
+	reset_gpio = devm_gpiod_get_optional(dev, "reset", GPIOD_OUT_HIGH);
+	if (IS_ERR_OR_NULL(reset_gpio))
+		return PTR_ERR_OR_ZERO(reset_gpio);
+
+	fsleep(1000);
+	gpiod_set_value_cansleep(reset_gpio, 0);
+	/* Wake-up time */
+	fsleep(10000);
+
+	return 0;
+}
+
+static int mc33xs2410_probe(struct spi_device *spi)
+{
+	struct mc33xs2410_pwm *mc33xs2410;
+	struct device *dev = &spi->dev;
+	struct pwm_chip *chip;
+	int ret;
+
+	chip = devm_pwmchip_alloc(dev, 4, sizeof(*mc33xs2410));
+	if (IS_ERR(chip))
+		return PTR_ERR(chip);
+
+	mc33xs2410 = mc33xs2410_from_chip(chip);
+	mc33xs2410->spi = spi;
+	chip->ops = &mc33xs2410_pwm_ops;
+
+	ret = mc33xs2410_reset(dev);
+	if (ret)
+		return ret;
+
+	/*
+	 * Disable watchdog and keep in mind that the watchdog won't trigger a
+	 * reset of the machine when running into an timeout, instead the
+	 * control over the outputs is handed over to the INx input logic
+	 * signals of the device. Disabling it here just deactivates this
+	 * feature until a proper solution is found.
+	 */
+	ret = mc33xs2410_write_reg(spi, MC33XS2410_WDT, 0x0);
+	if (ret < 0)
+		return dev_err_probe(dev, ret, "Failed to disable watchdog\n");
+
+	/* Transition to normal mode */
+	ret = mc33xs2410_modify_reg(spi, MC33XS2410_GLB_CTRL,
+				    MC33XS2410_GLB_CTRL_MODE,
+				    MC33XS2410_GLB_CTRL_MODE_NORMAL);
+	if (ret < 0)
+		return dev_err_probe(dev, ret,
+				     "Failed to transition to normal mode\n");
+
+	ret = devm_pwmchip_add(dev, chip);
+	if (ret < 0)
+		return dev_err_probe(dev, ret, "Failed to add pwm chip\n");
+
+	return 0;
+}
+
+static const struct spi_device_id mc33xs2410_spi_id[] = {
+	{ "mc33xs2410" },
+	{ }
+};
+MODULE_DEVICE_TABLE(spi, mc33xs2410_spi_id);
+
+static const struct of_device_id mc33xs2410_of_match[] = {
+	{ .compatible = "nxp,mc33xs2410" },
+	{ }
+};
+MODULE_DEVICE_TABLE(of, mc33xs2410_of_match);
+
+static struct spi_driver mc33xs2410_driver = {
+	.driver = {
+		.name = "mc33xs2410-pwm",
+		.of_match_table = mc33xs2410_of_match,
+	},
+	.probe = mc33xs2410_probe,
+	.id_table = mc33xs2410_spi_id,
+};
+module_spi_driver(mc33xs2410_driver);
+
+MODULE_DESCRIPTION("NXP MC33XS2410 high-side switch driver");
+MODULE_AUTHOR("Dimitri Fedrau <dima.fedrau@gmail.com>");
+MODULE_LICENSE("GPL");
-- 
2.39.5

Re: [PATCH v6 2/2] pwm: add support for NXPs high-side switch MC33XS2410

[Uwe Kleine-König]

[-- Attachment #1: Type: text/plain, Size: 8414 bytes --]

Hello Dimitri,

On Fri, Sep 27, 2024 at 02:57:45PM +0200, Dimitri Fedrau wrote:
> diff --git a/drivers/pwm/pwm-mc33xs2410.c b/drivers/pwm/pwm-mc33xs2410.c
> new file mode 100644
> index 000000000000..f9a334a5e69b
> --- /dev/null
> +++ b/drivers/pwm/pwm-mc33xs2410.c
> @@ -0,0 +1,422 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/*
> + * Copyright (C) 2024 Liebherr-Electronics and Drives GmbH
> + *
> + * Reference Manual : https://www.nxp.com/docs/en/data-sheet/MC33XS2410.pdf
> + *
> + * Limitations:
> + * - Supports frequencies between 0.5Hz and 2048Hz with following steps:
> + *   - 0.5 Hz steps from 0.5 Hz to 32 Hz
> + *   - 2 Hz steps from 2 Hz to 128 Hz
> + *   - 8 Hz steps from 8 Hz to 512 Hz
> + *   - 32 Hz steps from 32 Hz to 2048 Hz
> + * - Cannot generate a 0 % duty cycle.
> + * - Always produces low output if disabled.
> + * - Configuration isn't atomic. When changing polarity, duty cycle or period
> + *   the data is taken immediately, counters not being affected, resulting in a
> + *   behavior of the output pin that is neither the old nor the new state,
> + *   rather something in between.
> + */
> +
> +#include <linux/bitfield.h>
> +#include <linux/delay.h>
> +#include <linux/err.h>
> +#include <linux/math64.h>
> +#include <linux/minmax.h>
> +#include <linux/module.h>
> +#include <linux/of.h>
> +#include <linux/pwm.h>
> +
> +#include <asm/unaligned.h>
> +
> +#include <linux/spi/spi.h>
> +
> +#define MC33XS2410_GLB_CTRL		0x00
> +#define MC33XS2410_GLB_CTRL_MODE	GENMASK(7, 6)
> +#define MC33XS2410_GLB_CTRL_MODE_NORMAL	FIELD_PREP(MC33XS2410_GLB_CTRL_MODE, 1)
> +#define MC33XS2410_PWM_CTRL1		0x05
> +#define MC33XS2410_PWM_CTRL1_POL_INV(x)	BIT(x)
> +#define MC33XS2410_PWM_CTRL3		0x07
> +/* x in { 0 ... 3 } */
> +#define MC33XS2410_PWM_CTRL3_EN(x)	BIT(4 + (x))
> +#define MC33XS2410_PWM_FREQ1		0x08
> +/* x in { 1 ... 4 } */
> +#define MC33XS2410_PWM_FREQ(x)		(MC33XS2410_PWM_FREQ1 + (x - 1))
> +#define MC33XS2410_PWM_FREQ_STEP_MASK	GENMASK(7, 6)
> +#define MC33XS2410_PWM_FREQ_COUNT_MASK	GENMASK(5, 0)
> +#define MC33XS2410_PWM_DC1		0x0c
> +/* x in { 1 ... 4 } */
> +#define MC33XS2410_PWM_DC(x)		(MC33XS2410_PWM_DC1 + (x - 1))
> +#define MC33XS2410_WDT			0x14
> +
> +#define MC33XS2410_WR			BIT(7)
> +#define MC33XS2410_RD_CTRL		BIT(7)
> +#define MC33XS2410_RD_DATA_MASK		GENMASK(13, 0)
> +
> +#define MC33XS2410_MIN_PERIOD		488282
> +#define MC33XS2410_MAX_PERIOD_STEP0	2000000000
> +/* x in { 0 ... 3 } */
> +#define MC33XS2410_MAX_PERIOD_STEP(x)	(MC33XS2410_MAX_PERIOD_STEP0 >> (2 * x))

Nitpick: These register definition become easier to parse for a human if
you indent the RHS of register fields one tab further and add an empty
line between the definitions for different registers.

MC33XS2410_PWM_DC1 is only used once, I'd hard-code it into the
definition of MC33XS2410_PWM_DC.

The register fields [7:4] in MC33XS2410_PWM_CTRL3 are called PWM_ON4 ..
PWM_ON1. So your x in { 0 ... 3 } is wrong. (Luckily, having some x
range over { 0 ... 3 } and others orver { 1 ... 4 } is prone to error
and confusion.)

Also I'd drop all _MASK suffixes.

For MC33XS2410_MAX_PERIOD_STEP maybe use a different variable name than
for the others. For the register definitions the range is over hwpwm
(which might be a good name there?), for MC33XS2410_MAX_PERIOD_STEP it's
about MC33XS2410_PWM_FREQ_STEP.

> +#define MC33XS2410_MAX_TRANSFERS	5
> +#define MC33XS2410_WORD_LEN		2
> +
> +struct mc33xs2410_pwm {
> +	struct spi_device *spi;
> +};
> +
> +static inline struct mc33xs2410_pwm *mc33xs2410_from_chip(struct pwm_chip *chip)
> +{
> +	return pwmchip_get_drvdata(chip);
> +}
> +
> +static int mc33xs2410_xfer_regs(struct spi_device *spi, bool read, u8 *reg,
> +				u16 *val, bool *ctrl, int len)

Should len better be unsigned?

Unless I missed something all ctrl[x] are always identical. If so
represent that by a single bool.

> +{
> +	struct spi_transfer t[MC33XS2410_MAX_TRANSFERS] = { { 0 } };
> +	u8 tx[MC33XS2410_MAX_TRANSFERS * MC33XS2410_WORD_LEN];
> +	u8 rx[MC33XS2410_MAX_TRANSFERS * MC33XS2410_WORD_LEN];
> +	int i, ret, reg_i, val_i;
> +
> +	if (!len)
> +		return 0;
> +
> +	if (read)
> +		len++;
> +
> +	if (len > MC33XS2410_MAX_TRANSFERS)
> +		return -EINVAL;
> +
> +	for (i = 0; i < len; i++) {
> +		reg_i = i * MC33XS2410_WORD_LEN;
> +		val_i = reg_i + 1;
> +		if (read) {
> +			if (i < len - 1) {
> +				tx[reg_i] = reg[i];
> +				tx[val_i] = ctrl[i] ? MC33XS2410_RD_CTRL : 0;
> +				t[i].tx_buf = &tx[reg_i];
> +			}
> +
> +			if (i > 0)
> +				t[i].rx_buf = &rx[reg_i - MC33XS2410_WORD_LEN];
> +		} else {
> +			tx[reg_i] = reg[i] | MC33XS2410_WR;
> +			tx[val_i] = val[i];
> +			t[i].tx_buf = &tx[reg_i];
> +		}
> +
> +		t[i].len = MC33XS2410_WORD_LEN;
> +		t[i].cs_change = 1;

Not sure if MC33XS2410_WORD_LEN really improves readability here.

Why is this done using $len transfers, wouldn't a single one do (and
maybe be more performant and not rely on a spi controller that supports
cs_change)?

> +	}
> +
> +	t[len - 1].cs_change = 0;
> +
> +	ret = spi_sync_transfer(spi, &t[0], len);
> +	if (ret < 0)
> +		return ret;
> +
> +	if (read) {
> +		for (i = 0; i < len - 1; i++) {
> +			reg_i = i * MC33XS2410_WORD_LEN;
> +			val[i] = FIELD_GET(MC33XS2410_RD_DATA_MASK,
> +					   get_unaligned_be16(&rx[reg_i]));
> +		}
> +	}
> +
> +	return 0;
> +}
> [...]
> +static
> +int mc33xs2410_read_reg(struct spi_device *spi, u8 reg, u16 *val, bool ctrl)

My personal opinion: Better break the line in the argument list or not
at all. Having a "static" on its own line looks ugly.

> +{
> +	return mc33xs2410_read_regs(spi, &reg, &ctrl, val, 1);
> +}
> [...]
> +static u64 mc33xs2410_pwm_get_period(u8 reg)
> +{
> [...]
> +	/* Convert frequency to period, considering the doubled frequency. */
> +	return DIV_ROUND_UP((u32)(2 * NSEC_PER_SEC), freq);

That u32 cast isn't needed.

> +}
> [...]
> +static int mc33xs2410_pwm_apply(struct pwm_chip *chip, struct pwm_device *pwm,
> +				const struct pwm_state *state)
> +{
> [...]
> +	/* frequency */
> +	val[0] = mc33xs2410_pwm_get_freq(period);
> +	/* Continue calculations with the possibly truncated period */
> +	period = mc33xs2410_pwm_get_period(val[0]);
> +
> +	/* duty cycle */
> +	duty_cycle = min(period, state->duty_cycle);
> +	rel_dc = mc33xs2410_pwm_get_relative_duty_cycle(period, duty_cycle);
> +	val[1] = rel_dc < 0 ? 0 : rel_dc;

Handling of the duty cycle is correct here, but misleading. I already
added a comment here that using val[1] = 0 if rel_dc < 0 is wrong and
just deleted it again after I saw (rel_dc >= 0) being used determining
the value for MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm). An explicit if block
would make this more obvious.

mc33xs2410_pwm_get_relative_duty_cycle() is simple enough and only used
once that I'd unroll it here.

> +	/* polarity */
> +	mask = MC33XS2410_PWM_CTRL1_POL_INV(pwm->hwpwm);
> +	val[2] = (state->polarity == PWM_POLARITY_INVERSED) ?
> +		 (val[2] | mask) : (val[2] & ~mask);
> +
> +	/* enable output */
> +	mask = MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm);
> +	val[3] = (state->enabled && rel_dc >= 0) ? (val[3] | mask) :
> +						   (val[3] & ~mask);
> +
> +	return mc33xs2410_write_regs(spi, reg, val, 4);
> +}
> +
> +static int mc33xs2410_pwm_get_state(struct pwm_chip *chip,
> +				    struct pwm_device *pwm,
> +				    struct pwm_state *state)
> +{
> [...]
> +	state->period = mc33xs2410_pwm_get_period(val[0]);
> +	state->polarity = (val[2] & MC33XS2410_PWM_CTRL1_POL_INV(pwm->hwpwm)) ?
> +			  PWM_POLARITY_INVERSED : PWM_POLARITY_NORMAL;
> +	state->enabled = !!(val[3] & MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm));
> +	mc33xs2410_pwm_set_relative_duty_cycle(state, val[1]);

No need to set state->duty_cycle = 0 if state->enabled is false. This is
another function I suggest to unroll as it hides more than it abstracts.

> +	return 0;
> +}
> +
> [...]
> +static int mc33xs2410_probe(struct spi_device *spi)
> +{
> [...]
> +	/* Transition to normal mode */
> +	ret = mc33xs2410_modify_reg(spi, MC33XS2410_GLB_CTRL,
> +				    MC33XS2410_GLB_CTRL_MODE,
> +				    MC33XS2410_GLB_CTRL_MODE_NORMAL);
> +	if (ret < 0)
> +		return dev_err_probe(dev, ret,
> +				     "Failed to transition to normal mode\n");

What is the effect of this register write if the PWM was already setup
by the bootloader?

> +
> +	ret = devm_pwmchip_add(dev, chip);
> +	if (ret < 0)
> +		return dev_err_probe(dev, ret, "Failed to add pwm chip\n");
> +
> +	return 0;
> +}

Best regards
Uwe

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v6 2/2] pwm: add support for NXPs high-side switch MC33XS2410

[Dimitri Fedrau]

Hello Uwe,

Am Tue, Oct 22, 2024 at 09:54:50AM +0200 schrieb Uwe Kleine-König:

[...]
> > +
> > +#define MC33XS2410_MIN_PERIOD		488282
> > +#define MC33XS2410_MAX_PERIOD_STEP0	2000000000
> > +/* x in { 0 ... 3 } */
> > +#define MC33XS2410_MAX_PERIOD_STEP(x)	(MC33XS2410_MAX_PERIOD_STEP0 >> (2 * x))
> 
> Nitpick: These register definition become easier to parse for a human if
> you indent the RHS of register fields one tab further and add an empty
> line between the definitions for different registers.
> 

Adding an empty line seems reasonable to me but the additional tab doesn't
help me to improve readability.

> MC33XS2410_PWM_DC1 is only used once, I'd hard-code it into the
> definition of MC33XS2410_PWM_DC.
> 

Ok. Should I do the same for MC33XS2410_PWM_FREQ1 and
MC33XS2410_MAX_PERIOD_STEP0 ?

> The register fields [7:4] in MC33XS2410_PWM_CTRL3 are called PWM_ON4 ..
> PWM_ON1. So your x in { 0 ... 3 } is wrong. (Luckily, having some x
> range over { 0 ... 3 } and others orver { 1 ... 4 } is prone to error
> and confusion.)
> 

Will fix it. Should I do the same for MC33XS2410_PWM_CTRL1_POL_INV ?

> Also I'd drop all _MASK suffixes.
> 

Ok.

> For MC33XS2410_MAX_PERIOD_STEP maybe use a different variable name than
> for the others. For the register definitions the range is over hwpwm
> (which might be a good name there?), for MC33XS2410_MAX_PERIOD_STEP it's
> about MC33XS2410_PWM_FREQ_STEP.
> 

What about MC33XS2410_PWM_MAX_PERIOD(x) ?

> > +#define MC33XS2410_MAX_TRANSFERS	5
> > +#define MC33XS2410_WORD_LEN		2
> > +
> > +struct mc33xs2410_pwm {
> > +	struct spi_device *spi;
> > +};
> > +
> > +static inline struct mc33xs2410_pwm *mc33xs2410_from_chip(struct pwm_chip *chip)
> > +{
> > +	return pwmchip_get_drvdata(chip);
> > +}
> > +
> > +static int mc33xs2410_xfer_regs(struct spi_device *spi, bool read, u8 *reg,
> > +				u16 *val, bool *ctrl, int len)
> 
> Should len better be unsigned?
> 

I switch to unsigned.

> Unless I missed something all ctrl[x] are always identical. If so
> represent that by a single bool.
> 

Yes, they are identical. I added the crtl[x] to be able read from ctrl and
diag registers. I will change it so it is represented by a single bool, if
the feature is needed in the future I can still add it.

> > +{
> > +	struct spi_transfer t[MC33XS2410_MAX_TRANSFERS] = { { 0 } };
> > +	u8 tx[MC33XS2410_MAX_TRANSFERS * MC33XS2410_WORD_LEN];
> > +	u8 rx[MC33XS2410_MAX_TRANSFERS * MC33XS2410_WORD_LEN];
> > +	int i, ret, reg_i, val_i;
> > +
> > +	if (!len)
> > +		return 0;
> > +
> > +	if (read)
> > +		len++;
> > +
> > +	if (len > MC33XS2410_MAX_TRANSFERS)
> > +		return -EINVAL;
> > +
> > +	for (i = 0; i < len; i++) {
> > +		reg_i = i * MC33XS2410_WORD_LEN;
> > +		val_i = reg_i + 1;
> > +		if (read) {
> > +			if (i < len - 1) {
> > +				tx[reg_i] = reg[i];
> > +				tx[val_i] = ctrl[i] ? MC33XS2410_RD_CTRL : 0;
> > +				t[i].tx_buf = &tx[reg_i];
> > +			}
> > +
> > +			if (i > 0)
> > +				t[i].rx_buf = &rx[reg_i - MC33XS2410_WORD_LEN];
> > +		} else {
> > +			tx[reg_i] = reg[i] | MC33XS2410_WR;
> > +			tx[val_i] = val[i];
> > +			t[i].tx_buf = &tx[reg_i];
> > +		}
> > +
> > +		t[i].len = MC33XS2410_WORD_LEN;
> > +		t[i].cs_change = 1;
> 
> Not sure if MC33XS2410_WORD_LEN really improves readability here.
> 

It is used throughout in the function and improves readability overall,
maybe not here but for consistency I would stick to it.

> Why is this done using $len transfers, wouldn't a single one do (and
> maybe be more performant and not rely on a spi controller that supports
> cs_change)?
> 

Without cs_change after every 16 bit, requests aren't processed by the
device. Reading/writing from/to device fails. The SPI controller therefore
must support cs_change. Single transfer is not possible because of the
cs_change after every 16bit.

> > +	}
> > +
> > +	t[len - 1].cs_change = 0;
> > +
> > +	ret = spi_sync_transfer(spi, &t[0], len);
> > +	if (ret < 0)
> > +		return ret;
> > +
> > +	if (read) {
> > +		for (i = 0; i < len - 1; i++) {
> > +			reg_i = i * MC33XS2410_WORD_LEN;
> > +			val[i] = FIELD_GET(MC33XS2410_RD_DATA_MASK,
> > +					   get_unaligned_be16(&rx[reg_i]));
> > +		}
> > +	}
> > +
> > +	return 0;
> > +}
> > [...]
> > +static
> > +int mc33xs2410_read_reg(struct spi_device *spi, u8 reg, u16 *val, bool ctrl)
> 
> My personal opinion: Better break the line in the argument list or not
> at all. Having a "static" on its own line looks ugly.
> 
Ok.

> > +{
> > +	return mc33xs2410_read_regs(spi, &reg, &ctrl, val, 1);
> > +}
> > [...]
> > +static u64 mc33xs2410_pwm_get_period(u8 reg)
> > +{
> > [...]
> > +	/* Convert frequency to period, considering the doubled frequency. */
> > +	return DIV_ROUND_UP((u32)(2 * NSEC_PER_SEC), freq);
> 
> That u32 cast isn't needed.
>
Will fix it.

> > +}
> > [...]
> > +static int mc33xs2410_pwm_apply(struct pwm_chip *chip, struct pwm_device *pwm,
> > +				const struct pwm_state *state)
> > +{
> > [...]
> > +	/* frequency */
> > +	val[0] = mc33xs2410_pwm_get_freq(period);
> > +	/* Continue calculations with the possibly truncated period */
> > +	period = mc33xs2410_pwm_get_period(val[0]);
> > +
> > +	/* duty cycle */
> > +	duty_cycle = min(period, state->duty_cycle);
> > +	rel_dc = mc33xs2410_pwm_get_relative_duty_cycle(period, duty_cycle);
> > +	val[1] = rel_dc < 0 ? 0 : rel_dc;
> 
> Handling of the duty cycle is correct here, but misleading. I already
> added a comment here that using val[1] = 0 if rel_dc < 0 is wrong and
> just deleted it again after I saw (rel_dc >= 0) being used determining
> the value for MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm). An explicit if block
> would make this more obvious.
>

Will add an explicit if block, should I do the same for the value for
MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm) ?

> mc33xs2410_pwm_get_relative_duty_cycle() is simple enough and only used
> once that I'd unroll it here.
> 

You are right, will fix it.

> > +	/* polarity */
> > +	mask = MC33XS2410_PWM_CTRL1_POL_INV(pwm->hwpwm);
> > +	val[2] = (state->polarity == PWM_POLARITY_INVERSED) ?
> > +		 (val[2] | mask) : (val[2] & ~mask);
> > +
> > +	/* enable output */
> > +	mask = MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm);
> > +	val[3] = (state->enabled && rel_dc >= 0) ? (val[3] | mask) :
> > +						   (val[3] & ~mask);
> > +
> > +	return mc33xs2410_write_regs(spi, reg, val, 4);
> > +}
> > +
> > +static int mc33xs2410_pwm_get_state(struct pwm_chip *chip,
> > +				    struct pwm_device *pwm,
> > +				    struct pwm_state *state)
> > +{
> > [...]
> > +	state->period = mc33xs2410_pwm_get_period(val[0]);
> > +	state->polarity = (val[2] & MC33XS2410_PWM_CTRL1_POL_INV(pwm->hwpwm)) ?
> > +			  PWM_POLARITY_INVERSED : PWM_POLARITY_NORMAL;
> > +	state->enabled = !!(val[3] & MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm));
> > +	mc33xs2410_pwm_set_relative_duty_cycle(state, val[1]);
> 
> No need to set state->duty_cycle = 0 if state->enabled is false. This is
> another function I suggest to unroll as it hides more than it abstracts.
> 

Function can be unrolled, but the check for state->enabled is needed. The
device is unable to generate a 0% duty cycle, so it is turned off to
generate a 0% duty cylce.

> > +	return 0;
> > +}
> > +
> > [...]
> > +static int mc33xs2410_probe(struct spi_device *spi)
> > +{
> > [...]
> > +	/* Transition to normal mode */
> > +	ret = mc33xs2410_modify_reg(spi, MC33XS2410_GLB_CTRL,
> > +				    MC33XS2410_GLB_CTRL_MODE,
> > +				    MC33XS2410_GLB_CTRL_MODE_NORMAL);
> > +	if (ret < 0)
> > +		return dev_err_probe(dev, ret,
> > +				     "Failed to transition to normal mode\n");
> 
> What is the effect of this register write if the PWM was already setup
> by the bootloader?
> 

When its setup is done in the bootloader and the watchdog is disabled in
the bootloader it shouldn't have any impact.

> > +
> > +	ret = devm_pwmchip_add(dev, chip);
> > +	if (ret < 0)
> > +		return dev_err_probe(dev, ret, "Failed to add pwm chip\n");
> > +
> > +	return 0;
> > +}
> 
> Best regards
> Uwe

Best regards
Dimitri

Re: [PATCH v6 2/2] pwm: add support for NXPs high-side switch MC33XS2410

[Uwe Kleine-König]

[-- Attachment #1: Type: text/plain, Size: 6329 bytes --]

Hello Dimitri,

On Wed, Oct 23, 2024 at 02:52:21PM +0200, Dimitri Fedrau wrote:
> Am Tue, Oct 22, 2024 at 09:54:50AM +0200 schrieb Uwe Kleine-König:
> 
> [...]
> > > +
> > > +#define MC33XS2410_MIN_PERIOD		488282
> > > +#define MC33XS2410_MAX_PERIOD_STEP0	2000000000
> > > +/* x in { 0 ... 3 } */
> > > +#define MC33XS2410_MAX_PERIOD_STEP(x)	(MC33XS2410_MAX_PERIOD_STEP0 >> (2 * x))
> > 
> > Nitpick: These register definition become easier to parse for a human if
> > you indent the RHS of register fields one tab further and add an empty
> > line between the definitions for different registers.
> 
> Adding an empty line seems reasonable to me but the additional tab doesn't
> help me to improve readability.

OK, fine for me.

> > MC33XS2410_PWM_DC1 is only used once, I'd hard-code it into the
> > definition of MC33XS2410_PWM_DC.
> 
> Ok. Should I do the same for MC33XS2410_PWM_FREQ1 and
> MC33XS2410_MAX_PERIOD_STEP0 ?

yepp.

> > The register fields [7:4] in MC33XS2410_PWM_CTRL3 are called PWM_ON4 ..
> > PWM_ON1. So your x in { 0 ... 3 } is wrong. (Luckily, having some x
> > range over { 0 ... 3 } and others orver { 1 ... 4 } is prone to error
> > and confusion.)
> 
> Will fix it. Should I do the same for MC33XS2410_PWM_CTRL1_POL_INV ?

I guess so, otherwise you don't get consistent ranges.

> > For MC33XS2410_MAX_PERIOD_STEP maybe use a different variable name than
> > for the others. For the register definitions the range is over hwpwm
> > (which might be a good name there?), for MC33XS2410_MAX_PERIOD_STEP it's
> > about MC33XS2410_PWM_FREQ_STEP.
> 
> What about MC33XS2410_PWM_MAX_PERIOD(x) ?

Consistency is trump.

> > > +#define MC33XS2410_MAX_TRANSFERS	5
> > > +#define MC33XS2410_WORD_LEN		2
> > > +
> > > +struct mc33xs2410_pwm {
> > > +	struct spi_device *spi;
> > > +};
> > > +
> > > +static inline struct mc33xs2410_pwm *mc33xs2410_from_chip(struct pwm_chip *chip)
> > > +{
> > > +	return pwmchip_get_drvdata(chip);
> > > +}
> > > +
> > > +static int mc33xs2410_xfer_regs(struct spi_device *spi, bool read, u8 *reg,
> > > +				u16 *val, bool *ctrl, int len)
> > 
> > Unless I missed something all ctrl[x] are always identical. If so
> > represent that by a single bool.
> 
> Yes, they are identical. I added the crtl[x] to be able read from ctrl and
> diag registers. I will change it so it is represented by a single bool, if
> the feature is needed in the future I can still add it.

ack.

> > > +{
> > > +	struct spi_transfer t[MC33XS2410_MAX_TRANSFERS] = { { 0 } };
> > > +	u8 tx[MC33XS2410_MAX_TRANSFERS * MC33XS2410_WORD_LEN];
> > > +	u8 rx[MC33XS2410_MAX_TRANSFERS * MC33XS2410_WORD_LEN];
> > > +	int i, ret, reg_i, val_i;
> > > +
> > > +	if (!len)
> > > +		return 0;
> > > +
> > > +	if (read)
> > > +		len++;
> > > +
> > > +	if (len > MC33XS2410_MAX_TRANSFERS)
> > > +		return -EINVAL;
> > > +
> > > +	for (i = 0; i < len; i++) {
> > > +		reg_i = i * MC33XS2410_WORD_LEN;
> > > +		val_i = reg_i + 1;
> > > +		if (read) {
> > > +			if (i < len - 1) {
> > > +				tx[reg_i] = reg[i];
> > > +				tx[val_i] = ctrl[i] ? MC33XS2410_RD_CTRL : 0;
> > > +				t[i].tx_buf = &tx[reg_i];
> > > +			}
> > > +
> > > +			if (i > 0)
> > > +				t[i].rx_buf = &rx[reg_i - MC33XS2410_WORD_LEN];
> > > +		} else {
> > > +			tx[reg_i] = reg[i] | MC33XS2410_WR;
> > > +			tx[val_i] = val[i];
> > > +			t[i].tx_buf = &tx[reg_i];
> > > +		}
> > > +
> > > +		t[i].len = MC33XS2410_WORD_LEN;
> > > +		t[i].cs_change = 1;
> > 
> > Not sure if MC33XS2410_WORD_LEN really improves readability here.
> 
> It is used throughout in the function and improves readability overall,
> maybe not here but for consistency I would stick to it.

Seems to be subjective.

> > Why is this done using $len transfers, wouldn't a single one do (and
> > maybe be more performant and not rely on a spi controller that supports
> > cs_change)?
> 
> Without cs_change after every 16 bit, requests aren't processed by the
> device. Reading/writing from/to device fails. The SPI controller therefore
> must support cs_change. Single transfer is not possible because of the
> cs_change after every 16bit.

There is SPI_CS_WORD for this usecase.
 
> > > +	/* polarity */
> > > +	mask = MC33XS2410_PWM_CTRL1_POL_INV(pwm->hwpwm);
> > > +	val[2] = (state->polarity == PWM_POLARITY_INVERSED) ?
> > > +		 (val[2] | mask) : (val[2] & ~mask);
> > > +
> > > +	/* enable output */
> > > +	mask = MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm);
> > > +	val[3] = (state->enabled && rel_dc >= 0) ? (val[3] | mask) :
> > > +						   (val[3] & ~mask);
> > > +
> > > +	return mc33xs2410_write_regs(spi, reg, val, 4);
> > > +}
> > > +
> > > +static int mc33xs2410_pwm_get_state(struct pwm_chip *chip,
> > > +				    struct pwm_device *pwm,
> > > +				    struct pwm_state *state)
> > > +{
> > > [...]
> > > +	state->period = mc33xs2410_pwm_get_period(val[0]);
> > > +	state->polarity = (val[2] & MC33XS2410_PWM_CTRL1_POL_INV(pwm->hwpwm)) ?
> > > +			  PWM_POLARITY_INVERSED : PWM_POLARITY_NORMAL;
> > > +	state->enabled = !!(val[3] & MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm));
> > > +	mc33xs2410_pwm_set_relative_duty_cycle(state, val[1]);
> > 
> > No need to set state->duty_cycle = 0 if state->enabled is false. This is
> > another function I suggest to unroll as it hides more than it abstracts.
> 
> Function can be unrolled, but the check for state->enabled is needed. The
> device is unable to generate a 0% duty cycle, so it is turned off to
> generate a 0% duty cylce.

What breaks if you drop the check for state->enabled?
 
> > > [...]
> > > +static int mc33xs2410_probe(struct spi_device *spi)
> > > +{
> > > [...]
> > > +	/* Transition to normal mode */
> > > +	ret = mc33xs2410_modify_reg(spi, MC33XS2410_GLB_CTRL,
> > > +				    MC33XS2410_GLB_CTRL_MODE,
> > > +				    MC33XS2410_GLB_CTRL_MODE_NORMAL);
> > > +	if (ret < 0)
> > > +		return dev_err_probe(dev, ret,
> > > +				     "Failed to transition to normal mode\n");
> > 
> > What is the effect of this register write if the PWM was already setup
> > by the bootloader?
> > 
> 
> When its setup is done in the bootloader and the watchdog is disabled in
> the bootloader it shouldn't have any impact.

ok.

Best regards
Uwe

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v6 2/2] pwm: add support for NXPs high-side switch MC33XS2410

[Dimitri Fedrau]

Hello Uwe,

Am Thu, Oct 24, 2024 at 11:19:16PM +0200 schrieb Uwe Kleine-König:
> Hello Dimitri,
> 
> On Wed, Oct 23, 2024 at 02:52:21PM +0200, Dimitri Fedrau wrote:
> > Am Tue, Oct 22, 2024 at 09:54:50AM +0200 schrieb Uwe Kleine-König:
> > > > +{
> > > > +	struct spi_transfer t[MC33XS2410_MAX_TRANSFERS] = { { 0 } };
> > > > +	u8 tx[MC33XS2410_MAX_TRANSFERS * MC33XS2410_WORD_LEN];
> > > > +	u8 rx[MC33XS2410_MAX_TRANSFERS * MC33XS2410_WORD_LEN];
> > > > +	int i, ret, reg_i, val_i;
> > > > +
> > > > +	if (!len)
> > > > +		return 0;
> > > > +
> > > > +	if (read)
> > > > +		len++;
> > > > +
> > > > +	if (len > MC33XS2410_MAX_TRANSFERS)
> > > > +		return -EINVAL;
> > > > +
> > > > +	for (i = 0; i < len; i++) {
> > > > +		reg_i = i * MC33XS2410_WORD_LEN;
> > > > +		val_i = reg_i + 1;
> > > > +		if (read) {
> > > > +			if (i < len - 1) {
> > > > +				tx[reg_i] = reg[i];
> > > > +				tx[val_i] = ctrl[i] ? MC33XS2410_RD_CTRL : 0;
> > > > +				t[i].tx_buf = &tx[reg_i];
> > > > +			}
> > > > +
> > > > +			if (i > 0)
> > > > +				t[i].rx_buf = &rx[reg_i - MC33XS2410_WORD_LEN];
> > > > +		} else {
> > > > +			tx[reg_i] = reg[i] | MC33XS2410_WR;
> > > > +			tx[val_i] = val[i];
> > > > +			t[i].tx_buf = &tx[reg_i];
> > > > +		}
> > > > +
> > > > +		t[i].len = MC33XS2410_WORD_LEN;
> > > > +		t[i].cs_change = 1;
> > > 
> > > Not sure if MC33XS2410_WORD_LEN really improves readability here.
> > 
> > It is used throughout in the function and improves readability overall,
> > maybe not here but for consistency I would stick to it.
> 
> Seems to be subjective.
>

I will get rid of it. Due to your proposal below, to use SPI_CS_WORD, the
code to read/write from/to the device can be simplified by using a single
transaction.

> > > Why is this done using $len transfers, wouldn't a single one do (and
> > > maybe be more performant and not rely on a spi controller that supports
> > > cs_change)?
> > 
> > Without cs_change after every 16 bit, requests aren't processed by the
> > device. Reading/writing from/to device fails. The SPI controller therefore
> > must support cs_change. Single transfer is not possible because of the
> > cs_change after every 16bit.
> 
> There is SPI_CS_WORD for this usecase.
>
Thanks, didn't know about it. Helps a lot to simplify the code to
read/write from/to device. Will switch to 16 bits per word and use
SPI_CS_WORD.

> > > > +	/* polarity */
> > > > +	mask = MC33XS2410_PWM_CTRL1_POL_INV(pwm->hwpwm);
> > > > +	val[2] = (state->polarity == PWM_POLARITY_INVERSED) ?
> > > > +		 (val[2] | mask) : (val[2] & ~mask);
> > > > +
> > > > +	/* enable output */
> > > > +	mask = MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm);
> > > > +	val[3] = (state->enabled && rel_dc >= 0) ? (val[3] | mask) :
> > > > +						   (val[3] & ~mask);
> > > > +
> > > > +	return mc33xs2410_write_regs(spi, reg, val, 4);
> > > > +}
> > > > +
> > > > +static int mc33xs2410_pwm_get_state(struct pwm_chip *chip,
> > > > +				    struct pwm_device *pwm,
> > > > +				    struct pwm_state *state)
> > > > +{
> > > > [...]
> > > > +	state->period = mc33xs2410_pwm_get_period(val[0]);
> > > > +	state->polarity = (val[2] & MC33XS2410_PWM_CTRL1_POL_INV(pwm->hwpwm)) ?
> > > > +			  PWM_POLARITY_INVERSED : PWM_POLARITY_NORMAL;
> > > > +	state->enabled = !!(val[3] & MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm));
> > > > +	mc33xs2410_pwm_set_relative_duty_cycle(state, val[1]);
> > > 
> > > No need to set state->duty_cycle = 0 if state->enabled is false. This is
> > > another function I suggest to unroll as it hides more than it abstracts.
> > 
> > Function can be unrolled, but the check for state->enabled is needed. The
> > device is unable to generate a 0% duty cycle, so it is turned off to
> > generate a 0% duty cylce.
> 
> What breaks if you drop the check for state->enabled?
>  
The device is unable to generate a 0% duty cycle, to support this you
proposed in an earlier review to disable the output. Without checking if
the output is disabled, the mc33xs2410_pwm_get_state function returns the
wrong duty cycle for a previously setted 0% duty cycle. A "0" value in the
MC33XS2410_PWM_DC register means that the relative duty cylce is 1/256. As
a result there are complaints if PWM_DEBUG is enabled.

Best regards,
Dimitri

Re: [PATCH v6 2/2] pwm: add support for NXPs high-side switch MC33XS2410

[Uwe Kleine-König]

[-- Attachment #1: Type: text/plain, Size: 989 bytes --]

Hello Dimitri,

On Sun, Nov 03, 2024 at 08:07:09PM +0100, Dimitri Fedrau wrote:
> Am Thu, Oct 24, 2024 at 11:19:16PM +0200 schrieb Uwe Kleine-König:
> > What breaks if you drop the check for state->enabled?
> >  
> The device is unable to generate a 0% duty cycle, to support this you
> proposed in an earlier review to disable the output. Without checking if
> the output is disabled, the mc33xs2410_pwm_get_state function returns the
> wrong duty cycle for a previously setted 0% duty cycle. A "0" value in the
> MC33XS2410_PWM_DC register means that the relative duty cylce is 1/256. As
> a result there are complaints if PWM_DEBUG is enabled.

I fail to follow. If .enabled=true + .duty_cycle=0 is requested you
disable. That's fine. However it shouldn't be necessary to use
state->enabled in .get_state(). I didn't look at the actual code, but if
you provide a sequence of writes to /sys that trigger a PWM_DEBUG
output, I'll take another look.

Best regards
Uwe

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v6 2/2] pwm: add support for NXPs high-side switch MC33XS2410

[Dimitri Fedrau]

Hello Uwe,

Am Sun, Nov 03, 2024 at 09:19:36PM +0100 schrieb Uwe Kleine-König:
> Hello Dimitri,
> 
> On Sun, Nov 03, 2024 at 08:07:09PM +0100, Dimitri Fedrau wrote:
> > Am Thu, Oct 24, 2024 at 11:19:16PM +0200 schrieb Uwe Kleine-König:
> > > What breaks if you drop the check for state->enabled?
> > >  
> > The device is unable to generate a 0% duty cycle, to support this you
> > proposed in an earlier review to disable the output. Without checking if
> > the output is disabled, the mc33xs2410_pwm_get_state function returns the
> > wrong duty cycle for a previously setted 0% duty cycle. A "0" value in the
> > MC33XS2410_PWM_DC register means that the relative duty cylce is 1/256. As
> > a result there are complaints if PWM_DEBUG is enabled.
> 
> I fail to follow. If .enabled=true + .duty_cycle=0 is requested you
> disable. That's fine. However it shouldn't be necessary to use
> state->enabled in .get_state(). I didn't look at the actual code, but if
> you provide a sequence of writes to /sys that trigger a PWM_DEBUG
> output, I'll take another look.
> 
Apply 0% duty cycle: .enabled=false + .duty_cycle=0
Below some writes triggering PWM_DEBUG output:

# echo 488282 > /sys/class/pwm/pwmchip3/pwm0/period
# echo 244140 > /sys/class/pwm/pwmchip3/pwm0/duty_cycle
# echo 0 > /sys/class/pwm/pwmchip3/pwm0/duty_cycle
[   91.813513] mc33xs2410-pwm spi0.0: .apply is supposed to round down duty_cycle (requested: 0/488282, applied: 1908/488282)

Best regards,
Dimitri

Re: [PATCH v6 2/2] pwm: add support for NXPs high-side switch MC33XS2410

[Uwe Kleine-König]

[-- Attachment #1: Type: text/plain, Size: 3164 bytes --]

On Sun, Nov 03, 2024 at 09:52:15PM +0100, Dimitri Fedrau wrote:
> Hello Uwe,
> 
> Am Sun, Nov 03, 2024 at 09:19:36PM +0100 schrieb Uwe Kleine-König:
> > Hello Dimitri,
> > 
> > On Sun, Nov 03, 2024 at 08:07:09PM +0100, Dimitri Fedrau wrote:
> > > Am Thu, Oct 24, 2024 at 11:19:16PM +0200 schrieb Uwe Kleine-König:
> > > > What breaks if you drop the check for state->enabled?
> > > >  
> > > The device is unable to generate a 0% duty cycle, to support this you
> > > proposed in an earlier review to disable the output. Without checking if
> > > the output is disabled, the mc33xs2410_pwm_get_state function returns the
> > > wrong duty cycle for a previously setted 0% duty cycle. A "0" value in the
> > > MC33XS2410_PWM_DC register means that the relative duty cylce is 1/256. As
> > > a result there are complaints if PWM_DEBUG is enabled.
> > 
> > I fail to follow. If .enabled=true + .duty_cycle=0 is requested you
> > disable. That's fine. However it shouldn't be necessary to use
> > state->enabled in .get_state(). I didn't look at the actual code, but if
> > you provide a sequence of writes to /sys that trigger a PWM_DEBUG
> > output, I'll take another look.
> > 
> Apply 0% duty cycle: .enabled=false + .duty_cycle=0
> Below some writes triggering PWM_DEBUG output:
> 
> # echo 488282 > /sys/class/pwm/pwmchip3/pwm0/period
> # echo 244140 > /sys/class/pwm/pwmchip3/pwm0/duty_cycle
> # echo 0 > /sys/class/pwm/pwmchip3/pwm0/duty_cycle
> [   91.813513] mc33xs2410-pwm spi0.0: .apply is supposed to round down duty_cycle (requested: 0/488282, applied: 1908/488282)

I don't understand that. We're talking about 

diff --git a/drivers/pwm/pwm-mc33xs2410.c b/drivers/pwm/pwm-mc33xs2410.c
index f9a334a5e69b..14f5f7312d0a 100644
--- a/drivers/pwm/pwm-mc33xs2410.c
+++ b/drivers/pwm/pwm-mc33xs2410.c
@@ -244,15 +244,6 @@ static int mc33xs2410_pwm_get_relative_duty_cycle(u64 period, u64 duty_cycle)
 	return duty_cycle - 1;
 }
 
-static void mc33xs2410_pwm_set_relative_duty_cycle(struct pwm_state *state,
-						   u16 duty_cycle)
-{
-	if (!state->enabled)
-		state->duty_cycle = 0;
-	else
-		state->duty_cycle = DIV_ROUND_UP_ULL((duty_cycle + 1) * state->period, 256);
-}
-
 static int mc33xs2410_pwm_apply(struct pwm_chip *chip, struct pwm_device *pwm,
 				const struct pwm_state *state)
 {
@@ -325,7 +316,7 @@ static int mc33xs2410_pwm_get_state(struct pwm_chip *chip,
 	state->polarity = (val[2] & MC33XS2410_PWM_CTRL1_POL_INV(pwm->hwpwm)) ?
 			  PWM_POLARITY_INVERSED : PWM_POLARITY_NORMAL;
 	state->enabled = !!(val[3] & MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm));
-	mc33xs2410_pwm_set_relative_duty_cycle(state, val[1]);
+	state->duty_cycle = DIV_ROUND_UP_ULL((duty_cycle + 1) * state->period, 256);
 	return 0;
 }
 
on top of your patch, right?

`echo 0 > /sys/class/pwm/pwmchip3/pwm0/duty_cycle` should result in
MC33XS2410_PWM_CTRL3 having MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm) cleared.
When mc33xs2410_pwm_get_state() is called then it returns state->enabled
= false and in that case the above mentioned warning doesn't trigger.

Where is the misunderstanding?

Best regards
Uwe

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: [PATCH v6 2/2] pwm: add support for NXPs high-side switch MC33XS2410

[Dimitri Fedrau]

Am Mon, Nov 04, 2024 at 09:52:51AM +0100 schrieb Uwe Kleine-König:
> On Sun, Nov 03, 2024 at 09:52:15PM +0100, Dimitri Fedrau wrote:
> > Hello Uwe,
> > 
> > Am Sun, Nov 03, 2024 at 09:19:36PM +0100 schrieb Uwe Kleine-König:
> > > Hello Dimitri,
> > > 
> > > On Sun, Nov 03, 2024 at 08:07:09PM +0100, Dimitri Fedrau wrote:
> > > > Am Thu, Oct 24, 2024 at 11:19:16PM +0200 schrieb Uwe Kleine-König:
> > > > > What breaks if you drop the check for state->enabled?
> > > > >  
> > > > The device is unable to generate a 0% duty cycle, to support this you
> > > > proposed in an earlier review to disable the output. Without checking if
> > > > the output is disabled, the mc33xs2410_pwm_get_state function returns the
> > > > wrong duty cycle for a previously setted 0% duty cycle. A "0" value in the
> > > > MC33XS2410_PWM_DC register means that the relative duty cylce is 1/256. As
> > > > a result there are complaints if PWM_DEBUG is enabled.
> > > 
> > > I fail to follow. If .enabled=true + .duty_cycle=0 is requested you
> > > disable. That's fine. However it shouldn't be necessary to use
> > > state->enabled in .get_state(). I didn't look at the actual code, but if
> > > you provide a sequence of writes to /sys that trigger a PWM_DEBUG
> > > output, I'll take another look.
> > > 
> > Apply 0% duty cycle: .enabled=false + .duty_cycle=0
> > Below some writes triggering PWM_DEBUG output:
> > 
> > # echo 488282 > /sys/class/pwm/pwmchip3/pwm0/period
> > # echo 244140 > /sys/class/pwm/pwmchip3/pwm0/duty_cycle
> > # echo 0 > /sys/class/pwm/pwmchip3/pwm0/duty_cycle
> > [   91.813513] mc33xs2410-pwm spi0.0: .apply is supposed to round down duty_cycle (requested: 0/488282, applied: 1908/488282)
> 
> I don't understand that. We're talking about 
> 
> diff --git a/drivers/pwm/pwm-mc33xs2410.c b/drivers/pwm/pwm-mc33xs2410.c
> index f9a334a5e69b..14f5f7312d0a 100644
> --- a/drivers/pwm/pwm-mc33xs2410.c
> +++ b/drivers/pwm/pwm-mc33xs2410.c
> @@ -244,15 +244,6 @@ static int mc33xs2410_pwm_get_relative_duty_cycle(u64 period, u64 duty_cycle)
>  	return duty_cycle - 1;
>  }
>  
> -static void mc33xs2410_pwm_set_relative_duty_cycle(struct pwm_state *state,
> -						   u16 duty_cycle)
> -{
> -	if (!state->enabled)
> -		state->duty_cycle = 0;
> -	else
> -		state->duty_cycle = DIV_ROUND_UP_ULL((duty_cycle + 1) * state->period, 256);
> -}
> -
>  static int mc33xs2410_pwm_apply(struct pwm_chip *chip, struct pwm_device *pwm,
>  				const struct pwm_state *state)
>  {
> @@ -325,7 +316,7 @@ static int mc33xs2410_pwm_get_state(struct pwm_chip *chip,
>  	state->polarity = (val[2] & MC33XS2410_PWM_CTRL1_POL_INV(pwm->hwpwm)) ?
>  			  PWM_POLARITY_INVERSED : PWM_POLARITY_NORMAL;
>  	state->enabled = !!(val[3] & MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm));
> -	mc33xs2410_pwm_set_relative_duty_cycle(state, val[1]);
> +	state->duty_cycle = DIV_ROUND_UP_ULL((duty_cycle + 1) * state->period, 256);
>  	return 0;
>  }
>  
> on top of your patch, right?
>
Yes.

> `echo 0 > /sys/class/pwm/pwmchip3/pwm0/duty_cycle` should result in
> MC33XS2410_PWM_CTRL3 having MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm) cleared.
> When mc33xs2410_pwm_get_state() is called then it returns state->enabled
> = false and in that case the above mentioned warning doesn't trigger.
>
Yes, as you explained. But the warning is shown.

> Where is the misunderstanding?
>

if (state->enabled && state->duty_cycle < s2.duty_cycle)
	dev_warn(pwmchip_parent(chip),
		".apply is supposed to round down duty_cycle (requested: %llu/%llu, applied: %llu/%llu)\n",
		state->duty_cycle, state->period,
		s2.duty_cycle, s2.period);

state has previously applied settings and is parameter of pwm_apply_debug,
in that case s2=s1, and s1 is returned by get_state:

state->enabled=true
state->duty_cycle=0
s2.enabled=false
s2.duty_cycle=1908

Due to the code the warning should be raised. If it shouldn't the check
should be different, something like if (state->enabled && s2.enabled &&
...)

Best regards,
Dimitri

Re: [PATCH v6 2/2] pwm: add support for NXPs high-side switch MC33XS2410

[Uwe Kleine-König]

[-- Attachment #1: Type: text/plain, Size: 1376 bytes --]

On Mon, Nov 04, 2024 at 02:07:53PM +0100, Dimitri Fedrau wrote:
> Am Mon, Nov 04, 2024 at 09:52:51AM +0100 schrieb Uwe Kleine-König:
> > `echo 0 > /sys/class/pwm/pwmchip3/pwm0/duty_cycle` should result in
> > MC33XS2410_PWM_CTRL3 having MC33XS2410_PWM_CTRL3_EN(pwm->hwpwm) cleared.
> > When mc33xs2410_pwm_get_state() is called then it returns state->enabled
> > = false and in that case the above mentioned warning doesn't trigger.
>
> Yes, as you explained. But the warning is shown.
> 
> > Where is the misunderstanding?
> 
> if (state->enabled && state->duty_cycle < s2.duty_cycle)
> 	dev_warn(pwmchip_parent(chip),
> 		".apply is supposed to round down duty_cycle (requested: %llu/%llu, applied: %llu/%llu)\n",
> 		state->duty_cycle, state->period,
> 		s2.duty_cycle, s2.period);
> 
> state has previously applied settings and is parameter of pwm_apply_debug,
> in that case s2=s1, and s1 is returned by get_state:
> 
> state->enabled=true
> state->duty_cycle=0
> s2.enabled=false
> s2.duty_cycle=1908
> 
> Due to the code the warning should be raised. If it shouldn't the check
> should be different, something like if (state->enabled && s2.enabled &&
> ...)

This should be fixed with https://lore.kernel.org/linux-pwm/20241105153521.1001864-2-u.kleine-koenig@baylibre.com/T/#u
Test feedback and review very welcome.

Best regards
Uwe

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]