Skip to contentSkip to conversation

add support for Rockchip SARADC

PATCH Superseded Michael Riesch 2 participants 6 messages Jun 23, 2021 – Jun 24, 2021

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover add support for Rockchip SARADC

Michael Riesch (2):
aiodev: add driver for Rockchip SARADC
arm: rk3568-evb1: add support for Rockchip SARADC

 arch/arm/dts/rk3568-evb1-v10.dts |   5 +
 arch/arm/dts/rk3568.dtsi         |  12 ++
 drivers/aiodev/Kconfig           |   7 ++
 drivers/aiodev/Makefile          |   1 +
 drivers/aiodev/rockchip_saradc.c | 206 +++++++++++++++++++++++++++++++
 5 files changed, 231 insertions(+)
 create mode 100644 drivers/aiodev/rockchip_saradc.c

--
2.20.1

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

#1 lore raw

PATCH v2 1/2 aiodev: add driver for Rockchip SARADC

This commit adds support for the Successive Approximation Register (SAR) ADCs that can be found in Rockchip SoCs, such as the RK3568.

---
v2:
- fix timeout loop
- remove spurious debug output
- revise reference voltage handling
- add handling of clocks

 drivers/aiodev/Kconfig           |   7 ++
 drivers/aiodev/Makefile          |   1 +
 drivers/aiodev/rockchip_saradc.c | 206 +++++++++++++++++++++++++++++++
 3 files changed, 214 insertions(+)
 create mode 100644 drivers/aiodev/rockchip_saradc.c
drivers/aiodev/Kconfig +7 -0
index 88d013aad..98223a8f9 100644
@@ -50,4 +50,11 @@ config STM32_ADC
 	  Support for ADC on STM32.  Supports simple one-shot readings
 	  rather than continuous sampling with DMA, etc.  ADC channels should be
 	  configured via device tree, using the kernel bindings.
+
+config ROCKCHIP_SARADC
+	tristate "Rockchip SARADC driver"
+	depends on ARCH_RK3568
+	help
+	  Support for Successive Approximation Register (SAR) ADC in Rockchip
+	  SoCs.
 endif
drivers/aiodev/Makefile +1 -0
index 52652f67b..1b480f6fa 100644
@@ -6,3 +6,4 @@ obj-$(CONFIG_MC13XXX_ADC) += mc13xxx_adc.o
 obj-$(CONFIG_QORIQ_THERMAL) += qoriq_thermal.o
 obj-$(CONFIG_AM335X_ADC) += am335x_adc.o
 obj-$(CONFIG_STM32_ADC) += stm32-adc.o stm32-adc-core.o
+obj-$(CONFIG_ROCKCHIP_SARADC) += rockchip_saradc.o
drivers/aiodev/rockchip_saradc.c new file +206 -0

206 lines not on this page · raw message

--
2.20.1

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

#2 lore raw

Re: [PATCH v2 1/2] aiodev: add driver for Rockchip SARADC

Hi,

Quoted 22 lines · Michael Riesch

This commit adds support for the Successive Approximation Register (SAR) ADCs that can be found in Rockchip SoCs, such as the RK3568.

---
v2:
- fix timeout loop
- remove spurious debug output
- revise reference voltage handling
- add handling of clocks

 drivers/aiodev/Kconfig           |   7 ++
 drivers/aiodev/Makefile          |   1 +
 drivers/aiodev/rockchip_saradc.c | 206 +++++++++++++++++++++++++++++++
 3 files changed, 214 insertions(+)
 create mode 100644 drivers/aiodev/rockchip_saradc.c
drivers/aiodev/Kconfig +0 -0
index 88d013aad..98223a8f9 100644

drivers/aiodev/Kconfig config STM32_ADC · line 56

Quoted 7 lines
drivers/aiodev/Kconfig +4 -0
@@ -50,4 +50,11 @@ config STM32_ADC
 	  Support for ADC on STM32.  Supports simple one-shot readings
 	  rather than continuous sampling with DMA, etc.  ADC channels should be
 	  configured via device tree, using the kernel bindings.
+
+config ROCKCHIP_SARADC
+	tristate "Rockchip SARADC driver"
+	depends on ARCH_RK3568

Please change to depends on ARCH_RK3568 || COMPILE_TEST depends on OFDEVICE

That way, this driver can be built with static analyzers for other architectures as well.

Quoted 33 lines
drivers/aiodev/Kconfig +3 -0
+	help
+	  Support for Successive Approximation Register (SAR) ADC in Rockchip
+	  SoCs.
 endif
drivers/aiodev/Makefile +1 -0
index 52652f67b..1b480f6fa 100644
@@ -6,3 +6,4 @@ obj-$(CONFIG_MC13XXX_ADC) += mc13xxx_adc.o
 obj-$(CONFIG_QORIQ_THERMAL) += qoriq_thermal.o
 obj-$(CONFIG_AM335X_ADC) += am335x_adc.o
 obj-$(CONFIG_STM32_ADC) += stm32-adc.o stm32-adc-core.o
+obj-$(CONFIG_ROCKCHIP_SARADC) += rockchip_saradc.o
drivers/aiodev/rockchip_saradc.c new file +14 -0
new file mode 100644
index 000000000..ae842eb1f
@@ -0,0 +1,206 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2021, WolfVision GmbH
+ * Author: Michael Riesch <[email protected]>
+ *
+ * Originally based on the Linux kernel v5.12 drivers/iio/adc/rockchip-saradc.c.
+ */
+
+#include <common.h>
+#include <aiodev.h>
+#include <linux/clk.h>
+#include <regulator.h>
+
+#define SARADC_DATA	       0x00

drivers/aiodev/rockchip_saradc.c line 24

Quoted 10 lines
drivers/aiodev/rockchip_saradc.c new file +10 -0
+
+#define SARADC_CTRL	       0x08
+#define SARADC_CTRL_IRQ_STATUS (1 << 6)
+#define SARADC_CTRL_IRQ_ENABLE (1 << 5)
+#define SARADC_CTRL_POWER_CTRL (1 << 3)
+#define SARADC_CTRL_CHN_MASK   0x07
+
+#define SARADC_DLY_PU_SOC      0x0c
+
+#define SARADC_TIMEOUT	       100 /* ms */

Nitpick: SARADC_TIMEOUT_NS (100 * MSECOND)

Makes it harder to misuse.

Quoted 29 lines
drivers/aiodev/rockchip_saradc.c new file +29 -0
+
+struct rockchip_saradc_cfg {
+	unsigned int num_bits;
+	unsigned int num_channels;
+};
+
+struct rockchip_saradc_data {
+	const struct rockchip_saradc_cfg *config;
+	void __iomem *base;
+	struct regulator *vref;
+	unsigned int ref_voltage_mv;
+	struct clk *cclk;
+	struct clk *pclk;
+	struct aiodevice aiodev;
+	struct aiochannel *channels;
+};
+
+static inline void rockchip_saradc_reg_wr(struct rockchip_saradc_data *data,
+					  u32 value, u32 reg)
+{
+	writel(value, data->base + reg);
+}
+
+static inline u32 rockchip_saradc_reg_rd(struct rockchip_saradc_data *data,
+					 u32 reg)
+{
+	return readl(data->base + reg);
+}
+

drivers/aiodev/rockchip_saradc.c line 63

Quoted 10 lines
drivers/aiodev/rockchip_saradc.c new file +10 -0
+static int rockchip_saradc_read(struct aiochannel *chan, int *val)
+{
+	struct rockchip_saradc_data *data;
+	u32 value = 0;
+	u32 control = 0;
+	u32 mask;
+	u64 start;
+
+	if (!chan || !val)
+		return -EINVAL;

You don't ever see chan == NULL here, so the check isn't needed.
A buggy aiodev user can indeed pass val == NULL, but they could pass (void *)1 as well and you would be none the wiser. It's expected, you are passed valid pointers here, so the checks can be dropped.

drivers/aiodev/rockchip_saradc.c line 67

Quoted 4 lines
drivers/aiodev/rockchip_saradc.c new file +4 -0
+
+	data = container_of(chan->aiodev, struct rockchip_saradc_data, aiodev);
+	if (!data)
+		return -EINVAL;

container_of can't fail. It's just pointer arithmetic.

Quoted 23 lines
drivers/aiodev/rockchip_saradc.c new file +23 -0
+
+	rockchip_saradc_reg_wr(data, 8, SARADC_DLY_PU_SOC);
+	rockchip_saradc_reg_wr(data,
+			       (chan->index & SARADC_CTRL_CHN_MASK) |
+				       SARADC_CTRL_IRQ_ENABLE |
+				       SARADC_CTRL_POWER_CTRL,
+			       SARADC_CTRL);
+
+	start = get_time_ns();
+	do {
+		control = rockchip_saradc_reg_rd(data, SARADC_CTRL);
+
+		if (is_timeout(start, SARADC_TIMEOUT * MSECOND))
+			return -ETIMEDOUT;
+	} while (!(control & SARADC_CTRL_IRQ_STATUS));
+
+	mask = (1 << data->config->num_bits) - 1;
+	value = rockchip_saradc_reg_rd(data, SARADC_DATA) & mask;
+	rockchip_saradc_reg_wr(data, 0, SARADC_CTRL);
+
+	*val = (value * data->ref_voltage_mv) / mask;
+
+	return 0;

drivers/aiodev/rockchip_saradc.c line 100

Quoted 10 lines
drivers/aiodev/rockchip_saradc.c new file +10 -0
+}
+
+static int rockchip_saradc_probe(struct device_d *dev)
+{
+	struct rockchip_saradc_data *data;
+	int i, ret;
+
+	data = xzalloc(sizeof(struct rockchip_saradc_data));
+	if (!data)
+		return -ENOMEM;

Allocator functions starting with x panic barebox, so data is always != NULL if they succeed. For big, frequent allocations calloc can be used instead, which also zeroes, but fails with return NULL. For small one-time allocations xzalloc is ok, but there is no point in checking its return value.

drivers/aiodev/rockchip_saradc.c line 103

Quoted 3 lines
drivers/aiodev/rockchip_saradc.c new file +3 -0
+
+	data->config = device_get_match_data(dev);
+	if (!data->config) {

Nitpick: This can't ever happen.

Quoted 7 lines
drivers/aiodev/rockchip_saradc.c new file +7 -0
+		ret = -EINVAL;
+		goto fail_data;
+	}
+	data->aiodev.hwdev = dev;
+	data->aiodev.read = rockchip_saradc_read;
+
+	data->base = dev_request_mem_region(dev, 0);

drivers/aiodev/rockchip_saradc.c line 120

Quoted 10 lines
drivers/aiodev/rockchip_saradc.c new file +10 -0
+	if (IS_ERR(data->base)) {
+		ret = PTR_ERR(data->base);
+		goto fail_data;
+	}
+
+	data->vref = regulator_get(dev, "vref");
+	if (IS_ERR(data->vref)) {
+		dev_err(dev, "can't get vref-supply: %pe\n", data->vref);
+		ret = PTR_ERR(data->vref);
+		goto fail_data;

Nitpick: You don't release the memory region you requested above. So if you fail to get the regulator here and the probe is deferred, you won't be able to reach this point here again as the second request for memory would fail without releasing the first. Quite a few drivers don't care about this failure mode, but given that you take care to free your allocations, you may want to revert the others operations you do as well.

Quoted 29 lines
drivers/aiodev/rockchip_saradc.c new file +29 -0
+	}
+
+	ret = regulator_enable(data->vref);
+	if (ret < 0) {
+		dev_err(dev, "can't enable vref-supply value: %d\n", ret);
+		goto fail_data;
+	}
+
+	ret = regulator_get_voltage(data->vref);
+	if (ret < 0) {
+		dev_warn(dev, "can't get vref-supply value: %d\n", ret);
+		/* use default value as fallback */
+		ret = 1800000;
+	}
+	data->ref_voltage_mv = ret / 1000;
+
+	data->cclk = clk_get(dev, "saradc");
+	if (IS_ERR(data->cclk)) {
+		dev_err(dev, "can't get converter clock: %pe\n", data->cclk);
+		ret = PTR_ERR(data->cclk);
+		goto fail_data;
+	}
+
+	ret = clk_enable(data->cclk);
+	if (ret < 0) {
+		dev_err(dev, "can't enable converter clock: %d\n", ret);
+		goto fail_data;
+	}
+

drivers/aiodev/rockchip_saradc.c line 159

Quoted 10 lines
drivers/aiodev/rockchip_saradc.c new file +10 -0
+	data->pclk = clk_get(dev, "apb_pclk");
+	if (IS_ERR(data->pclk)) {
+		dev_err(dev, "can't get peripheral clock: %pe\n", data->pclk);
+		ret = PTR_ERR(data->pclk);
+		goto fail_data;
+	}
+
+	ret = clk_enable(data->pclk);
+	if (ret < 0) {
+		dev_err(dev, "can't enable peripheral clk: %d\n", ret);

FYI, you can have ERR_PTR(ret) here with %pe, to get an error string.

drivers/aiodev/rockchip_saradc.c line 161

Quoted 2 lines
drivers/aiodev/rockchip_saradc.c new file +2 -0
+		goto fail_data;
+	}

Nitpick: There is also clk_bulk_get and clk_bulk_enable to handle multiple clocks at once, but for only two clocks, it's ok to do it separately.

Quoted 53 lines
drivers/aiodev/rockchip_saradc.c new file +45 -1
+
+	data->aiodev.num_channels = data->config->num_channels;
+	data->channels =
+		xzalloc(sizeof(*data->channels) * data->aiodev.num_channels);
+	data->aiodev.channels = xmalloc(sizeof(*data->aiodev.channels) *
+					data->aiodev.num_channels);
+	for (i = 0; i < data->aiodev.num_channels; i++) {
+		data->aiodev.channels[i] = &data->channels[i];
+		data->channels[i].unit = "mV";
+	}
+
+	rockchip_saradc_reg_wr(data, 0, SARADC_CTRL);
+
+	ret = aiodevice_register(&data->aiodev);
+	if (ret)
+		goto fail_channels;
+
+	dev_info(dev, "registered as %s\n", dev_name(&data->aiodev.dev));
+	return 0;
+
+fail_channels:
+	kfree(data->channels);
+	kfree(data->aiodev.channels);
+
+fail_data:
+	kfree(data);
+	return ret;
+}
+
+static const struct rockchip_saradc_cfg rk3568_saradc_cfg = {
+	.num_bits = 10,
+	.num_channels = 8,
+};
+
+static const struct of_device_id of_rockchip_saradc_match[] = {
+	{ .compatible = "rockchip,rk3568-saradc", .data = &rk3568_saradc_cfg },
+	{ /* end */ }
+};
+
+static struct driver_d rockchip_saradc_driver = {
+	.name = "rockchip_saradc",
+	.probe = rockchip_saradc_probe,
+	.of_compatible = DRV_OF_COMPAT(of_rockchip_saradc_match),
+};
+device_platform_driver(rockchip_saradc_driver);
--

2.20.1

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

Signature · 9 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#3 lore raw

PATCH v2 2/2 arm: rk3568-evb1: add support for Rockchip SARADC

In reply to Michael Riesch

---
v2:
- add initial version

 arch/arm/dts/rk3568-evb1-v10.dts |  5 +++++
 arch/arm/dts/rk3568.dtsi         | 12 ++++++++++++
 2 files changed, 17 insertions(+)
arch/arm/dts/rk3568-evb1-v10.dts +5 -0
index ca6f9c280..b8b4c85bf 100644
@@ -485,3 +485,8 @@
 &uart2 {
 	status = "okay";
 };
+
+&saradc {
+	vref-supply = <&vcca_1v8>;
+	status = "okay";
+};
arch/arm/dts/rk3568.dtsi +12 -0
index 34d0b8a08..8a75a7e96 100644
@@ -817,6 +817,18 @@
 		status = "disabled";
 	};
 
+	saradc: saradc@fe720000 {
+		compatible = "rockchip,rk3568-saradc";
+		reg = <0x0 0xfe720000 0x0 0x100>;
+		interrupts = <GIC_SPI 93 IRQ_TYPE_LEVEL_HIGH>;
+		clocks = <&cru CLK_SARADC>, <&cru PCLK_SARADC>;
+		clock-names = "saradc", "apb_pclk";
+		resets = <&cru SRST_P_SARADC>;
+		reset-names = "saradc-apb";
+		#io-channel-cells = <1>;
+		status = "disabled";
+	};
+
 	pinctrl: pinctrl {
 		compatible = "rockchip,rk3568-pinctrl";
 		rockchip,grf = <&grf>;

--
2.20.1

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

#4 lore raw

Re: [PATCH v2 2/2] arm: rk3568-evb1: add support for Rockchip SARADC

Quoted 28 lines · Michael Riesch

---
v2:
- add initial version

 arch/arm/dts/rk3568-evb1-v10.dts |  5 +++++
 arch/arm/dts/rk3568.dtsi         | 12 ++++++++++++
 2 files changed, 17 insertions(+)
arch/arm/dts/rk3568-evb1-v10.dts +5 -0
index ca6f9c280..b8b4c85bf 100644
@@ -485,3 +485,8 @@
 &uart2 {
 	status = "okay";
 };
+
+&saradc {
+	vref-supply = <&vcca_1v8>;
+	status = "okay";
+};
arch/arm/dts/rk3568.dtsi +0 -0
index 34d0b8a08..8a75a7e96 100644
@@ -817,6 +817,18 @@
 		status = "disabled";

arch/arm/dts/rk3568.dtsi line 827

Quoted 10 lines
arch/arm/dts/rk3568.dtsi +8 -0
 	};
 
+	saradc: saradc@fe720000 {
+		compatible = "rockchip,rk3568-saradc";
+		reg = <0x0 0xfe720000 0x0 0x100>;
+		interrupts = <GIC_SPI 93 IRQ_TYPE_LEVEL_HIGH>;
+		clocks = <&cru CLK_SARADC>, <&cru PCLK_SARADC>;
+		clock-names = "saradc", "apb_pclk";
+		resets = <&cru SRST_P_SARADC>;
+		reset-names = "saradc-apb";

Keep in mind the driver you added does _not_ trigger this reset.

Quoted 15 lines
arch/arm/dts/rk3568.dtsi +4 -1
+		#io-channel-cells = <1>;
+		status = "disabled";
+	};
+
 	pinctrl: pinctrl {
 		compatible = "rockchip,rk3568-pinctrl";
 		rockchip,grf = <&grf>;
--

2.20.1

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

Signature · 9 lines
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox
#5 lore raw

Re: [PATCH v2 2/2] arm: rk3568-evb1: add support for Rockchip SARADC

Hello Ahmad,

Thanks for your review, I'll prepare a v3 as soon as I have time.

Quoted 33 lines · Ahmad Fatoum
Quoted 30 lines · Michael Riesch

---
v2:
- add initial version

 arch/arm/dts/rk3568-evb1-v10.dts |  5 +++++
 arch/arm/dts/rk3568.dtsi         | 12 ++++++++++++
 2 files changed, 17 insertions(+)
arch/arm/dts/rk3568-evb1-v10.dts +5 -0
index ca6f9c280..b8b4c85bf 100644
@@ -485,3 +485,8 @@
 &uart2 {
 	status = "okay";
 };
+
+&saradc {
+	vref-supply = <&vcca_1v8>;
+	status = "okay";
+};
arch/arm/dts/rk3568.dtsi +0 -0
index 34d0b8a08..8a75a7e96 100644
@@ -817,6 +817,18 @@
 		status = "disabled";
 	};
 
Quoted 10 lines

arch/arm/dts/rk3568.dtsi line 827

Quoted 8 lines
arch/arm/dts/rk3568.dtsi +8 -0
+	saradc: saradc@fe720000 {
+		compatible = "rockchip,rk3568-saradc";
+		reg = <0x0 0xfe720000 0x0 0x100>;
+		interrupts = <GIC_SPI 93 IRQ_TYPE_LEVEL_HIGH>;
+		clocks = <&cru CLK_SARADC>, <&cru PCLK_SARADC>;
+		clock-names = "saradc", "apb_pclk";
+		resets = <&cru SRST_P_SARADC>;
+		reset-names = "saradc-apb";

Keep in mind the driver you added does _not_ trigger this reset.

Duly noted. I considered implementing the reset in the driver (see v1-thread) but did not arrive at the conclusion that the reset is really required here.

Regards, Michael

Quoted 11 lines
Quoted 11 lines
arch/arm/dts/rk3568.dtsi +4 -1
+		#io-channel-cells = <1>;
+		status = "disabled";
+	};
+
 	pinctrl: pinctrl {
 		compatible = "rockchip,rk3568-pinctrl";
 		rockchip,grf = <&grf>;
--

2.20.1

Quoted 6 lines
Quoted 5 lines

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox

_______________________________________________
barebox mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/barebox