Re: [PATCH 2/6] thermal/drivers/tsens: Add TSENS enable and calibration support for V2

From: Praveenkumar I
Date: Tue Jul 11 2023 - 05:24:49 EST


On 7/10/2023 8:32 PM, Dmitry Baryshkov wrote:
On Mon, 10 Jul 2023 at 16:22, Praveenkumar I <quic_ipkumar@xxxxxxxxxxx> wrote:

On 7/10/2023 4:49 PM, Dmitry Baryshkov wrote:
On 10/07/2023 13:37, Praveenkumar I wrote:
SoCs without RPM have to enable sensors and calibrate from the kernel.
Though TSENS IP supports 16 sensors, not all are used. So added
sensors_to_en in tsens data help enable the relevant sensors.

Added new calibration function for V2 as the tsens.c calib function
only supports V1.

Signed-off-by: Praveenkumar I <quic_ipkumar@xxxxxxxxxxx>
---
drivers/thermal/qcom/tsens-v2.c | 116 ++++++++++++++++++++++++++++++++
drivers/thermal/qcom/tsens.c | 37 +++++++++-
drivers/thermal/qcom/tsens.h | 56 +++++++++++++++
3 files changed, 208 insertions(+), 1 deletion(-)

diff --git a/drivers/thermal/qcom/tsens-v2.c
b/drivers/thermal/qcom/tsens-v2.c
index 29a61d2d6ca3..db48b1d95348 100644
--- a/drivers/thermal/qcom/tsens-v2.c
+++ b/drivers/thermal/qcom/tsens-v2.c
@@ -6,11 +6,20 @@
#include <linux/bitops.h>
#include <linux/regmap.h>
+#include <linux/nvmem-consumer.h>
#include "tsens.h"
/* ----- SROT ------ */
#define SROT_HW_VER_OFF 0x0000
#define SROT_CTRL_OFF 0x0004
+#define SROT_MEASURE_PERIOD 0x0008
+#define SROT_Sn_CONVERSION 0x0060
+#define V2_SHIFT_DEFAULT 0x0003
+#define V2_SLOPE_DEFAULT 0x0cd0
+#define V2_CZERO_DEFAULT 0x016a
+#define ONE_PT_SLOPE 0x0cd0
+#define TWO_PT_SHIFTED_GAIN 921600
+#define ONE_PT_CZERO_CONST 94
/* ----- TM ------ */
#define TM_INT_EN_OFF 0x0004
@@ -59,6 +68,16 @@ static const struct reg_field
tsens_v2_regfields[MAX_REGFIELDS] = {
/* CTRL_OFF */
[TSENS_EN] = REG_FIELD(SROT_CTRL_OFF, 0, 0),
[TSENS_SW_RST] = REG_FIELD(SROT_CTRL_OFF, 1, 1),
+ [SENSOR_EN] = REG_FIELD(SROT_CTRL_OFF, 3, 18),
+ [CODE_OR_TEMP] = REG_FIELD(SROT_CTRL_OFF, 21, 21),
+
+ /* MAIN_MEASURE_PERIOD */
+ [MAIN_MEASURE_PERIOD] = REG_FIELD(SROT_MEASURE_PERIOD, 0, 7),
+
+ /* Sn Conversion */
+ REG_FIELD_FOR_EACH_SENSOR16(SHIFT, SROT_Sn_CONVERSION, 23, 24),
+ REG_FIELD_FOR_EACH_SENSOR16(SLOPE, SROT_Sn_CONVERSION, 10, 22),
+ REG_FIELD_FOR_EACH_SENSOR16(CZERO, SROT_Sn_CONVERSION, 0, 9),
/* ----- TM ------ */
/* INTERRUPT ENABLE */
@@ -104,6 +123,103 @@ static const struct reg_field
tsens_v2_regfields[MAX_REGFIELDS] = {
[TRDY] = REG_FIELD(TM_TRDY_OFF, 0, 0),
};
+static int tsens_v2_calibration(struct tsens_priv *priv)
+{
+ struct device *dev = priv->dev;
+ u32 mode, base0, base1;
+ u32 slope, czero;
+ char name[15];
+ int i, j, ret;
+
+ if (priv->num_sensors > MAX_SENSORS)
+ return -EINVAL;
+
+ ret = nvmem_cell_read_variable_le_u32(priv->dev, "mode", &mode);
+ if (ret == -ENOENT)
+ dev_warn(priv->dev, "Calibration data not present in DT\n");
+ if (ret < 0)
+ return ret;
+
+ dev_dbg(priv->dev, "calibration mode is %d\n", mode);
+
+ ret = nvmem_cell_read_variable_le_u32(priv->dev, "base0", &base0);
+ if (ret < 0)
+ return ret;
+
+ ret = nvmem_cell_read_variable_le_u32(priv->dev, "base1", &base1);
+ if (ret < 0)
+ return ret;
+
+ /* Read offset values and allocate SHIFT, SLOPE & CZERO regmap
for enabled sensors */
+ for (i = 0; i < priv->num_sensors; i++) {
+ if (!(priv->sensors_to_en & (0x1 << i)))
+ continue;
+
+ ret = snprintf(name, sizeof(name), "s%d_offset",
priv->sensor[i].hw_id);
+ if (ret < 0)
+ return ret;
+
+ ret = nvmem_cell_read_variable_le_u32(priv->dev, name,
&priv->sensor[i].offset);
+ if (ret)
+ return ret;
+
+ for (j = SHIFT_0; j <= CZERO_0; j++) {
+ int idx = (i * 3) + j;
+
+ priv->rf[idx] = devm_regmap_field_alloc(dev,
priv->srot_map,
+ priv->fields[idx]);
+ if (IS_ERR(priv->rf[idx]))
+ return PTR_ERR(priv->rf[idx]);
I think, allocating data structures for 48 regfields, which are
written just once, to be an overkill.
Can we change it to single field for each sensor. For example,
CONVERSION_0 instead of SHIFT_0, SLOPE_0 and CZERO_0? This way it will
be max 16 regfields.
If you move writing of the registers to the loop, you won't need
regfields. You can just call regmap_update_bits. The point is that you
don't have to allocate a one-time instance.
Sure, will update in next patch.

+ }
+ }
+
+ /* Based on calib mode, program SHIFT, SLOPE and CZERO for
enabled sensors */
+ switch (mode) {
+ case TWO_PT_CALIB:
+ slope = (TWO_PT_SHIFTED_GAIN / (base1 - base0));
+
+ for (i = 0; i < priv->num_sensors; i++) {
+ if (!(priv->sensors_to_en & (0x1 << i)))
+ continue;
+
+ int idx = i * 3;
+
+ czero = (base0 + priv->sensor[i].offset - ((base1 -
base0) / 3));
+ regmap_field_write(priv->rf[SHIFT_0 + idx],
V2_SHIFT_DEFAULT);
+ regmap_field_write(priv->rf[SLOPE_0 + idx], slope);
+ regmap_field_write(priv->rf[CZERO_0 + idx], czero);
+ }
+ fallthrough;
+ case ONE_PT_CALIB2:
+ for (i = 0; i < priv->num_sensors; i++) {
+ if (!(priv->sensors_to_en & (0x1 << i)))
+ continue;
+
+ int idx = i * 3;
+
+ czero = base0 + priv->sensor[i].offset -
ONE_PT_CZERO_CONST;
+ regmap_field_write(priv->rf[SHIFT_0 + idx],
V2_SHIFT_DEFAULT);
+ regmap_field_write(priv->rf[SLOPE_0 + idx], ONE_PT_SLOPE);
+ regmap_field_write(priv->rf[CZERO_0 + idx], czero);
+ }
+ break;
+ default:
+ dev_dbg(priv->dev, "calibrationless mode\n");
+ for (i = 0; i < priv->num_sensors; i++) {
+ if (!(priv->sensors_to_en & (0x1 << i)))
+ continue;
+
+ int idx = i * 3;
+
+ regmap_field_write(priv->rf[SHIFT_0 + idx],
V2_SHIFT_DEFAULT);
+ regmap_field_write(priv->rf[SLOPE_0 + idx],
V2_SLOPE_DEFAULT);
+ regmap_field_write(priv->rf[CZERO_0 + idx],
V2_CZERO_DEFAULT);
+ }
+ }
This code iterates over the sensors field several times. Please
consider extracting a function that handles all setup for a single
sensor, then calling it in a loop (I should probably do the same for
tsens-v0/v1 too).
Sure. After reading the mode0, base0 and base1 from QFPROM, we can call
a function in a loop to setup the calibration for each sensor.
+
+ return 0;
+}
+
static const struct tsens_ops ops_generic_v2 = {
.init = init_common,
.get_temp = get_temp_tsens_valid,
diff --git a/drivers/thermal/qcom/tsens.c b/drivers/thermal/qcom/tsens.c
index 98c356acfe98..169690355dad 100644
--- a/drivers/thermal/qcom/tsens.c
+++ b/drivers/thermal/qcom/tsens.c
@@ -974,7 +974,7 @@ int __init init_common(struct tsens_priv *priv)
ret = regmap_field_read(priv->rf[TSENS_EN], &enabled);
if (ret)
goto err_put_device;
- if (!enabled) {
+ if (!enabled && !priv->sensors_to_en) {
dev_err(dev, "%s: device not enabled\n", __func__);
ret = -ENODEV;
goto err_put_device;
@@ -1006,6 +1006,40 @@ int __init init_common(struct tsens_priv *priv)
goto err_put_device;
}
+ /* Do TSENS initialization if required */
+ if (priv->sensors_to_en) {
Maybe it would be better to explicitly add VER_2_X_NO_RPM and check it
here?
Sure, will add a separate version macro.
+ priv->rf[CODE_OR_TEMP] = devm_regmap_field_alloc(dev,
priv->srot_map,
+ priv->fields[CODE_OR_TEMP]);
+ if (IS_ERR(priv->rf[CODE_OR_TEMP])) {
+ ret = PTR_ERR(priv->rf[CODE_OR_TEMP]);
+ goto err_put_device;
+ }
+
+ priv->rf[MAIN_MEASURE_PERIOD] =
+ devm_regmap_field_alloc(dev, priv->srot_map,
+ priv->fields[MAIN_MEASURE_PERIOD]);
+ if (IS_ERR(priv->rf[MAIN_MEASURE_PERIOD])) {
+ ret = PTR_ERR(priv->rf[MAIN_MEASURE_PERIOD]);
+ goto err_put_device;
+ }
+
+ regmap_field_write(priv->rf[TSENS_SW_RST], 0x1);
+
+ /* Update measure period to 2ms */
+ regmap_field_write(priv->rf[MAIN_MEASURE_PERIOD], 0x1);
+
+ /* Enable available sensors */
+ regmap_field_write(priv->rf[SENSOR_EN], priv->sensors_to_en);
+
+ /* Real temperature format */
+ regmap_field_write(priv->rf[CODE_OR_TEMP], 0x1);
+
+ regmap_field_write(priv->rf[TSENS_SW_RST], 0x0);
+
+ /* Enable TSENS */
+ regmap_field_write(priv->rf[TSENS_EN], 0x1);
+ }
+
/* This loop might need changes if enum regfield_ids is
reordered */
for (j = LAST_TEMP_0; j <= UP_THRESH_15; j += 16) {
for (i = 0; i < priv->feat->max_sensors; i++) {
@@ -1282,6 +1316,7 @@ static int tsens_probe(struct platform_device
*pdev)
priv->dev = dev;
priv->num_sensors = num_sensors;
+ priv->sensors_to_en = data->sensors_to_en;
priv->ops = data->ops;
for (i = 0; i < priv->num_sensors; i++) {
if (data->hw_ids)
diff --git a/drivers/thermal/qcom/tsens.h b/drivers/thermal/qcom/tsens.h
index 2805de1c6827..f8897bc8944e 100644
--- a/drivers/thermal/qcom/tsens.h
+++ b/drivers/thermal/qcom/tsens.h
@@ -168,6 +168,58 @@ enum regfield_ids {
TSENS_SW_RST,
SENSOR_EN,
CODE_OR_TEMP,
+ /* MEASURE_PERIOD */
+ MAIN_MEASURE_PERIOD,
+
+ /* Sn_CONVERSION */
+ SHIFT_0,
+ SLOPE_0,
+ CZERO_0,
+ SHIFT_1,
+ SLOPE_1,
+ CZERO_1,
+ SHIFT_2,
+ SLOPE_2,
+ CZERO_2,
+ SHIFT_3,
+ SLOPE_3,
+ CZERO_3,
+ SHIFT_4,
+ SLOPE_4,
+ CZERO_4,
+ SHIFT_5,
+ SLOPE_5,
+ CZERO_5,
+ SHIFT_6,
+ SLOPE_6,
+ CZERO_6,
+ SHIFT_7,
+ SLOPE_7,
+ CZERO_7,
+ SHIFT_8,
+ SLOPE_8,
+ CZERO_8,
+ SHIFT_9,
+ SLOPE_9,
+ CZERO_9,
+ SHIFT_10,
+ SLOPE_10,
+ CZERO_10,
+ SHIFT_11,
+ SLOPE_11,
+ CZERO_11,
+ SHIFT_12,
+ SLOPE_12,
+ CZERO_12,
+ SHIFT_13,
+ SLOPE_13,
+ CZERO_13,
+ SHIFT_14,
+ SLOPE_14,
+ CZERO_14,
+ SHIFT_15,
+ SLOPE_15,
+ CZERO_15,
/* ----- TM ------ */
/* TRDY */
@@ -524,6 +576,7 @@ struct tsens_features {
/**
* struct tsens_plat_data - tsens compile-time platform data
* @num_sensors: Number of sensors supported by platform
+ * @sensors_to_en: Sensors to be enabled. Each bit represent a sensor
* @ops: operations the tsens instance supports
* @hw_ids: Subset of sensors ids supported by platform, if not the
first n
* @feat: features of the IP
@@ -531,6 +584,7 @@ struct tsens_features {
*/
struct tsens_plat_data {
const u32 num_sensors;
+ const u16 sensors_to_en;
There is already a similar field, hw_ids. Can it be used instead?
Yes, it can be used. I missed to check this hw_ids. Will change the
num_sensors to 5 and use the hw_ids.
const struct tsens_ops *ops;
unsigned int *hw_ids;
struct tsens_features *feat;
@@ -551,6 +605,7 @@ struct tsens_context {
* struct tsens_priv - private data for each instance of the tsens IP
* @dev: pointer to struct device
* @num_sensors: number of sensors enabled on this device
+ * @sensors_to_en: sensors to be enabled. Each bit represents a sensor
* @tm_map: pointer to TM register address space
* @srot_map: pointer to SROT register address space
* @tm_offset: deal with old device trees that don't address TM and
SROT
@@ -569,6 +624,7 @@ struct tsens_context {
struct tsens_priv {
struct device *dev;
u32 num_sensors;
+ u16 sensors_to_en;
struct regmap *tm_map;
struct regmap *srot_map;
u32 tm_offset;
--
Thanks,
Praveenkumar