Message-ID: <a02ee642-a88d-eb9a-ff22-18bdf341f73a@ti.com>
Date:   Thu, 4 Nov 2021 08:37:17 +0530
From:   "J, KEERTHY" <j-keerthy@...com>
To:     Daniel Lezcano <daniel.lezcano@...aro.org>, <robh+dt@...nel.org>,
        <rui.zhang@...el.com>, <amitk@...nel.org>, <kristo@...nel.org>
CC:     <linux-pm@...r.kernel.org>, <devicetree@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v4 4/4] thermal: k3_j72xx_bandgap: Add the bandgap driver
 support



On 11/3/2021 8:29 PM, Daniel Lezcano wrote:
> On 03/11/2021 10:19, Keerthy wrote:
>> Add VTM thermal support. In the Voltage Thermal Management
>> Module(VTM), K3 J72XX supplies a voltage reference and a temperature
>> sensor feature that are gathered in the band gap voltage and
>> temperature sensor (VBGAPTS) module. The band gap provides current and
>> voltage reference for its internal circuits and other analog IP
>> blocks. The analog-to-digital converter (ADC) produces an output value
>> that is proportional to the silicon temperature.
>>
>> Currently reading temperatures only is supported.  There are no
>> active/passive cooling agent supported.
>>
>> J721e SoCs have errata i2128: https://www.ti.com/lit/pdf/sprz455
>>
>> The VTM Temperature Monitors (TEMPSENSORs) are trimmed during production,
>> with the resulting values stored in software-readable registers. Software
>> should use these  register values when translating the Temperature
>> Monitor output codes to temperature values.
>>
>> It has an involved workaround. Software needs to read the error codes for
>> -40C, 30C, 125C from the efuse for each device & derive a new look up table
>> for adc to temperature conversion. Involved calculating slopes & constants
>> using 3 different straight line equations with adc refernce codes as the
>> y-axis & error codes in the x-axis.
>>
>> -40C to 30C
>> 30C to 125C
>> 125C to 150C
>>
>> With the above 2 line equations we derive the full look-up table to
>> workaround the errata i2128 for j721e SoC.
>>
>> Tested temperature reading on J721e SoC & J7200 SoC.
> 
> I made a bunch of comments, but this driver needs some reworking to be
> simplified.
> 
> The driver is very similar to the k3_bandgap.c and we found exactly the
> same table+3values-avg approach. After simplifying this driver, it
> should merge easily with k3_bandgap.c.

k3_bandgap.c was my first choice as well but did not want to clutter 
that up with a newer errata implementation. Hence chose to have a new 
driver for j72xx. I am okay either ways.

> 
>> [daniel.lezcano@...aro.org: Generate look-up tables run-time]
>> Signed-off-by: Daniel Lezcano <daniel.lezcano@...aro.org>
>> Signed-off-by: Keerthy <j-keerthy@...com>
>> ---
>>
>> Changes in v4:
>>
>>    * Fixed warnings with w=1 during compilation.
>>
>> Changes in v3:
>>
>>    * Removed static look up tables & added functions to dynamically generate them.
>>      Thanks Daniel! Saved 12KB of static data & almost 200 lines of driver code.
>>
>>
>>   drivers/thermal/Makefile           |   2 +-
>>   drivers/thermal/k3_j72xx_bandgap.c | 634 +++++++++++++++++++++++++++++
>>   2 files changed, 635 insertions(+), 1 deletion(-)
>>   create mode 100644 drivers/thermal/k3_j72xx_bandgap.c
>>
>> diff --git a/drivers/thermal/Makefile b/drivers/thermal/Makefile
>> index 82fc3e616e54..6dcaf2028ddc 100644
>> --- a/drivers/thermal/Makefile
>> +++ b/drivers/thermal/Makefile
>> @@ -28,7 +28,7 @@ thermal_sys-$(CONFIG_CPU_IDLE_THERMAL)	+= cpuidle_cooling.o
>>   # devfreq cooling
>>   thermal_sys-$(CONFIG_DEVFREQ_THERMAL) += devfreq_cooling.o
>>   
>> -obj-$(CONFIG_K3_THERMAL)	+= k3_bandgap.o
>> +obj-$(CONFIG_K3_THERMAL)	+= k3_bandgap.o k3_j72xx_bandgap.o
>>   # platform thermal drivers
>>   obj-y				+= broadcom/
>>   obj-$(CONFIG_THERMAL_MMIO)		+= thermal_mmio.o
>> diff --git a/drivers/thermal/k3_j72xx_bandgap.c b/drivers/thermal/k3_j72xx_bandgap.c
>> new file mode 100644
>> index 000000000000..3d3786f5058d
>> --- /dev/null
>> +++ b/drivers/thermal/k3_j72xx_bandgap.c
>> @@ -0,0 +1,634 @@
>> +// SPDX-License-Identifier: GPL-2.0
>> +/*
>> + * TI Bandgap temperature sensor driver for J72XX SoC Family
>> + *
>> + * Copyright (C) 2021 Texas Instruments Incorporated - http://www.ti.com/
>> + */
>> +
>> +#include <linux/module.h>
>> +#include <linux/init.h>
>> +#include <linux/kernel.h>
>> +#include <linux/pm_runtime.h>
>> +#include <linux/err.h>
>> +#include <linux/types.h>
>> +#include <linux/spinlock.h>
>> +#include <linux/of_platform.h>
>> +#include <linux/io.h>
>> +#include <linux/thermal.h>
>> +#include <linux/of.h>
>> +#include <linux/delay.h>
>> +#include <linux/math.h>
>> +#include <linux/slab.h>
>> +
>> +#define K3_VTM_DEVINFO_PWR0_OFFSET		0x4
>> +#define K3_VTM_DEVINFO_PWR0_CVD_CT_MASK		0xf
>> +#define K3_VTM_DEVINFO_PWR0_TEMPSENS_CT_MASK	0xf0
>> +#define K3_VTM_TMPSENS0_CTRL_OFFSET		0x300
>> +#define K3_VTM_MISC_CTRL_OFFSET			0xc
>> +#define K3_VTM_TMPSENS_STAT_OFFSET		0x8
>> +#define K3_VTM_ANYMAXT_OUTRG_ALERT_EN		0x1
>> +#define K3_VTM_MISC_CTRL2_OFFSET		0x10
>> +#define K3_VTM_REGS_PER_TS			0x10
>> +#define K3_VTM_TS_STAT_DTEMP_MASK		0x3ff
>> +#define K3_VTM_MAX_NUM_TS			8
>> +#define K3_VTM_TMPSENS_CTRL_CBIASSEL		BIT(0)
>> +#define K3_VTM_TMPSENS_CTRL_SOC			BIT(5)
>> +#define K3_VTM_TMPSENS_CTRL_CLRZ		BIT(6)
>> +#define K3_VTM_TMPSENS_CTRL_CLKON_REQ		BIT(7)
>> +#define K3_VTM_TMPSENS_CTRL_MAXT_OUTRG_EN	BIT(11)
>> +
>> +#define K3_VTM_CORRECTION_TEMP_CNT		3
>> +
>> +#define K3_VTM_ADC_BEGIN_VAL			6
>> +#define K3_VTM_ADC_END_VAL			944
>> +
>> +#define MINUS40CREF				5
>> +#define PLUS30CREF				253
>> +#define PLUS125CREF				730
>> +#define PLUS150CREF				940
>> +
>> +#define MAX_TEMP				123000
>> +#define COOL_DOWN_TEMP				105000
>> +
>> +//#define DEBUG_VTM
> 
> I suggest to remove DEBUG_VTM and replace the dev_info by dev_dbg in the
> print lookup table. There are also messages to reduce the verbosity
> level to debug (commented inline).

Okay.

> 
>> +
>> +static int *ref_table;
>> +static int *derived_table;
> 
> These tables could be passed around via function parameters and their
> life cycle is just during the probe. No need to declare them globally.

Okay.

> 
>> +
>> +s64 golden_factors[] = {
>> +	-490019999999999936,
>> +	3251200000000000,
>> +	-1705800000000,
>> +	603730000,
>> +	-92627,
>> +};
>> +
>> +s64 pvt_wa_factors[] = {
>> +	-415230000000000000,
>> +	3126600000000000,
>> +	-1157800000000,
>> +};
>> +
>> +static s64 int_power(unsigned long long base, unsigned int exp)
>> +{
>> +	s64 result = 1;
>> +
>> +	while (exp) {
>> +		if (exp & 1)
>> +			result *= base;
>> +		exp >>= 1;
>> +		base *= base;
>> +	}
>> +
>> +	return result;
>> +}
> 
> This function is already implemented.
> 
> include/linux/math.h:u64 int_pow(u64 base, unsigned int exp);

I will pick that up.

> 
>> +static int compute_value(int index, s64 *factors, int nr_factors, int reduction)
>> +{
>> +	s64 value = 0;
>> +	int i;
>> +
>> +	for (i = 0; i < nr_factors; i++)
>> +		value += factors[i] * int_power(index, i);
>> +
>> +	return (int)(value / int_power(10, reduction));
>> +}
>> +
>> +static void init_table(bool wa, int *mytable)
> 
> s/mytable/table/
> 
> Instead of passing a boolean, let the caller pass the right parameters,
> that makes the code nicer.
> 
> static void init_table(int *table, int size, s64 *factors, int nr_factors)
> {
> 	int i;
> 
> 	for (i = 0; i < size; i++)
> 		table[i] = polynomial(factors, nr_factors,
> 					FACTORS_REDUCTION);
> }

Okay.

> 
>> +{
>> +	s64 *factors;
>> +	const int size = 1024;
>> +	int factors_size;
>> +	int i;
>> +
>> +	if (wa) {
>> +		factors_size = 3;
>> +		factors = pvt_wa_factors;
>> +	} else {
>> +		factors_size = 5;
>> +		factors = golden_factors;
>> +	}
>> +
>> +	for (i = 0; i < size; i++)
>> +		mytable[i] = compute_value(i, factors, factors_size, 13);
>> +}
>> +
>> +static void fill_table(int err1, int err2, int ref1, int ref2, int idx1,
>> +		       int idx2)
> 
> I suggest to create a self explanatory structure and pass a couple of
> them as parameter with the derived and ref tables.
> 
> Given what does this function, it would make sense rename it to
> something more explicit along with a short explanation of why this is
> needed.

Okay

> 
>> +{
>> +	int m, c, num, den, i, err;
>> +
>> +	/**
> 
> '/**' is for kernel documentation, should be replaced by /*

Okay

> 
>> +	 * Calculate the slope with adc values read from the register
>> +	 * as the y-axis param and err in adc value as x-axis param
>> +	 */
>> +	if (err1 != err2) {
>> +		num = ref2 - ref1;
>> +		den = err2 - err1;
>> +		m = num / den;
> 
> 	if den > num then 'm' is zero and err = (i - c) / m will crash
> 	if ref1 and ref2 are equal, 'm' will be zero also

Thanks for catching. I will add them as conditions in the if.

> 
>> +		c = ref2 - m * err2;
>> +		for (i = idx1; i <= idx2; i++) {
>> +			err = (i - c) / m;
>> +			if (((i + err1) < 0) || ((i + err1) > 1023))
>> +				continue;
>> +			derived_table[i] = ref_table[i + err];
>> +		}
>> +	} else { /* Constant error take care of divide by zero */
>> +		for (i = idx1; i < idx2; i++) {
> 
> 		Why the first loop has i<=idx2 and this one is i<idx2 ?

I will make them <=.

> 
>> +			if (((i + err1) < 0) || ((i + err1) > 1023))
>> +				continue;
>> +			derived_table[i] = ref_table[i + err1];
>> +		}
> 
> Can be this loop put outside of the if the else ?

Okay. I can add a return in the if & else can be removed.



> 
>> +	}
>> +}
>> +
>> +static int prep_lookup_table(int minus40cerr, int plus30cerr, int plus125cerr,
>> +			     int plus150cerr)
>> +{
>> +	int inc, i;
>> +
>> +	/* Fill up the lookup table region -40C to +30C */
>> +	fill_table(minus40cerr, plus30cerr, MINUS40CREF, PLUS30CREF, 0,
>> +		   PLUS30CREF);
>> +	/* Fill up the lookup table region +30C to +125C */
>> +	fill_table(plus30cerr, plus125cerr, PLUS30CREF, PLUS125CREF, PLUS30CREF,
>> +		   PLUS125CREF);
>> +	/* Fill up the lookup table region +125C to +150C */
>> +	fill_table(plus125cerr, plus150cerr, PLUS125CREF, PLUS150CREF, PLUS125CREF,
>> +		   PLUS150CREF);
>> +
>> +	/* Get to the first valid temperature */
>> +	i = 0;
>> +	while (!derived_table[i])
>> +		i++;
> 
> 	zero is not a valid temperature ?

Basically assumption is we start somewhere near -40C. If there are zeros 
it means those indicies need to b backfilled with 300 milli degree steps.

> 
>> +
>> +	/*
>> +	 * Get to the last zero index and back fill the temperature for
>> +	 * sake of continuity
>> +	 */
>> +	if (i) {
>> +		/* 300 milli celsius steps */
>> +		while (i--)
>> +			derived_table[i] = derived_table[i + 1] - 300;
>> +		/* case 0 */
>> +		derived_table[i] = derived_table[i + 1] - 300;
>> +	}
>> +
>> +	/*
>> +	 * Fill the last trailing 0s which are unfilled with increments of
>> +	 * 100 milli celsius till 1023 code
>> +	 */
>> +	i = 1023;
> 
> Please, give the table and const table size in parameters.

Okay

> 
>> +	while (!derived_table[i])
>> +		i--;
>> +
>> +	i++;
>> +	inc = 1;
>> +	while (i < 1024) {
>> +		derived_table[i] = derived_table[i - 1] + inc * 100;
>> +		i++;
>> +	}
> 
> Can you explain why the above is done ?

This is the ending section of the table if some of the indicies are not 
populated then as per errata implementation fill the last remaining 
idicies with 100 milli degree centigrade increments to have the table 
complete.

> 
>> +	return 0;
>> +}
>> +
>> +struct k3_thermal_data;
> not necessary
> 
> Below, "not necessary" == the field is assigned but then never used,
> "not used" == the field is never used at all.
> 
>> +struct k3_j72xx_bandgap {
>> +	struct device *dev;
> 
> 	not necessary
> 
>> +	void __iomem *base;
>> +	void __iomem *cfg2_base;
>> +	void __iomem *fuse_base;
>> +	const struct k3_j72xx_bandgap_data	*conf;
>> +	spinlock_t lock; /* shields this struct */
> 
> 	not used
> 
>> +	int ts_cnt;
> 
> 	not used
> 
>> +	struct k3_thermal_data *ts_data[K3_VTM_MAX_NUM_TS];
> 
> 	not used
>> +};
>> +
>> +/* common data structures */
>> +struct k3_thermal_data {
>> +	struct thermal_zone_device *ti_thermal;
> 
> 	not necessary
> 
>> +	struct thermal_cooling_device *cool_dev;
> 
> 	not used
> 
>> +	struct k3_j72xx_bandgap *bgp;
>> +	enum thermal_device_mode mode;
> 
> 	not used
> 
>> +	int sensor_id;
> 
> 	not used
> 
>> +	u32 ctrl_offset;
>> +	u32 stat_offset;
>> +	int prev_temp;
> 
> 	not necessary
> 
>> +	int ct_offsets[K3_VTM_CORRECTION_TEMP_CNT];
>> +	int ct_bm[K3_VTM_CORRECTION_TEMP_CNT];
> 
> 	Both are only local to the function using it
> 
>> +	int ref_adc_val[3];
> 
> 	not used
> 
>> +	int act_adc_val[3];
> 
> 	not used
> 
>> +	int slope[2];
> 
> 	not used
> 
>> +	int coeff[2];
> 
> 	not used
> 
>> +};

Okay for all the above params. I will optimize them.

>> +
>> +static int two_cmp(int tmp, int mask)
>> +{
>> +	tmp = ~(tmp);
>> +	tmp &= mask;
>> +	tmp += 1;
>> +
>> +	/* Return negative value */
>> +	return (0 - tmp);
>> +}
> 
> Can you explain ?

Two's compliment implementation. Some of the error values in the Efuse 
are in 2's complement notation & hence the above function is used to 
convert.

> 
>> +
>> +static unsigned int vtm_get_best_value(unsigned int s0, unsigned int s1,
>> +				       unsigned int s2)
>> +{
>> +	int d01 = abs(s0 - s1);
>> +	int d02 = abs(s0 - s2);
>> +	int d12 = abs(s1 - s2);
>> +
>> +	if (d01 <= d02 && d01 <= d12)
>> +		return (s0 + s1) / 2;
>> +
>> +	if (d02 <= d01 && d02 <= d12)
>> +		return (s0 + s2) / 2;
>> +
>> +	return (s1 + s2) / 2;
>> +}
>> +
>> +static inline int k3_bgp_read_temp(struct k3_thermal_data *devdata,
>> +				   int *temp)
>> +{
>> +	struct k3_j72xx_bandgap *bgp;
>> +	unsigned int dtemp, s0, s1, s2;
>> +
>> +	bgp = devdata->bgp;
>> +	/*
>> +	 * Errata is applicable for am654 pg 1.0 silicon/J7ES. There
>> +	 * is a variation of the order for certain degree centigrade on AM654.
>> +	 * Work around that by getting the average of two closest
>> +	 * readings out of three readings everytime we want to
>> +	 * report temperatures.
>> +	 *
>> +	 * Errata workaround.
>> +	 */
>> +	s0 = readl(bgp->base + devdata->stat_offset) &
>> +		K3_VTM_TS_STAT_DTEMP_MASK;
>> +	s1 = readl(bgp->base + devdata->stat_offset) &
>> +		K3_VTM_TS_STAT_DTEMP_MASK;
>> +	s2 = readl(bgp->base + devdata->stat_offset) &
>> +		K3_VTM_TS_STAT_DTEMP_MASK;
>> +	dtemp = vtm_get_best_value(s0, s1, s2);
>> +
>> +	if (dtemp < 0 || dtemp > 1023)
>> +		return -EINVAL;
>> +
>> +	*temp = derived_table[dtemp];
>> +
>> +	return 0;
>> +}
>> +
>> +/* Get temperature callback function for thermal zone */
>> +static int k3_thermal_get_temp(void *devdata, int *temp)
>> +{
>> +	struct k3_thermal_data *data = devdata;
>> +	int ret = 0;
>> +
>> +	ret = k3_bgp_read_temp(data, temp);
>> +	if (ret)
>> +		return ret;
>> +
>> +	data->prev_temp = *temp;
> 
> 	where is it used ?

Okay this will be needed if trend computation was needed. I will 
optimize this one as well.

> 
>> +
>> +	return ret;
>> +}
>> +
>> +static const struct thermal_zone_of_device_ops k3_of_thermal_ops = {
>> +	.get_temp = k3_thermal_get_temp,
>> +};
>> +
>> +static int k3_j72xx_bandgap_temp_to_adc_code(int temp)
>> +{
>> +	int low = 0, high = 1023, mid;
>> +
>> +	if (temp > 160000 || temp < -50000)
>> +		return -EINVAL;
> 
> How the caller does the difference between a negative temperature and
> this error ?

The converted ADC code has to be between 0 - 1023. Any negative value 
returned is an error. Does that make sense?

> 
>> +	/* Binary search to find the adc code */
>> +	while (low < (high - 1)) {
>> +		mid = (low + high) / 2;
>> +		if (temp <= derived_table[mid])
>> +			high = mid;
>> +		else
>> +			low = mid;
>> +	}
>> +
>> +	return mid;
>> +}
>> +
>> +static void get_efuse_values(int id, struct k3_thermal_data *data, int *err,
>> +			     struct k3_j72xx_bandgap *bgp)
>> +{
>> +	int i, tmp, pow;
>> +
>> +	/* Populate efuse reg offsets & Bit masks for -40C, 30C, 125C */
>> +	switch (id) {
>> +	case 0:
>> +		data->ct_offsets[0] = 0x0;
>> +		data->ct_offsets[1] = 0x8;
>> +		data->ct_offsets[2] = 0x4;
>> +		data->ct_bm[0] = 0x3f;
>> +		data->ct_bm[1] = 0x1fe000;
>> +		data->ct_bm[2] = 0x1ff;
>> +		break;
>> +
>> +	case 1:
>> +		data->ct_offsets[0] = 0x0;
>> +		data->ct_offsets[1] = 0x8;
>> +		data->ct_offsets[2] = 0x4;
>> +		data->ct_bm[0] = 0xfc0;
>> +		data->ct_bm[1] = 0x1fe00000;
>> +		data->ct_bm[2] = 0x3fe00;
>> +		break;
>> +
>> +	case 2:
>> +		data->ct_offsets[0] = 0x0;
>> +		data->ct_offsets[1] = -1;
>> +		data->ct_offsets[2] = 0x4;
>> +		data->ct_bm[0] = 0x3f000;
>> +		data->ct_bm[1] = 0x7f800000;
>> +		data->ct_bm[2] = 0x7fc0000;
>> +		break;
>> +
>> +	case 3:
>> +		data->ct_offsets[0] = 0x0;
>> +		data->ct_offsets[1] = 0xC;
>> +		data->ct_offsets[2] = -1; /* Spread across 2 registers */
>> +		data->ct_bm[0] = 0xfc0000;
>> +		data->ct_bm[1] = 0x1fe0;
>> +		data->ct_bm[2] = 0x1f800000;
>> +		break;
>> +
>> +	case 4:
>> +		data->ct_offsets[0] = 0x0;
>> +		data->ct_offsets[1] = 0xc;
>> +		data->ct_offsets[2] = 0x8;
>> +		data->ct_bm[0] = 0x3f000000;
>> +		data->ct_bm[1] = 0x1fe000;
>> +		data->ct_bm[2] = 0x1ff0;
>> +		break;
>> +	}
>> +
>> +	for (i = 0; i < 3; i++) {
>> +		/* Extract the offset value using bit-mask */
>> +		if (data->ct_offsets[i] == -1 && i == 1) {
>> +			/* 25C offset Case of Sensor 2 split between 2 regs */
>> +			tmp = (readl(bgp->fuse_base + 0x8) & 0xE0000000) >> (29);
>> +			tmp |= ((readl(bgp->fuse_base + 0xC) & 0x1F) << 3);
>> +			pow = tmp & 0x80;
>> +		} else if (data->ct_offsets[i] == -1 && i == 2) {
>> +			/* 125C Case of Sensor 3 split between 2 regs */
>> +			tmp = (readl(bgp->fuse_base + 0x4) & 0xF8000000) >> (27);
>> +			tmp |= ((readl(bgp->fuse_base + 0x8) & 0xF) << 5);
>> +			pow = tmp & 0x100;
>> +		} else {
>> +			tmp = readl(bgp->fuse_base + data->ct_offsets[i]);
>> +			tmp &= data->ct_bm[i];
>> +			tmp = tmp >> __ffs(data->ct_bm[i]);
>> +
>> +			/* Obtain the sign bit pow*/
>> +			pow = data->ct_bm[i] >> __ffs(data->ct_bm[i]);
>> +			pow += 1;
>> +			pow /= 2;
>> +		}
>> +
>> +		/* Check for negative value */
>> +		if (tmp & pow) {
>> +			/* 2's complement value */
>> +			tmp = two_cmp(tmp, data->ct_bm[i] >> __ffs(data->ct_bm[i]));
>> +		}
>> +		err[i] = tmp;
> 
> Can you explain what is doing the code above ?

The efuse will have error values in 2's complement notation. The above 
it to get the MSB which will decide the sign.

> 
>> +	}
>> +}
>> +
>> +#ifdef DEBUG_VTM
> 
> Remove

Okay.

> 
>> +static void print_look_up_table(struct device *dev)
>> +{
>> +	int i;
>> +
>> +	dev_info(dev, "The contents of derived array\n");
> 	
> 	dev_dbg
> 
>> +	dev_info(dev, "Code   Temperaturei\n");
> 
> 	dev_dbg
> 
>> +	for (i = 0; i <= 1023; i++)
>> +		dev_info(dev, "%d       %d\n", i, derived_table[i]);
> 
> 	dev_dbg

Okay for all the above devv_info --> dev_dbg

> 
>> +}
>> +#endif
>> +
>> +static const struct of_device_id of_k3_j72xx_bandgap_match[];
>> +
>> +struct k3_j72xx_bandgap_j72xx_data {
> 
> Fix name please

Ah okay. I will fix.

> 
>> +	unsigned int	workaround;
> 
> 	s/workaround/has_errata_<num>/

okay

> 
>> +};
>> +
>> +static int k3_j72xx_bandgap_probe(struct platform_device *pdev)
>> +{
>> +	int ret = 0, cnt, val, id, reg_cnt = 0, j, table_size = 1024;
>> +	int err[3], high_max, low_max;
>> +	struct resource *res;
>> +	struct device *dev = &pdev->dev;
>> +	struct k3_j72xx_bandgap *bgp;
>> +	struct k3_thermal_data *data;
>> +	int workaround_needed = 0;
>> +	const struct of_device_id *of_id;
>> +	const struct k3_j72xx_bandgap_j72xx_data *driver_data;
>> +
>> +	bgp = devm_kzalloc(&pdev->dev, sizeof(*bgp), GFP_KERNEL);
>> +	if (!bgp)
>> +		return -ENOMEM;
>> +
>> +	bgp->dev = dev;
>> +	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>> +	bgp->base = devm_ioremap_resource(dev, res);
>> +	if (IS_ERR(bgp->base))
>> +		return PTR_ERR(bgp->base);
>> +
>> +	res = platform_get_resource(pdev, IORESOURCE_MEM, 1);
>> +	bgp->cfg2_base = devm_ioremap_resource(dev, res);
>> +	if (IS_ERR(bgp->cfg2_base))
>> +		return PTR_ERR(bgp->cfg2_base);
>> +
>> +	res = platform_get_resource(pdev, IORESOURCE_MEM, 2);
>> +	bgp->fuse_base = devm_ioremap_resource(dev, res);
>> +	if (IS_ERR(bgp->fuse_base))
>> +		return PTR_ERR(bgp->fuse_base);
>> +
>> +	of_id = of_match_device(of_k3_j72xx_bandgap_match, &pdev->dev);
>> +	if (of_id) {
>> +		driver_data = of_id->data;
>> +		workaround_needed = driver_data->workaround;
>> +	}
> 
> of_device_get_match_data(dev) ?

okay

> 
>> +	pm_runtime_enable(dev);
>> +	ret = pm_runtime_get_sync(dev);
>> +	if (ret < 0) {
>> +		pm_runtime_put_noidle(dev);
>> +		pm_runtime_disable(dev);
>> +		return ret;
>> +	}
>> +
>> +	/* Get the sensor count in the VTM */
>> +	val = readl(bgp->base + K3_VTM_DEVINFO_PWR0_OFFSET);
>> +	cnt = val & K3_VTM_DEVINFO_PWR0_TEMPSENS_CT_MASK;
>> +	cnt >>= __ffs(K3_VTM_DEVINFO_PWR0_TEMPSENS_CT_MASK);
>> +	bgp->ts_cnt = cnt;
>> +
>> +	data = devm_kcalloc(bgp->dev, cnt, sizeof(*data), GFP_KERNEL);
>> +	if (!data) {
>> +		ret = -ENOMEM;
>> +		goto err_alloc;
>> +	}
>> +
>> +	ref_table = devm_kzalloc(bgp->dev, sizeof(*ref_table) * table_size,
>> +				 GFP_KERNEL);
>> +	if (!ref_table) {
>> +		ret = -ENOMEM;
>> +		goto err_alloc;
>> +	}
>> +
>> +	derived_table = devm_kzalloc(bgp->dev, sizeof(*derived_table) * table_size,
>> +				     GFP_KERNEL);
>> +	if (!derived_table) {
>> +		ret = -ENOMEM;
>> +		goto err_alloc;
>> +	}
>> +
>> +	/* Workaround not needed if bit30/bit31 is set even for J721e */
>> +	if (workaround_needed && (readl(bgp->fuse_base + 0x0) & 0xc0000000) == 0xc0000000)
>> +		workaround_needed = false;
> 
> Does it mean the same hardware may have or not the errata? Is there a
> way to distinguish them from the compatible string ?

Yes Same hardware same silicon revision some may have & some might not. 
This is the prescribed way to check if existance of errata.

> 
>> +
>> +	if (!workaround_needed) {
>> +		dev_info(bgp->dev, "work around NOT needed!!\n");
>> +		init_table(false, ref_table);
>> +	} else {
>> +		dev_info(bgp->dev, "work around needed!!\n");
>> +		init_table(true, ref_table);
>> +	}
> 
> 	dev_dbg(bgp->dev, "Work around %sneeded\n",
> 		workaround_needed ? "not " : "");

okay

> 
>> +
>> +	/* Register the thermal sensors */
>> +	for (id = 0; id < cnt; id++) {
>> +		data[id].sensor_id = id;
>> +		data[id].bgp = bgp;
>> +		data[id].ctrl_offset = K3_VTM_TMPSENS0_CTRL_OFFSET + id * 0x20;
>> +		data[id].stat_offset = data[id].ctrl_offset +
>> +					K3_VTM_TMPSENS_STAT_OFFSET;
>> +
>> +		if (!workaround_needed)
>> +			goto prepare_lookup;
> 
> 		if (workaround_needed) {
> 
>> +		/* ref adc values for -40C, 30C & 125C respectively */
>> +		data[id].ref_adc_val[0] = MINUS40CREF;
>> +		data[id].ref_adc_val[1] = PLUS30CREF;
>> +		data[id].ref_adc_val[2] = PLUS125CREF;
>> +		get_efuse_values(id, &data[id], err, bgp);
> 
> 		}

okay

>> +
>> +prepare_lookup:
>> +		if (id == 0 && workaround_needed) {
>> +			prep_lookup_table(err[0], err[1], err[2], 0);
>> +		} else if (id == 0 && !workaround_needed) {
>> +			for (j = 0; j < table_size; j++)
>> +				derived_table[j] = ref_table[j];
> 
> 		memcpy ?

okay

> 
>> +		}
>> +
>> +		val = readl(data[id].bgp->cfg2_base + data[id].ctrl_offset);
>> +		val |= (K3_VTM_TMPSENS_CTRL_MAXT_OUTRG_EN |
>> +			K3_VTM_TMPSENS_CTRL_SOC |
>> +			K3_VTM_TMPSENS_CTRL_CLRZ | BIT(4));
>> +		writel(val, data[id].bgp->cfg2_base + data[id].ctrl_offset);
>> +
>> +		bgp->ts_data[id] = &data[id];
>> +		data[id].ti_thermal =
>> +		devm_thermal_zone_of_sensor_register(bgp->dev, id,
>> +						     &data[id],
>> +						     &k3_of_thermal_ops);
>> +		if (IS_ERR(data[id].ti_thermal)) {
>> +			dev_err(bgp->dev, "thermal zone device is NULL\n");
>> +			ret = PTR_ERR(data[id].ti_thermal);
>> +			goto err_alloc;
>> +		}
>> +
>> +		reg_cnt++;
> 
> 		not used

okay

>> +
>> +		/* Initialize Previous temp */
>> +		k3_thermal_get_temp(&data[id], &data[id].prev_temp);
> 
> 		why ? where is used prev_temp ?

This can be otimized.

>> +	}
>> +
>> +	/*
>> +	 * Program TSHUT thresholds
>> +	 * Step 1: set the thresholds to ~123C and 105C WKUP_VTM_MISC_CTRL2
>> +	 * Step 2: WKUP_VTM_TMPSENS_CTRL_j set the MAXT_OUTRG_EN  bit
>> +	 *         This is already taken care as per of init
>> +	 * Step 3: WKUP_VTM_MISC_CTRL set the ANYMAXT_OUTRG_ALERT_EN  bit
>> +	 */
>> +	high_max = k3_j72xx_bandgap_temp_to_adc_code(MAX_TEMP);
>> +	low_max = k3_j72xx_bandgap_temp_to_adc_code(COOL_DOWN_TEMP);
> 
> I guess high_max is the hardware shutdown temperature, right ?

correct.

> 
> But what is low_max ?

It is high_max - hysteresis. May be i should change to high_max_hyst?

> 
>> +	writel((low_max << 16) | high_max, data[0].bgp->cfg2_base +
>> +	       K3_VTM_MISC_CTRL2_OFFSET);
>> +	mdelay(100);
>> +	writel(K3_VTM_ANYMAXT_OUTRG_ALERT_EN, data[0].bgp->cfg2_base +
>> +	       K3_VTM_MISC_CTRL_OFFSET);
>> +
>> +	platform_set_drvdata(pdev, bgp);
>> +
>> +#ifdef DEBUG_VTM
> 
> Remove ifdef

okay

Thanks for the detailed review. I will work on the comments.

Regards,
Keerthy
> 
>> +	print_look_up_table(dev);
>> +#endif
>> +	/*
>> +	 * Now that the derived_table has the appropriate look up values
>> +	 * Free up the ref_table
>> +	 */
>> +	kfree(ref_table);
>> +
>> +	return 0;
>> +
>> +err_alloc:
>> +	pm_runtime_put_sync(&pdev->dev);
>> +	pm_runtime_disable(&pdev->dev);
>> +
>> +	return ret;
>> +}
>> +
>> +static int k3_j72xx_bandgap_remove(struct platform_device *pdev)
>> +{
>> +	pm_runtime_put_sync(&pdev->dev);
>> +	pm_runtime_disable(&pdev->dev);
>> +
>> +	return 0;
>> +}
>> +
>> +const struct k3_j72xx_bandgap_j72xx_data k3_j72xx_bandgap_j721e_data = {
>> +	.workaround = 1,
>> +};
>> +
>> +const struct k3_j72xx_bandgap_j72xx_data k3_j72xx_bandgap_j7200_data = {
>> +	.workaround = 0,
>> +};
>> +
>> +static const struct of_device_id of_k3_j72xx_bandgap_match[] = {
>> +	{
>> +		.compatible = "ti,j721e-vtm",
>> +		.data = (void *)&k3_j72xx_bandgap_j721e_data,
>> +	},
>> +	{
>> +		.compatible = "ti,j7200-vtm",
>> +		.data = (void *)&k3_j72xx_bandgap_j7200_data,
>> +	},
>> +	{ /* sentinel */ },
>> +};
>> +MODULE_DEVICE_TABLE(of, of_k3_j72xx_bandgap_match);
>> +
>> +static struct platform_driver k3_j72xx_bandgap_sensor_driver = {
>> +	.probe = k3_j72xx_bandgap_probe,
>> +	.remove = k3_j72xx_bandgap_remove,
>> +	.driver = {
>> +		.name = "k3-j72xx-soc-thermal",
>> +		.of_match_table	= of_k3_j72xx_bandgap_match,
>> +	},
>> +};
>> +
>> +module_platform_driver(k3_j72xx_bandgap_sensor_driver);
>> +
>> +MODULE_DESCRIPTION("K3 bandgap temperature sensor driver");
>> +MODULE_LICENSE("GPL v2");
>> +MODULE_AUTHOR("J Keerthy <j-keerthy@...com>");
>>
> 
> 

Powered by blists - more mailing lists