Message-ID: <ac5b3982-a658-e05b-1b5c-3aeeda1585ed@loongson.cn>
Date:   Wed, 14 Jun 2023 16:03:43 +0800
From:   zhuyinbo <zhuyinbo@...ngson.cn>
To:     Daniel Lezcano <daniel.lezcano@...aro.org>,
        "Rafael J . Wysocki" <rafael@...nel.org>,
        Amit Kucheria <amitk@...nel.org>,
        Zhang Rui <rui.zhang@...el.com>,
        Rob Herring <robh+dt@...nel.org>,
        Krzysztof Kozlowski <krzysztof.kozlowski+dt@...aro.org>,
        linux-pm@...r.kernel.org, devicetree@...r.kernel.org,
        linux-kernel@...r.kernel.org
Cc:     Jianmin Lv <lvjianmin@...ngson.cn>, wanghongliang@...ngson.cn,
        Liu Peibao <liupeibao@...ngson.cn>,
        loongson-kernel@...ts.loongnix.cn,
        zhanghongchen <zhanghongchen@...ngson.cn>, zhuyinbo@...ngson.cn
Subject: Re: [PATCH v14 1/2] thermal: loongson-2: add thermal management
 support


Hi Daniel,

Thank you very much for your feedback and suggestions. Below, I have
some comments, please review.

在 2023/6/12 下午10:22, Daniel Lezcano 写道:
> On 26/04/2023 08:20, Yinbo Zhu wrote:
>> This patch adds the support for Loongson-2 thermal sensor controller,
>> which can support maximum 4 sensors, each sensor contains a sampling
>> register and a control register. The sampling register is used to obtain
>> the temperature in real time, the control register GATE field is used to
>> set the threshold of high or low temperature, when the input temperature
>> is higher than the high temperature threshold or lower than the low
>> temperature threshold, an interrupt will occur.
>>
>> Signed-off-by: zhanghongchen <zhanghongchen@...ngson.cn>
>> Signed-off-by: Yinbo Zhu <zhuyinbo@...ngson.cn>
>> ---
>> Change in v14:
>>         1. Add back depends on COMPILE_TEST.
>>         2. The implementation of devm_thermal_add_hwmon_sysfs has 
>> changed in
>>            recent community code that cause compile fail issue and 
>> this verison
>>            add a dev args in devm_thermal_add_hwmon_sysfs to fix 
>> compile issue.
>> Change in v13:
>>         1. Add a description about that how works the sensor.
>>         2. Drop the COMPILE_TEST.
>>         3. Rework the help prograph in LOONGSON2_THERMAL Kconfig.
>>         4. Drop the 'tzd' 'irq' and 'pdev' element in 
>> loongson2_thermal_data.
>>         5. Drop the reset of variable in loongson2_thermal_set.
>>         6. Drop the function loongson2_thermal_get_sensor_id.
>>         7. Drop the function loongson2_thermal_alarm_irq.
>>         8. Rework the devm_thermal_of_zone_register.
>>         9. Pass 'tzd' instead of 'data' in devm_request_threaded_irq.
>>         10. Drop the "data->tzd->tzp->no_hwmon = false".
>>         11. Drop the loongson2_thermal_remove.
>>         12. Add the sensor id in the of_device_id data field.
>>         13. Drop the save and restore function.
>> Change in v12:
>>         1. Fixup it about min and max.
>>         2. Use dev_err_probe replace dev_err in 
>> devm_request_threaded_irq context.
>> Change in v11:
>>         1. Add min() and max() to replace related code in function
>>            loongson2_thermal_set.
>>         2. Add dev_err_probe to to replace related code for function
>>            return value use devm_thermal_of_zone_register.
>>         3. Replace thermal_add_hwmon_sysfs with 
>> devm_thermal_add_hwmon_sysfs
>>            and use dev_warn replace dev_err in this context.
>> Change in v10:
>>         1. Add all history change log information.
>> Change in v9:
>>         1. Switch new API that use devm_thermal_of_zone_register
>>            to replace previous interfaces.
>>         2. Add depend on LOONGARCH || COMPILE_TEST.
>> Change in v8:
>>                  1. Replace string loongson2/Loongson2/LOONGSON2 with 
>> loongson-2/
>>                     Loongson-2/LOONGSON-2 in Kconfig and commit log 
>> and MAINTAINERS
>>            files.
>> Change in v7:
>>         1. Split the modification of patch 3 and merge it into this 
>> patch.
>>         2. Remove the unless code annotation to fix the compile warning
>>            when compile C code with W=1.
>> Change in v6:
>>         1. NO change, but other patch in this series of patches set has
>>            changes.
>> Change in v5:
>>         1. NO change, but other patch in this series of patches set has
>>            changes.
>> Change in v4:
>>         1. Fixup the compatible.
>> Change in v3:
>>         1. Add a function to gain sensor id an remove dts id.
>> Change in v2:
>>         1. Remove error msg printing when addr ioremap has error.
>>         2. Make loongson2 thermal driver was built-in by default.
>>         3. Replace ls2k with loongson2.
>>         4. Remove CONFIG_PM_SLEEP and set pm function type was
>>            __maybe_unused.
>>
>>   MAINTAINERS                         |   7 ++
>>   drivers/thermal/Kconfig             |  12 ++
>>   drivers/thermal/Makefile            |   1 +
>>   drivers/thermal/loongson2_thermal.c | 166 ++++++++++++++++++++++++++++
>>   4 files changed, 186 insertions(+)
>>   create mode 100644 drivers/thermal/loongson2_thermal.c
>>
>> diff --git a/MAINTAINERS b/MAINTAINERS
>> index 25a0981c74b6..b3a76acd2caf 100644
>> --- a/MAINTAINERS
>> +++ b/MAINTAINERS
>> @@ -12143,6 +12143,13 @@ S:    Maintained
>>   F:    
>> Documentation/devicetree/bindings/pinctrl/loongson,ls2k-pinctrl.yaml
>>   F:    drivers/pinctrl/pinctrl-loongson2.c
>> +LOONGSON-2 SOC SERIES THERMAL DRIVER
>> +M:    zhanghongchen <zhanghongchen@...ngson.cn>
>> +M:    Yinbo Zhu <zhuyinbo@...ngson.cn>
>> +L:    linux-pm@...r.kernel.org
>> +S:    Maintained
>> +F:    drivers/thermal/loongson2_thermal.c
>> +
>>   LOONGSON GPIO DRIVER
>>   M:    Yinbo Zhu <zhuyinbo@...ngson.cn>
>>   L:    linux-gpio@...r.kernel.org
>> diff --git a/drivers/thermal/Kconfig b/drivers/thermal/Kconfig
>> index 4cd7ab707315..c4de94e972f4 100644
>> --- a/drivers/thermal/Kconfig
>> +++ b/drivers/thermal/Kconfig
>> @@ -502,4 +502,16 @@ config KHADAS_MCU_FAN_THERMAL
>>         If you say yes here you get support for the FAN controlled
>>         by the Microcontroller found on the Khadas VIM boards.
>> +config LOONGSON2_THERMAL
>> +    tristate "Loongson-2 SoC series thermal driver"
>> +    depends on LOONGARCH || COMPILE_TEST
>> +    depends on OF
>> +    help
>> +      Support for Thermal driver found on Loongson-2 SoC series 
>> platforms.
>> +      The thermal driver realizes get_temp and set_trips function, which
>> +      are used to obtain the temperature of the current node and set the
>> +      temperature range to trigger the interrupt. When the input 
>> temperature
>> +      is higher than the high temperature threshold or lower than the 
>> low
>> +      temperature threshold, the interrupt will occur.
>> +
>>   endif
>> diff --git a/drivers/thermal/Makefile b/drivers/thermal/Makefile
>> index eed300e83d48..4edde4df647b 100644
>> --- a/drivers/thermal/Makefile
>> +++ b/drivers/thermal/Makefile
>> @@ -62,3 +62,4 @@ obj-$(CONFIG_UNIPHIER_THERMAL)    += uniphier_thermal.o
>>   obj-$(CONFIG_AMLOGIC_THERMAL)     += amlogic_thermal.o
>>   obj-$(CONFIG_SPRD_THERMAL)    += sprd_thermal.o
>>   obj-$(CONFIG_KHADAS_MCU_FAN_THERMAL)    += khadas_mcu_fan.o
>> +obj-$(CONFIG_LOONGSON2_THERMAL)    += loongson2_thermal.o
>> diff --git a/drivers/thermal/loongson2_thermal.c 
>> b/drivers/thermal/loongson2_thermal.c
>> new file mode 100644
>> index 000000000000..6a338e6e490e
>> --- /dev/null
>> +++ b/drivers/thermal/loongson2_thermal.c
>> @@ -0,0 +1,166 @@
>> +// SPDX-License-Identifier: GPL-2.0+
>> +/*
>> + * Author: zhanghongchen <zhanghongchen@...ngson.cn>
>> + *         Yinbo Zhu <zhuyinbo@...ngson.cn>
>> + * Copyright (C) 2022-2023 Loongson Technology Corporation Limited
>> + */
>> +
>> +#include <linux/cpufreq.h>
> 
> Is it really needed ?


okay, I will check it.

> 
>> +#include <linux/delay.h>
> 
> Is it really needed ?


okay, I will check it.

> 
>> +#include <linux/interrupt.h>
>> +#include <linux/module.h>
>> +#include <linux/platform_device.h>
>> +#include <linux/io.h>
>> +#include <linux/of_device.h>
>> +#include <linux/thermal.h>
>> +#include "thermal_hwmon.h"
>>
>> +#define LOONGSON2_SOC_MAX_SENSOR_NUM            4
>> +
>> +#define LOONGSON2_TSENSOR_CTRL_HI            0x0
>> +#define LOONGSON2_TSENSOR_CTRL_LO            0x8
>> +#define LOONGSON2_TSENSOR_STATUS            0x10
>> +#define LOONGSON2_TSENSOR_OUT                0x14
> 
> Please use BIT() macros


okay, I got it.

> 
>> +struct loongson2_thermal_data {
>> +    int id;
>> +    void __iomem *regs;
>> +};
> 
> It would make sense to replace the id by the precomputed register offset 
> value.
> 
>> +static int loongson2_thermal_set(struct loongson2_thermal_data *data,
>> +                    int low, int high, bool enable)
>> +{
>> +    u64 reg_ctrl = 0;
>> +    int reg_off = data->id * 2;
>> +
>> +    if (low > high)
>> +        return -EINVAL;
> 
> I don't think this is supposed to happen


okay, I will remove it.

> 
>> +    low = max(low, -40);
>> +    high = min(high, 125);
> 
> You can use clamp() alternatively (your call)


okay, I got it, I will use it.

> 
>> +
>> +    low += 100;
>> +    high += 100;
>> +
>> +    reg_ctrl = low;
>> +    reg_ctrl |= enable ? 0x100 : 0;
>> +    writew(reg_ctrl, data->regs + LOONGSON2_TSENSOR_CTRL_LO + reg_off);
>> +
>> +    reg_ctrl = high;
>> +    reg_ctrl |= enable ? 0x100 : 0;
>> +    writew(reg_ctrl, data->regs + LOONGSON2_TSENSOR_CTRL_HI + reg_off);
> 
> Is the 'enable' boolean really useful?


Yes, this 'enable' was to enable thermal irq.

> 
> Wouldn't be the sensor trip points disabled by default at reset time?
> 


Only here will thermal irq be enabled throughout the entire driver, and
actual testing has shown that interrupts are valid, so this is
meaningful.

> If it is the case then we can get ride of this variable and make the 
> routine simpler
> 
>> +    return 0;
>> +}
>> +
>> +static int loongson2_thermal_get_temp(struct thermal_zone_device *tz, 
>> int *temp)
>> +{
>> +    u32 reg_val;
>> +    struct loongson2_thermal_data *data = tz->devdata;
>> +
>> +    reg_val = readl(data->regs + LOONGSON2_TSENSOR_OUT);
> 
> Seems like there is no offset for the sensor id here ?


There is no need for a sensor ID here.

There are some things that I didn't describe clearly, which made you
misunderstand. Actually, the temperature sensor of 2K1000 is like this:

There are 4 sets of temperature interrupt controllers, only one set of
temperature sampling registers. a sets of temperature interrupt
controllers was considered a sensor, which sensor include 3 register as
follows, where "SEL" represents which sensor is referenced, In 2k1000
datasheet, which "SEL" must be 0.

1) GATE:
2) EN:
3) SEL:

> 
>> +    *temp = ((reg_val & 0xff) - 100) * 1000;
> 
> Please use macros in units.h for conversion and change literals by macro


okay, I got it.

> 
>> +    return 0;
>> +}
>> +
>> +static irqreturn_t loongson2_thermal_irq_thread(int irq, void *dev)
>> +{
>> +    struct thermal_zone_device *tzd = dev;
>> +    struct loongson2_thermal_data *data = tzd->devdata;
> 
> The init function tells me whatever the sensor triggering the interrupt 
> we update the same thermal zone, not necessarily the one which reach the 
> trip point.


Although there are four sets of registers, but, only one set temperature
interrupt controllers was valid in 2k1000. and the condition for
interruption is that the temperature reaches trip point.

> 
>> +    /* clear interrupt */
>> +    writeb(0x3, data->regs + LOONGSON2_TSENSOR_STATUS);
> 
> Replace the literal value by a macro.


okay, I got it .

> 
>> +    thermal_zone_device_update(tzd, THERMAL_EVENT_UNSPECIFIED);
>> +
>> +    return IRQ_HANDLED;
>> +}
>> +
>> +static int loongson2_thermal_set_trips(struct thermal_zone_device 
>> *tz, int low, int high)
>> +{
>> +    struct loongson2_thermal_data *data = tz->devdata;
>> +
>> +    return loongson2_thermal_set(data, low/1000, high/1000, true);
> 
> Please use macros in units.h for conversion


okay, I got it.

> 
>> +}
>> +
>> +static const struct thermal_zone_device_ops loongson2_of_thermal_ops = {
>> +    .get_temp = loongson2_thermal_get_temp,
>> +    .set_trips = loongson2_thermal_set_trips,
>> +};
>> +
>> +static int loongson2_thermal_probe(struct platform_device *pdev)
>> +{
>> +    struct device *dev = &pdev->dev;
>> +    struct resource *res;
>> +    struct loongson2_thermal_data *data;
>> +    struct thermal_zone_device *tzd;
>> +    int ret, irq, i;
>> +
>> +    data = devm_kzalloc(dev, sizeof(*data), GFP_KERNEL);
>> +    if (!data)
>> +        return -ENOMEM;
>> +
>> +    data->id = (uintptr_t)device_get_match_data(dev);
> 
> The changelog says there is up to 4 sensors on the controller but the 
> initialization code seems to handle only one.
> 
> Either you skip this data at all and let the loop to fill the id when 
> the thermal zone successfully registered or you add a eg. a mask with 
> the sensor (eg. 0x3 for sensor 0 and 1) and use those value in the loop.


At present, this code seems appropriate because although 2k1000 has four
sets of temperature interrupt controllers, only the 0th set is valid.

> 
> 
>> +    res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>> +    data->regs = devm_ioremap(dev, res->start, resource_size(res));
>> +    if (IS_ERR(data->regs))
>> +        return PTR_ERR(data->regs);
>> +
>> +    /* get irq */
>> +    irq = platform_get_irq(pdev, 0);
>> +    if (irq < 0)
>> +        return irq;
>> +
>> +    if (data->id > LOONGSON2_SOC_MAX_SENSOR_NUM - 1 || data->id < 0) {
>> +        dev_err(dev, "sensor id error,must be in <0 ~ %d>\n",
>> +                LOONGSON2_SOC_MAX_SENSOR_NUM - 1);
>> +        return -EINVAL;
>> +    }
> 
> As this is coming from a static initialization, this test is not needed.


okay, I got it.

> 
>> +    writeb(0xff, data->regs + LOONGSON2_TSENSOR_STATUS);
> 
> Replace the literal value by a macro


okay, I got it.

> 
>> +    loongson2_thermal_set(data, 0, 0, false);
>> +
>> +    for (i = 0; i < LOONGSON2_SOC_MAX_SENSOR_NUM; i++) {
>> +        tzd = devm_thermal_of_zone_register(dev, i, data,
>> +            &loongson2_of_thermal_ops);
>> +
>> +        if (!IS_ERR(tzd))
>> +            break;
> 
> Assuming the loop aims to register all the supported thermal zones, the 
> conditions should be inverted.


This code seems appropriate, the reason is as above.

> 
>> +        if (PTR_ERR(tzd) != ENODEV)
>> +            continue;
>  >
>> +        return dev_err_probe(dev, PTR_ERR(tzd), "failed to register");
>> +    }
>> +
>> +    ret = devm_request_threaded_irq(dev, irq, NULL, 
>> loongson2_thermal_irq_thread,
>> +            IRQF_ONESHOT, "loongson2_thermal", tzd);
>> +    if (ret < 0)
>> +        return dev_err_probe(dev, ret, "failed to request alarm irq\n")
> 
> So there is one interrupt but 4 sensors. I assume there is one interrupt 
> per controller but I don't see in the interrupt handler routine the code 
> to figure out which sensor triggered the interrupt.


This code seems appropriate, the reason is as above.

> 
>> +    if (devm_thermal_add_hwmon_sysfs(dev, tzd))
>> +        dev_warn(dev, "Failed to add hwmon sysfs attributes\n");
>> +
>> +    return 0;
>> +}
>> +
>> +static const struct of_device_id of_loongson2_thermal_match[] = {
>> +    { .compatible = "loongson,ls2k-thermal", .data = (const void *)0},
> 
> May be better to replace that with a structure pointer but it is up to 
> you to decide


okay, I got it.


Thanks,
Yinbo

Powered by blists - more mailing lists