On Sat, Jul 06, 2024 at 07:52:47AM GMT, Guenter Roeck wrote:
> On 7/2/24 19:30, Inochi Amaoto wrote:
> > SG2042 use an external MCU to provide basic hardware information
> > and thermal sensors.
> >
> > Add driver support for the onboard MCU of SG2042.
> >
> > Signed-off-by: Inochi Amaoto <inochiama@xxxxxxxxxxx>
> > ---
> > Documentation/hwmon/index.rst | 1 +
> > Documentation/hwmon/sgmcu.rst | 44 +++
> > drivers/hwmon/Kconfig | 11 +
> > drivers/hwmon/Makefile | 1 +
> > drivers/hwmon/sgmcu.c | 585 ++++++++++++++++++++++++++++++++++
> > 5 files changed, 642 insertions(+)
> > create mode 100644 Documentation/hwmon/sgmcu.rst
> > create mode 100644 drivers/hwmon/sgmcu.c
> >
> > diff --git a/Documentation/hwmon/index.rst b/Documentation/hwmon/index.rst
> > index 03d313af469a..189626b3a055 100644
> > --- a/Documentation/hwmon/index.rst
> > +++ b/Documentation/hwmon/index.rst
> > @@ -203,6 +203,7 @@ Hardware Monitoring Kernel Drivers
> > sch5636
> > scpi-hwmon
> > sfctemp
> > + sgmcu
> > sht15
> > sht21
> > sht3x
> > diff --git a/Documentation/hwmon/sgmcu.rst b/Documentation/hwmon/sgmcu.rst
> > new file mode 100644
> > index 000000000000..5669dcfb2a33
> > --- /dev/null
> > +++ b/Documentation/hwmon/sgmcu.rst
> > @@ -0,0 +1,44 @@
> > +.. SPDX-License-Identifier: GPL-2.0
> > +
> > +Kernel driver sgmcu
> > +=====================
> > +
> > +Supported chips:
> > +
> > + * Onboard MCU for sg2042
> > +
> > + Addresses scanned: -
> > +
> > + Prefix: 'sgmcu'
> > +
> > +Authors:
> > +
> > + - Inochi Amaoto <inochiama@xxxxxxxxxxx>
> > +
> > +Description
> > +-----------
> > +
> > +This driver supprts hardware monitoring for onboard MCU with
> > +PMBus interface.
> > +
> > +Usage Notes
> > +-----------
> > +
> > +This driver does not auto-detect devices. You will have to instantiate
> > +the devices explicitly.
> > +Please see Documentation/i2c/instantiating-devices.rst for details.
> > +
> > +Platform data support
> > +---------------------
> > +
> > +The driver supports standard PMBus driver platform data.
> > +
>
> It doesn't. Where does PMBus come into play ? Is this a cut-and-paste problem ?
>
Yes, It should be smbus/i2c, I forgot to modify.
> > +Sysfs Attributes
> > +----------------
> > +
> > +================= =============================================
> > +temp1_input Measured temperature of SoC
> > +temp1_crit Critical high temperature
> > +temp1_crit_hyst hysteresis temperature restore from Critical
> > +temp2_input Measured temperature of the base board
> > +================= =============================================
> > diff --git a/drivers/hwmon/Kconfig b/drivers/hwmon/Kconfig
> > index e14ae18a973b..1100dd11f7f5 100644
> > --- a/drivers/hwmon/Kconfig
> > +++ b/drivers/hwmon/Kconfig
> > @@ -2079,6 +2079,17 @@ config SENSORS_SFCTEMP
> > This driver can also be built as a module. If so, the module
> > will be called sfctemp.
> >
> > +config SENSORS_SGMCU
> > + tristate "Sophgo onboard MCU support"
> > + depends on I2C
> > + depends on ARCH_SOPHGO || COMPILE_TEST
> > + help
> > + Support for onboard MCU of Sophgo SoCs. This mcu provides power
> > + control and some basic information.
> > +
> > + This driver can be built as a module. If so, the module
> > + will be called sgmcu.
> > +
> > config SENSORS_SURFACE_FAN
> > tristate "Surface Fan Driver"
> > depends on SURFACE_AGGREGATOR
> > diff --git a/drivers/hwmon/Makefile b/drivers/hwmon/Makefile
> > index e3f25475d1f0..e9b78ff8338e 100644
> > --- a/drivers/hwmon/Makefile
> > +++ b/drivers/hwmon/Makefile
> > @@ -195,6 +195,7 @@ obj-$(CONFIG_SENSORS_SCH56XX_COMMON)+= sch56xx-common.o
> > obj-$(CONFIG_SENSORS_SCH5627) += sch5627.o
> > obj-$(CONFIG_SENSORS_SCH5636) += sch5636.o
> > obj-$(CONFIG_SENSORS_SFCTEMP) += sfctemp.o
> > +obj-$(CONFIG_SENSORS_SGMCU) += sgmcu.o
> > obj-$(CONFIG_SENSORS_SL28CPLD) += sl28cpld-hwmon.o
> > obj-$(CONFIG_SENSORS_SHT15) += sht15.o
> > obj-$(CONFIG_SENSORS_SHT21) += sht21.o
> > diff --git a/drivers/hwmon/sgmcu.c b/drivers/hwmon/sgmcu.c
> > new file mode 100644
> > index 000000000000..d941d6fe741f
> > --- /dev/null
> > +++ b/drivers/hwmon/sgmcu.c
> > @@ -0,0 +1,585 @@
> > +// SPDX-License-Identifier: GPL-2.0-only
> > +/*
> > + * Copyright (c) 2024 Inochi Amaoto <inochiama@xxxxxxxxxxx>
> > + *
> > + * Sophgo power control mcu for SG2042
> > + */
> > +
> > +#include <linux/debugfs.h>
> > +#include <linux/kernel.h>
> > +#include <linux/module.h>
> > +#include <linux/i2c.h>
> > +#include <linux/err.h>
> > +#include <linux/hwmon.h>
> > +
> Alphabetic include file order, please.
>
> > +/* fixed MCU registers */
> > +#define REG_BOARD_TYPE 0x00
> > +#define REG_MCU_FIRMWARE_VERSION 0x01
> > +#define REG_PCB_VERSION 0x02
> > +#define REG_PWR_CTRL 0x03
> > +#define REG_SOC_TEMP 0x04
> > +#define REG_BOARD_TEMP 0x05
> > +#define REG_RST_COUNT 0x0a
> > +#define REG_UPTIME 0x0b
> > +#define REG_RESET_REASON 0x0d
> > +#define REG_MCU_TYPE 0x18
> > +#define REG_CRITICAL_ACTIONS 0x65
> > +#define REG_CRITICAL_TEMP 0x66
> > +#define REG_REPOWER_TEMP 0x67
> > +
> > +#define CRITICAL_ACTION_REBOOT 0x1
> > +#define CRITICAL_ACTION_POWEROFF 0x2
> > +
>
> Please use BIT() for bit masks.
>
> > +#define DEFAULT_REPOWER_TEMP 60
> > +#define MAX_REPOWER_TEMP 100
> > +
> > +#define sg2042_mcu_read_byte(client, reg) \
> > + i2c_smbus_read_byte_data(client, reg)
> > +#define sg2042_mcu_write_byte(client, reg, value) \
> > + i2c_smbus_write_byte_data(client, reg, value)
> > +#define sg2042_mcu_read_block(client, reg, array) \
> > + i2c_smbus_read_i2c_block_data(client, reg, sizeof(array), array)
> > +
>
> Pointless defines. See below.
>
> > +#define DEFINE_MCU_ATTR_READ_FUNC(_name, _type, _format) \
> > + static ssize_t _name##_show(struct device *dev, \
> > + struct device_attribute *attr, \
> > + char *buf) \
> > + { \
> > + struct sg2042_mcu_data *mcu = dev_get_drvdata(dev); \
> > + _type ret; \
> > + ret = sg2042_mcu_get_##_name(mcu->client); \
> > + if (ret < 0) \
> > + return ret; \
> > + return sprintf(buf, _format "\n", ret); \
> > + }
> > +
> > +#define DEFINE_MCU_DEBUG_ATTR_READ_FUNC(_name, _type, _format) \
> > + static int _name##_show(struct seq_file *seqf, \
> > + void *unused) \
> > + { \
> > + struct sg2042_mcu_data *mcu = seqf->private; \
> > + _type ret; \
> > + ret = sg2042_mcu_get_##_name(mcu->client); \
> > + if (ret < 0) \
> > + return ret; \
> > + seq_printf(seqf, _format "\n", ret); \
> > + return 0; \
> > + }
> > +
> > +#define _CREATE_DEBUG_ENTRY(name, perm, d, data) \
> > + debugfs_create_file(#name, perm, d, data, &name##_fops)
> > +
>
> Pointless define. Please do not replace API function names with your own;
> that just makes it harder to find callers.
>
> > +struct sg2042_mcu_board_data {
> > + u8 id;
> > + const char *name;
> > +};
> > +
> > +struct sg2042_mcu_data {
> > + struct i2c_client *client;
> > + const struct sg2042_mcu_board_data *board_info;
> > + struct dentry *debugfs;
> > +};
> > +
> > +static const struct sg2042_mcu_board_data sg2042_boards_data[] = {
> > + {
> > + .id = 0x80,
> > + .name = "SG2042 evb x8",
> > + },
> > + {
> > + .id = 0x81,
> > + .name = "SG2042R evb",
> > + },
> > + {
> > + .id = 0x83,
> > + .name = "SG2042 evb x4",
> > + },
> > + {
> > + .id = 0x90,
> > + .name = "Milk-V Pioneer",
> > + },
> > +};
> > +
> > +static const char *sg2042_mcu_reset_reason[8] = {
> > + "Power supply overheat",
> > + "Power supply failure",
> > + "12V power supply failure",
> > + "Reset commant",
>
> command
>
> > + "Unknown",
> > + "Unknown",
> > + "Unknown",
> > + "SoC overheat",
> > +};
> > +
> > +static struct dentry *sgmcu_debugfs;
> > +
> > +static int sg2042_mcu_get_board_type(struct i2c_client *client)
> > +{
> > + return sg2042_mcu_read_byte(client, REG_BOARD_TYPE);
> > +}
> > +
> > +static int sg2042_mcu_get_firmware_version(struct i2c_client *client)
> > +{
> > + return sg2042_mcu_read_byte(client, REG_MCU_FIRMWARE_VERSION);
> > +}
> > +
> > +static int sg2042_mcu_get_pcb_version(struct i2c_client *client)
> > +{
> > + return sg2042_mcu_read_byte(client, REG_PCB_VERSION);
> > +}
> > +
> > +static int sg2042_mcu_get_soc_temp(struct i2c_client *client)
> > +{
> > + return sg2042_mcu_read_byte(client, REG_SOC_TEMP);
> > +}
> > +
> > +static int sg2042_mcu_get_board_temp(struct i2c_client *client)
> > +{
> > + return sg2042_mcu_read_byte(client, REG_BOARD_TEMP);
> > +}
> > +
> > +static int sg2042_mcu_get_reset_count(struct i2c_client *client)
> > +{
> > + return sg2042_mcu_read_byte(client, REG_RST_COUNT);
> > +}
> > +
>
> Those one-line functions are just as pointless.
> i2c_smbus_read_byte_data(client, REG_RST_COUNT);
> does exactly the same without extra function and define.
>
> > +static s32 sg2042_mcu_get_uptime(struct i2c_client *client)
> > +{
> > + int ret;
> > + u8 time_val[2];
> > +
> > + ret = sg2042_mcu_read_block(client, REG_UPTIME, time_val);
> > + if (ret < 0)
> > + return ret;
> > +
> > + return (s32)(time_val[0]) + ((s32)(time_val[1]) << 8);
> > +}
> > +
>
> This is the only one of those functions which adds at least some value.
>
> > +static int sg2042_mcu_get_reset_reason(struct i2c_client *client)
> > +{
> > + return sg2042_mcu_read_byte(client, REG_RESET_REASON);
> > +}
> > +
> > +static int sg2042_mcu_get_mcu_type(struct i2c_client *client)
> > +{
> > + return sg2042_mcu_read_byte(client, REG_MCU_TYPE);
> > +}
> > +
> > +static int sg2042_mcu_get_soc_crit_action(struct i2c_client *client)
> > +{
> > + return sg2042_mcu_read_byte(client, REG_CRITICAL_ACTIONS);
> > +}
> > +
> > +static int sg2042_mcu_get_soc_crit_temp(struct i2c_client *client)
> > +{
> > + return sg2042_mcu_read_byte(client, REG_CRITICAL_TEMP);
> > +}
> > +
> > +static int sg2042_mcu_get_soc_hyst_temp(struct i2c_client *client)
> > +{
> > + return sg2042_mcu_read_byte(client, REG_REPOWER_TEMP);
> > +}
> > +
> > +static int sg2042_mcu_set_soc_crit_action(struct i2c_client *client,
> > + u8 value)
> > +{
> > + return sg2042_mcu_write_byte(client, REG_CRITICAL_ACTIONS, value);
> > +}
> > +
> > +static int sg2042_mcu_set_soc_crit_temp(struct i2c_client *client,
> > + u8 value)
> > +{
> > + return sg2042_mcu_write_byte(client, REG_CRITICAL_TEMP, value);
> > +}
> > +
> > +static int sg2042_mcu_set_soc_hyst_temp(struct i2c_client *client,
> > + u8 value)
> > +{
> > + return sg2042_mcu_write_byte(client, REG_REPOWER_TEMP, value);
> > +}
> > +
> > +DEFINE_MCU_ATTR_READ_FUNC(reset_count, int, "%d");
> > +DEFINE_MCU_ATTR_READ_FUNC(uptime, s32, "%d");
> > +
> > +static ssize_t reset_reason_show(struct device *dev,
> > + struct device_attribute *attr,
> > + char *buf)
> > +{
> > + struct sg2042_mcu_data *mcu = dev_get_drvdata(dev);
> > + int ret, val, i;
> > +
> > + val = sg2042_mcu_get_reset_reason(mcu->client);
> > + if (val < 0)
> > + return val;
> > +
> > + ret = sprintf(buf, "Reason: 0x%02x\n", val);
> > +
> > + for (i = 0; i < ARRAY_SIZE(sg2042_mcu_reset_reason); i++) {
> > + if (val & BIT(i))
> > + ret += sprintf(buf + ret, "bit %d: %s\n", i,
> > + sg2042_mcu_reset_reason[i]);
> > + }
> > +
> > + return ret;
> > +}
>
> This violates sysfs standards (one word only). If you want that level of detail,
> add it to debugfs.
>
I think it would be better to change name and leave it in the sysfs.
If a suitable world is not found, I will move it to debugfs.
> > +
> > +static ssize_t critical_action_show(struct device *dev,
> > + struct device_attribute *attr,
> > + char *buf)
> > +{
> > + struct sg2042_mcu_data *mcu = dev_get_drvdata(dev);
> > + int ret;
> > + const char *action;
> > +
> > + ret = sg2042_mcu_get_soc_crit_action(mcu->client);
> > + if (ret < 0)
> > + return ret;
> > +
> > + if (ret == CRITICAL_ACTION_REBOOT)
> > + action = "reboot";
> > + else if (ret == CRITICAL_ACTION_POWEROFF)
> > + action = "poweroff";
> > + else
> > + action = "unknown";
> > +
> > + return sprintf(buf, "%s\n", action);
> > +}
> > +
> > +static ssize_t critical_action_store(struct device *dev,
> > + struct device_attribute *attr,
> > + const char *buf, size_t count)
> > +{
> > + struct sg2042_mcu_data *mcu = dev_get_drvdata(dev);
> > + int value;
> > +
> > + if (sysfs_streq("reboot", buf))
> > + value = CRITICAL_ACTION_REBOOT;
> > + else if (sysfs_streq("poweroff", buf))
> > + value = CRITICAL_ACTION_POWEROFF;
> > + else
> > + return -EINVAL;
> > +
> > + return sg2042_mcu_set_soc_crit_action(mcu->client, value);
> > +}
> > +
> > +static DEVICE_ATTR_RO(reset_count);
> > +static DEVICE_ATTR_RO(uptime);
> > +static DEVICE_ATTR_RO(reset_reason);
> > +static DEVICE_ATTR_RW(critical_action);
> > +
> > +DEFINE_MCU_DEBUG_ATTR_READ_FUNC(firmware_version, int, "0x%02x");
> > +DEFINE_MCU_DEBUG_ATTR_READ_FUNC(pcb_version, int, "0x%02x");
> > +
> > +static int board_type_show(struct seq_file *seqf, void *unused)
> > +{
> > + struct sg2042_mcu_data *mcu = seqf->private;
> > +
> > + seq_printf(seqf, "%s\n", mcu->board_info->name ?: "Unknown");
> > +
> > + return 0;
> > +}
> > +
> > +static int mcu_type_show(struct seq_file *seqf, void *unused)
> > +{
> > + struct sg2042_mcu_data *mcu = seqf->private;
> > + int ret;
> > +
> > + ret = sg2042_mcu_get_mcu_type(mcu->client);
> > + if (ret < 0)
> > + return ret;
> > +
> > + seq_puts(seqf, ret ? "GD32\n" : "STM32\n");
> > +
> > + return 0;
> > +}
> > +
> > +DEFINE_SHOW_ATTRIBUTE(firmware_version);
> > +DEFINE_SHOW_ATTRIBUTE(pcb_version);
> > +DEFINE_SHOW_ATTRIBUTE(mcu_type);
> > +DEFINE_SHOW_ATTRIBUTE(board_type);
> > +
> > +// TODO: to debugfs
> > +
>
> If there is a TODO left, move the driver to drivers/staging/
> and keep it there until it is complete.
>
My fault, it is already finished and can be removed.
> > +static struct attribute *sg2042_mcu_attrs[] = {
> > + &dev_attr_reset_count.attr,
> > + &dev_attr_uptime.attr,
> > + &dev_attr_reset_reason.attr,
> > + &dev_attr_critical_action.attr,
> > + NULL
> > +};
> > +
> > +static const struct attribute_group sg2042_mcu_attr_group = {
> > + .attrs = sg2042_mcu_attrs,
> > +};
> > +
> > +static const struct hwmon_channel_info * const sg2042_mcu_info[] = {
> > + HWMON_CHANNEL_INFO(chip, HWMON_C_REGISTER_TZ | HWMON_C_UPDATE_INTERVAL),
> > + HWMON_CHANNEL_INFO(temp, HWMON_T_INPUT | HWMON_T_CRIT |
> > + HWMON_T_CRIT_HYST,
> > + HWMON_T_INPUT),
> > + NULL
> > +};
> > +
> > +static int sg2042_mcu_read_temp(struct device *dev,
> > + u32 attr, int channel,
> > + long *val)
> > +{
> > + struct sg2042_mcu_data *mcu = dev_get_drvdata(dev);
> > + long tmp;
>
> Why long ?
>
just use to calculate the value.
> > +
> > + switch (attr) {
> > + case hwmon_temp_input:
> > + switch (channel) {
> > + case 0:
> > + tmp = sg2042_mcu_get_soc_temp(mcu->client);
> > + if (tmp < 0)
> > + return tmp;
> > + *val = tmp * 1000;
> > + break;
> > + case 1:
> > + tmp = sg2042_mcu_get_board_temp(mcu->client);
> > + if (tmp < 0)
> > + return tmp;
> > + *val = tmp * 1000;
> > + break;
> > + default:
> > + return -EOPNOTSUPP;
> > + }
>
> This would be much simpler written with something like
> tmp = i2c_smbus_read_byte_data(client, channel ? REG_SOC_TEMP : REG_BOARD_TEMP);
> or even
> reg = channel ? REG_SOC_TEMP : REG_BOARD_TEMP;
> followed by unified read handling below.
>
You are right, thanks.
> > + break;
> > + case hwmon_temp_crit:
> > + if (channel)
> > + return -EOPNOTSUPP;
> > +
>
> Those checks are unnecessary if the is_visible() function
> does its job. If it doesn't, it is broken and needs to be fixed.
> Returning -EOPNOTSUPP is ok for default: cases, because _something_
> needs to be done, but not as result of extra and unnecessary checks
> like this one.
>
Thanks for this explanation. I will change it.
> > + tmp = sg2042_mcu_get_soc_crit_temp(mcu->client);
> > + if (tmp < 0)
> > + return tmp;
> > + *val = tmp * 1000;
> > + break;
> > + case hwmon_temp_crit_hyst:
> > + if (channel)
> > + return -EOPNOTSUPP;
> > +
> > + tmp = sg2042_mcu_get_soc_hyst_temp(mcu->client);
> > + if (tmp < 0)
> > + return tmp;
> > + *val = tmp * 1000;
> > + break;
> > + default:
> > + return -EOPNOTSUPP;
> > + }
>
> There is a lot of repetition here. Setting the register in the switch statement, followed
> by
> tmp = i2c_smbus_read_byte_data(client, reg);
> if (tmp < 0)
> return tmp;
> *val = tmp * 1000;
> would be much simpler.
>
>
> > + return 0;
> > +}
> > +
> > +static int sg2042_mcu_read(struct device *dev,
> > + enum hwmon_sensor_types type,
> > + u32 attr, int channel, long *val)
> > +{
> > + switch (type) {
> > + case hwmon_chip:
> > + if (attr != hwmon_chip_update_interval)
> > + return -EOPNOTSUPP;
> > + *val = 1000;
> > + break;
>
> Pointless attribute. Attributes are only valuable if 1) used and
> 2) written into the chip. This one isn't used for anything.
>
OK. I will remove this.
> > + case hwmon_temp:
> > + return sg2042_mcu_read_temp(dev, attr, channel, val);
> > + default:
> > + return -EOPNOTSUPP;
> > + }
> > + return 0;
> > +}
> > +
> > +static int sg2042_mcu_write(struct device *dev,
> > + enum hwmon_sensor_types type,
> > + u32 attr, int channel, long val)
> > +{
> > + struct sg2042_mcu_data *mcu = dev_get_drvdata(dev);
> > + u8 down_temp, repower_temp;
> > + int ret;
> > +
> > + if (type != hwmon_temp || attr != hwmon_temp_crit || !channel)
> > + return -EOPNOTSUPP;
> > +
> Pointless check if the is_visible() function does its job.
>
> Also, this is inconsistent: It only accepts writes if channel > 0,
> but the is_visible code only makes the attributes available for
> channel 0. On top of that, how does this expession ever allow writing
> hwmon_temp_crit_hyst ?
>
> Either I am missing something, or you did not test this code.
>
This is my fault. Only channel 0 can be writed.
> > + switch (attr) {
> > + case hwmon_temp_crit:
> > + ret = sg2042_mcu_get_soc_hyst_temp(mcu->client);
> > + if (ret < 0)
> > + repower_temp = DEFAULT_REPOWER_TEMP;
>
> Why is this not an error ? If some of the supported boards
> don't support it, the value should not be read in the first place,
> and the hyst attribute should not be there.
>
At least for now, all the chip this driver support can get
hyst_temp. So with your advice, I will issue an error.
> > + else
> > + repower_temp = ret;
> > +
> > + down_temp = val / 1000;
>
> val needs to be range checked.
>
> > + if (down_temp < repower_temp)
> > + return -EINVAL;
> > +
> > + return sg2042_mcu_set_soc_crit_temp(mcu->client,
> > + (u8)(val / 1000));
>
> A read followed by a write needs to be mutex protected because the other value
> could be changed at the same time from another process. I am not sure if those
> checks are valuable or even make sense, but if you want that compexity you'll
> have to add mutex protection.
>
Thanks, this is what I forgot.
> Also, what is the point of recalculating down_temp ?
>
After some real test, I found the sg2042 can not handle the case
when "down_temp < repower_temp", it will keep closing. To avoid
this, I added this check. This is the same for the repower_temp.
> > + case hwmon_temp_crit_hyst:
> > + ret = sg2042_mcu_get_soc_crit_temp(mcu->client);
> > + if (ret < 0)
> > + return -ENODEV;
> > +
>
> Do not overwrite error codes. If the attribute does not exist for some of the
> boards, it should not be created in the first place, and no attempt should be
> made to read it.
>
> > + down_temp = ret;
> > + repower_temp = val / 1000;
>
> val needs to be range checked.
>
> > + if (down_temp < repower_temp)
> > + return -EINVAL;
> > +
> > + return sg2042_mcu_set_soc_hyst_temp(mcu->client,
> > + (u8)(val / 1000));
>
> What is the point of recalculating repower_temp ?
>
> > + default:
> > + return -EOPNOTSUPP;
> > + }
> > +}
> > +
> > +static umode_t sg2042_mcu_is_visible(const void *_data,
> > + enum hwmon_sensor_types type,
> > + u32 attr, int channel)
> > +{
> > + switch (type) {
> > + case hwmon_chip:
> > + if (attr == hwmon_chip_update_interval)
> > + return 0444;
> > + break;
> > + case hwmon_temp:
> > + switch (attr) {
> > + case hwmon_temp_input:
> > + if (channel < 2)
> > + return 0444;
>
> channel is always < 2.
>
> > + break;
> > + case hwmon_temp_crit:
> > + case hwmon_temp_crit_hyst:
> > + if (channel == 0)
> > + return 0664;
> > + break;
> > + default:
> > + return 0;
> > + }
> > + break;
> > + default:
> > + return 0;
> > + }
> > + return 0;
> > +}
> > +
> > +static const struct hwmon_ops sg2042_mcu_ops = {
> > + .is_visible = sg2042_mcu_is_visible,
> > + .read = sg2042_mcu_read,
> > + .write = sg2042_mcu_write,
> > +};
> > +
> > +static const struct hwmon_chip_info sg2042_mcu_chip_info = {
> > + .ops = &sg2042_mcu_ops,
> > + .info = sg2042_mcu_info,
> > +};
> > +
> > +static void sg2042_mcu_debugfs_init(struct sg2042_mcu_data *mcu,
> > + struct device *dev)
> > +{
> > + mcu->debugfs = debugfs_create_dir(dev_name(dev), sgmcu_debugfs);
> > + if (mcu->debugfs) {
> > + _CREATE_DEBUG_ENTRY(firmware_version, 0444, mcu->debugfs, mcu);
> > + _CREATE_DEBUG_ENTRY(pcb_version, 0444, mcu->debugfs, mcu);
> > + _CREATE_DEBUG_ENTRY(mcu_type, 0444, mcu->debugfs, mcu);
> > + _CREATE_DEBUG_ENTRY(board_type, 0444, mcu->debugfs, mcu);
> > + }
> > +}
> > +
> > +static int sg2042_mcu_check_board(u8 id)
> > +{
> > + int i;
> > +
> > + for (i = 0; i < ARRAY_SIZE(sg2042_boards_data); i++) {
> > + if (sg2042_boards_data[i].id == id)
> > + return i;
> > + }
> > +
> > + return -ENODEV;
> > +}
> > +
> > +static int sg2042_mcu_i2c_probe(struct i2c_client *client)
> > +{
> > + int ret;
> > + struct device *dev = &client->dev;
> > + struct sg2042_mcu_data *mcu;
> > + struct device *hwmon_dev;
> > +
> > + if (!i2c_check_functionality(client->adapter, I2C_FUNC_SMBUS_BYTE_DATA |
> > + I2C_FUNC_SMBUS_BLOCK_DATA))
> > + return -EIO;
> > +
> > + ret = sg2042_mcu_get_board_type(client);
> > + if (ret < 0)
> > + return ret;
> > +
> > + ret = sg2042_mcu_check_board(ret);
> > + if (ret < 0)
> > + return ret;
> > +
> > + mcu = devm_kmalloc(dev, sizeof(*mcu), GFP_KERNEL);
> > + if (!mcu)
> > + return -ENOMEM;
> > +
> > + mcu->client = client;
> > + mcu->board_info = &sg2042_boards_data[ret];
> > +
> > + ret = sysfs_create_group(&dev->kobj, &sg2042_mcu_attr_group);
> > + if (ret < 0)
> > + return ret;
> > +
>
> Why not use .dev_groups provided by struct device_driver ?
>
> > + i2c_set_clientdata(client, mcu);
> > +
> > + hwmon_dev = devm_hwmon_device_register_with_info(dev, client->name,
> > + mcu,
> > + &sg2042_mcu_chip_info,
> > + NULL);
> > +
> > + sg2042_mcu_debugfs_init(mcu, dev);
> > +
>
> This leaves the debugfs files orphan if the device (not the driver) is removed.
> Also, it is still created even if devm_hwmon_device_register_with_info() returned
> an error.
>
OK, I forgot this condition, I will add necessary check, thanks.
> > + return PTR_ERR_OR_ZERO(hwmon_dev);
> > +}
> > +
> > +static void sg2042_mcu_i2c_remove(struct i2c_client *client)
> > +{
> > + struct device *dev = &client->dev;
> > +
> > + sysfs_remove_group(&dev->kobj, &sg2042_mcu_attr_group);
> > +}
> > +
> > +static const struct i2c_device_id sg2042_mcu_id[] = {
> > + { "sg2042_hwmon_mcu", 0 },
> > + {},
> > +};
> > +MODULE_DEVICE_TABLE(i2c, sg2042_mcu_id);
> > +
> > +static const struct of_device_id sg2042_mcu_of_id[] = {
> > + { .compatible = "sophgo,sg2042-hwmon-mcu" },
>
> Based on sg2042_boards_data, this seems wrong. Devicetree data should distinguish the supported
> boards. That is its whole point. The same is true for sg2042_mcu_id; there should be separate
> entries for each of the supported boards.
>
> > + {},
> > +};
> > +MODULE_DEVICE_TABLE(of, sg2042_mcu_of_id);
> > +
> > +static struct i2c_driver sg2042_mcu_driver = {
> > + .driver = {
> > + .name = "sg2042-mcu",
> > + .of_match_table = sg2042_mcu_of_id,
> > + },
> > + .probe = sg2042_mcu_i2c_probe,
> > + .remove = sg2042_mcu_i2c_remove,
> > + .id_table = sg2042_mcu_id,
> > +};
> > +
> > +static int __init sg2042_mcu_init(void)
> > +{
> > + sgmcu_debugfs = debugfs_create_dir("sgmcu", NULL);
> > + return i2c_add_driver(&sg2042_mcu_driver);
> > +}
> > +
> > +static void __exit sg2042_mcu_exit(void)
> > +{
> > + debugfs_remove_recursive(sgmcu_debugfs);
> > + i2c_del_driver(&sg2042_mcu_driver);
> > +}
> > +
> > +module_init(sg2042_mcu_init);
> > +module_exit(sg2042_mcu_exit);
> > +
> > +MODULE_AUTHOR("Inochi Amaoto <inochiama@xxxxxxxxxxx>");
> > +MODULE_DESCRIPTION("MCU I2C driver for SG2042 soc platform");
> > +MODULE_LICENSE("GPL");
> > --
> > 2.45.2
> >
> >
>
