Fan fault reading/clearing works fine.
Tested-by: Václav Kubernát <kubernat@xxxxxxxxx>
st 26. 5. 2021 v 17:41 odesílatel Guenter Roeck <linux@xxxxxxxxxxxx> napsal:
>
> Fault bits in MAX31790 are sticky and have to be cleared explicitly.
> A write operation into either the 'Target Duty Cycle' register or the
> 'Target Count' register is necessary to clear a fault.
>
> At the same time, we can never clear cached fault status values before
> reading them because the companion fault status for any given fan is
> cleared as well when clearing a fault.
>
> Cc: Jan Kundrát <jan.kundrat@xxxxxxxxx>
> Cc: Václav Kubernát <kubernat@xxxxxxxxx>
> Signed-off-by: Guenter Roeck <linux@xxxxxxxxxxxx>
> ---
> drivers/hwmon/max31790.c | 16 +++++++++++++++-
> 1 file changed, 15 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/hwmon/max31790.c b/drivers/hwmon/max31790.c
> index 19651feb40fb..8e4fd9b7c889 100644
> --- a/drivers/hwmon/max31790.c
> +++ b/drivers/hwmon/max31790.c
> @@ -80,7 +80,7 @@ static struct max31790_data *max31790_update_device(struct device *dev)
> MAX31790_REG_FAN_FAULT_STATUS1);
> if (rv < 0)
> goto abort;
> - data->fault_status = rv & 0x3F;
> + data->fault_status |= rv & 0x3F;
>
> rv = i2c_smbus_read_byte_data(client,
> MAX31790_REG_FAN_FAULT_STATUS2);
> @@ -184,7 +184,21 @@ static int max31790_read_fan(struct device *dev, u32 attr, int channel,
> *val = rpm;
> return 0;
> case hwmon_fan_fault:
> + mutex_lock(&data->update_lock);
> *val = !!(data->fault_status & (1 << channel));
> + data->fault_status &= ~(1 << channel);
> + /*
> + * If a fault bit is set, we need to write into one of the fan
> + * configuration registers to clear it. Note that this also
> + * clears the fault for the companion channel if enabled.
> + */
> + if (*val) {
> + int reg = MAX31790_REG_TARGET_COUNT(channel % NR_CHANNEL);
> +
> + i2c_smbus_write_byte_data(data->client, reg,
> + data->target_count[channel % NR_CHANNEL] >> 8);
> + }
> + mutex_unlock(&data->update_lock);
> return 0;
> default:
> return -EOPNOTSUPP;
> --
> 2.25.1
>
