Hello.
I agree that it makes sense to swap yes_ranges and no_ranges. I tested
that, but it seems like it doesn't have an effect, I could still see
fan*_input changing (that's where I don't want caching). Does caching
work automatically? As in, all registers are cached by default in
regmap, and only registers that are in the volatile yes_ranges aren't?
Václav
čt 22. 4. 2021 v 3:31 odesílatel Guenter Roeck <linux@xxxxxxxxxxxx> napsal:
>
> On 4/12/21 7:59 PM, Václav Kubernát wrote:
> > Converting the driver to use regmap makes it more generic. It also makes
> > it a lot easier to debug through debugfs.
> >
> > Signed-off-by: Václav Kubernát <kubernat@xxxxxxxxx>
> > ---
> > drivers/hwmon/Kconfig | 1 +
> > drivers/hwmon/max31790.c | 254 ++++++++++++++++++++-------------------
> > 2 files changed, 133 insertions(+), 122 deletions(-)
> >
> > diff --git a/drivers/hwmon/Kconfig b/drivers/hwmon/Kconfig
> > index 1ecf697d8d99..9f11d036c316 100644
> > --- a/drivers/hwmon/Kconfig
> > +++ b/drivers/hwmon/Kconfig
> > @@ -1095,6 +1095,7 @@ config SENSORS_MAX6697
> > config SENSORS_MAX31790
> > tristate "Maxim MAX31790 sensor chip"
> > depends on I2C
> > + select REGMAP_I2C
> > help
> > If you say yes here you get support for 6-Channel PWM-Output
> > Fan RPM Controller.
> > diff --git a/drivers/hwmon/max31790.c b/drivers/hwmon/max31790.c
> > index 2c6b333a28e9..e3765ce4444a 100644
> > --- a/drivers/hwmon/max31790.c
> > +++ b/drivers/hwmon/max31790.c
> > @@ -12,6 +12,7 @@
> > #include <linux/init.h>
> > #include <linux/jiffies.h>
> > #include <linux/module.h>
> > +#include <linux/regmap.h>
> > #include <linux/slab.h>
> >
> > /* MAX31790 registers */
> > @@ -46,92 +47,53 @@
> >
> > #define NR_CHANNEL 6
> >
> > +#define MAX31790_REG_USER_BYTE_67 0x67
> > +
> > +#define BULK_TO_U16(msb, lsb) (((msb) << 8) + (lsb))
> > +#define U16_MSB(num) (((num) & 0xFF00) >> 8)
> > +#define U16_LSB(num) ((num) & 0x00FF)
> > +
> > +static const struct regmap_range max31790_ro_range = {
> > + .range_min = MAX31790_REG_TACH_COUNT(0),
> > + .range_max = MAX31790_REG_PWMOUT(0) - 1,
> > +};
> > +
> > +static const struct regmap_access_table max31790_wr_table = {
> > + .no_ranges = &max31790_ro_range,
> > + .n_no_ranges = 1,
> > +};
> > +
> > +static const struct regmap_range max31790_volatile_ranges[] = {
> > + regmap_reg_range(MAX31790_REG_TACH_COUNT(0), MAX31790_REG_TACH_COUNT(12)),
> > + regmap_reg_range(MAX31790_REG_FAN_FAULT_STATUS2, MAX31790_REG_FAN_FAULT_STATUS1),
> > +};
> > +
> > +static const struct regmap_access_table max31790_volatile_table = {
> > + .no_ranges = max31790_volatile_ranges,
> > + .n_no_ranges = 2,
> > + .n_yes_ranges = 0
> > +};
>
> Looks like my reply to this got lost. Other regmap code suggests that
> volatile register ranges are identified with yes_ranges, not with no_ranges.
> "no" seems to mean "not volatile". Please verify and confirm if the
> above code does what you want it to do.
>
> Thanks,
> Guenter
